{ "name": "Content Lexicon", "url": "https://contentlexicon.com", "publisher": "Contentious", "license": { "name": "CC BY 4.0", "url": "https://creativecommons.org/licenses/by/4.0/", "terms": "The text of every entry is published under CC BY 4.0: quote it, share it and adapt it, as long as you credit the Lexicon. The drawings aren't covered." }, "summary": "https://contentlexicon.com/lexicon.json", "count": 167, "updated": "2026-10-20T15:43:00.000Z", "terms": [ { "term": "A/B testing", "alsoCalled": [ "Split testing" ], "slug": "a-b-testing", "definition": "A way to compare two or more live variants with users, useful when the question is specific and the measure actually matters.", "category": "Research and evidence", "related": [ "branch" ], "updated": "2026-10-11T06:23:00.000Z", "url": "https://contentlexicon.com/a-b-testing", "markdown": "https://contentlexicon.com/a-b-testing.md", "credit": "A/B testing, from the Content Lexicon by Contentious (https://contentlexicon.com/a-b-testing), CC BY 4.0", "body": "In an A/B test, one group might see version A of a page, [email](https://contentlexicon.com/email) or journey while another sees version B. The team then looks at what people actually did, rather than relying only on opinion, preference or whoever was most confident in the meeting.\n\nFor content work, A/B testing can be useful when the question is specific, the audience is large enough and the team knows what kind of evidence would be meaningful. It can help compare [calls to action](https://contentlexicon.com/call-to-action), page structures, subject lines, onboarding flows or different ways of explaining a choice.\n\nThe trap is thinking that A/B testing automatically tells you what is better. It usually tells you what performed better under particular conditions, for a particular audience, against a particular metric. If the metric is too narrow, the test may reward worse content: more clicks, less understanding; more sign-ups, more regret; more attention, less trust.\n\n## Why it matters for content\n\nA/B testing matters because content decisions are often argued from taste. One person likes the short version. Another wants more detail. Someone senior prefers a phrase they saw on a competitor’s [website](https://contentlexicon.com/website) and suddenly everyone is pretending that counts as strategy.\n\nA good test can bring evidence into that argument. It can show whether people notice a message, understand an offer, complete a task or choose one route more often than another. Used carefully, it helps teams learn from behaviour rather than endless internal debate.\n\nBut [content teams](https://contentlexicon.com/content-team) need to be careful about what they are testing. Not every content question can or should be reduced to a conversion rate. Some content exists to reduce anxiety, explain complexity, build confidence, prevent mistakes or help people make better decisions. Those outcomes may not show up neatly in a quick A/B test.\n\n## How AI changes the picture\n\nAI makes A/B testing easier to overproduce. It can generate headlines, summaries, calls to action, email variants and page introductions at speed. That can be useful when a team has a clear hypothesis and a disciplined way to review the options.\n\nIt can also create a fog of plausible alternatives. If the team tests variations without understanding the purpose of the content, the test becomes theatre: lots of variants, lots of numbers and very little judgement. The danger is not that AI makes bad options. The danger is that it makes too many mediocre ones, quickly enough to make them look intentional.\n\nAI can help with A/B testing when it is used to sharpen hypotheses, draft sensible variants and summarise results. It should not be used as an excuse to test everything or to avoid deciding what good content is supposed to do.\n\n## Questions people ask\n\n**Is A/B testing the same as branching?**\n\nNo. A/B testing compares two or more live variants with users to see which performs better. A [branch](https://contentlexicon.com/branch) is a separate working version where changes can be developed, reviewed and tested before they become live. A/B testing can happen from a branch, but branching is about controlled change rather than optimisation.\n\n**What makes a good A/B test?**\n\nA good A/B test starts with a clear hypothesis, a meaningful measure and a real decision the team is prepared to make afterwards. If nobody knows what will change as a result, the test is probably just decorative analytics.\n\n**Can content teams use A/B testing without huge traffic?**\n\nSometimes, but small samples make it harder to trust the result. With lower traffic, teams may learn more from moderated research, [usability testing](https://contentlexicon.com/user-testing), customer conversations or careful analysis of support and behaviour data.\n\n**What can go wrong?**\n\nTeams can test trivial differences, stop too early, overread weak results or optimise for the wrong metric. The biggest risk is mistaking “this won a test” for “this is better content”.\n\n## Learn more\n\n- [A/B Testing 101](https://www.nngroup.com/articles/ab-testing/) – Nielsen Norman Group’s overview of when A/B testing is useful and what it can and cannot tell you.\n- [A/B testing](https://www.optimizely.com/optimization-glossary/ab-testing/) – a glossary-style explainer of the method and its language, from an experimentation vendor.\n- [Controlled experiments on the web: survey and practical guide](https://ai.stanford.edu/~ronnyk/2009controlledExperimentsOnTheWebSurvey.pdf) – Ron Kohavi and colleagues' paper on online controlled experiments." }, { "term": "Accessibility", "alsoCalled": [ "a11y" ], "slug": "accessibility", "definition": "Making content and services usable by as many people as possible, including disabled people.", "category": "Editorial craft", "related": [ "alt-text", "audio", "content-standards", "frontend", "inclusive-language", "pdf", "plain-language", "readability", "style-guide", "ux", "wcag", "website" ], "updated": "2026-10-06T15:45:00.000Z", "url": "https://contentlexicon.com/accessibility", "markdown": "https://contentlexicon.com/accessibility.md", "credit": "Accessibility, from the Content Lexicon by Contentious (https://contentlexicon.com/accessibility), CC BY 4.0", "body": "In content work, accessibility comes down to clear structure, meaningful links, readable language, useful headings, [alt text](https://contentlexicon.com/alt-text), captions, transcripts and formats that work with assistive technology.\n\nAccessibility is often discussed as if it belonged only to interface design or development. In practice, content decisions affect it constantly: what a link says, whether a heading helps someone navigate, whether instructions are understandable, whether an image has a useful description, whether a transcript exists and whether a page makes sense when read out of visual order.\n\n## Why it matters for content\n\nAccessible content is easier for more people to use. That includes people using screen readers, keyboard navigation, captions, magnification or voice control, but it also includes people who are tired, stressed, unfamiliar with the subject, using a phone in difficult conditions or skimming in a hurry. Accessibility improves reach, usability and quality at the same time.\n\nThe stakes are highest on task-focused pages, service content, forms, legal information, support material and any page where a missed instruction or hidden meaning changes what someone can understand or do. That covers most of the pages people come for. In the UK, public sector bodies also have a legal duty to meet [WCAG](https://contentlexicon.com/wcag) under the Public Sector Bodies Accessibility Regulations 2018, and the Equality Act 2010 applies to service providers more widely. The law is the floor, though, not the reason.\n\n## How AI changes the picture\n\nClear headings, explicit labels, useful alt text and readable structure all help machines as well as people. Accessibility is still for people first; the change is that inaccessible content now fails in more places.\n\nThen there are the accessibility questions AI raises itself: whether generated summaries preserve meaning, whether automated image descriptions are good enough, whether chat interfaces provide usable answers and whether organisations are checking AI output for clarity rather than assuming fluency equals accessibility.\n\n## Questions people ask\n\n**Is accessibility mainly about screen readers?**\n\nScreen readers are one part of it. Accessibility also covers captions, keyboard use, contrast, structure, clarity, transcripts and many other ways people access content.\n\n**Is accessibility the same as WCAG?**\n\nNot quite. WCAG is a technical standard: a set of guidelines with testable success criteria. Accessibility is the broader goal of making content and services usable by more people.\n\n**Does accessibility limit creativity?**\n\nIt limits careless choices. Good accessibility usually sharpens communication rather than narrowing it.\n\n**Is accessibility only for disabled people?**\n\nNo. Disabled people are a central reason for it, but accessible content also tends to be clearer and more usable for everyone else.\n\n## Learn more\n\n- [WCAG 2 Overview](https://www.w3.org/WAI/standards-guidelines/wcag/) – the main introduction to the web accessibility standard.\n- [Writing for Web Accessibility](https://www.w3.org/WAI/tips/writing/) – W3C guidance on content decisions that affect accessibility.\n- [Home Office Posters](https://github.com/UKHomeOffice/posters) – dos and don'ts posters for seven access needs, from anxiety to screenreader users. A visual resource for explaining access needs quickly, available to adapt with credit for non-commercial use under a CC BY-NC-SA licence." }, { "term": "Agentic AI", "alsoCalled": [], "slug": "agentic-ai", "definition": "AI that takes actions and works through multi-step tasks towards a goal, rather than only answering a question.", "category": "AI and tooling", "related": [ "agile", "ai-governance", "conversational-design", "graph-engineering", "guardrails", "human-in-the-loop", "llm", "mcp", "prompt", "skill", "system-one-models", "system-prompt" ], "updated": "2026-10-10T17:11:00.000Z", "url": "https://contentlexicon.com/agentic-ai", "markdown": "https://contentlexicon.com/agentic-ai.md", "credit": "Agentic AI, from the Content Lexicon by Contentious (https://contentlexicon.com/agentic-ai), CC BY 4.0", "body": "An AI agent selects and carries out steps towards a goal, using results from one action to decide what to do next. Its tools might let it search sources, query a [database](https://contentlexicon.com/database) or change a record. Software agents predate generative AI; many current systems use a [large language model](https://contentlexicon.com/llm) to interpret requests and direct the work.\n\n‘Agentic’ is used for varying degrees of autonomy. A predefined [workflow](https://contentlexicon.com/workflow) follows a route set by people; an agent can choose or adapt its steps within the tools and permissions it has. A clear, repeatable task may be better served by a fixed workflow. A run may end when the goal is met, a configured limit is reached, it is cancelled or the system cannot proceed without help.\n\n## Why it matters for content\n\nA weekly review can use a simple date filter to identify overdue guidance. An agent could then examine each page, choose which approved sources to check and change its next step when it finds a conflict, for example by requesting an editor’s decision rather than drafting through the uncertainty. The schedule is fixed; the route through the evidence can vary.\n\nConnecting these steps can reduce the coordination needed between finding a problem and preparing a useful response. The benefit is greater where findings require different follow-up actions, rather than the same sequence every time. A drafting tool produces material for someone to act on; an agent can move the work towards a reviewable result.\n\nPreparing a private draft and updating a public policy need different permissions and approval rules. Teams should specify what the agent may change, which sources it may use and which actions require approval. Tool permissions should enforce those boundaries where possible; instructions alone don’t prevent an unauthorised action.\n\n## How agentic AI changes the picture\n\nAgentic AI can carry work from a brief through research, drafting, checks and preparation for publication. Instead of operating a separate tool at each stage and moving material between them, a team can delegate some of that coordination. An agent can respond to findings, repeat a failed step or send an unresolved issue for review, within a process that still has fixed rules and approval points.\n\nMulti-agent setups divide work between agents with different roles, instructions or tools. One might gather source material while another drafts and a third checks the result against editorial criteria. Their hand-offs need to carry evidence and unresolved questions, not just finished text. A reviewer agent may share the drafting agent’s blind spots, so another pass isn’t automatically an independent check. Adding agents can also increase coordination and cost; the division of work should serve the task.\n\nFor teams using agents, content work includes designing and maintaining that process. Editorial standards need to become usable instructions and evaluation criteria, supported by appropriate sources. Content people need to test what the system produces, investigate failures and improve how the work happens, alongside technical colleagues. The skills involved extend beyond writing and [editing](https://contentlexicon.com/editing) individual pieces to shaping the systems that produce them.\n\n## Questions people ask\n\n**How do agents fit into content workflows?**\n\nAn agent can handle one stage or several connected stages. A workflow might have fixed review and approval points while allowing an agent to choose which source to consult or which check to run next. Some steps remain predictable; others adapt to what the agent finds.\n\n**Is an AI agent just a chatbot?**\n\nA chatbot provides a conversational interface; behind it may be a fixed workflow or an agent. An agent can also work without a chat interface. Using a tool alone doesn’t show that the system chooses its own route.\n\n**Can an agent publish content?**\n\nAn agent can publish if its tools and permissions allow it. Whether it should depends on the risk, the permissions and the review process. Filing a private draft summary is one thing; updating a public policy page is quite another.\n\n**How much autonomy should an agent have?**\n\nMatch autonomy to the task and the consequences of error. An agent may select research steps independently but require approval before sending a message or publishing a change. Test its performance on realistic cases, inspect failures and keep responsibility with a named owner.\n\n## Learn more\n\n- [Agentic AI, explained](https://mitsloan.mit.edu/ideas-made-to-matter/agentic-ai-explained) – MIT Sloan on systems that perceive, reason and act across multi-step tasks, with examples from banking and retail and the risks organisations need to manage.\n- [Building Effective AI Agents](https://www.anthropic.com/engineering/building-effective-agents) – Anthropic’s 2024 distinction between predefined workflows and agents that choose their steps, with a preference for the simplest system that works; the page notes that its tooling has since changed.\n- [I think “agent” may finally have a widely enough agreed upon definition to be useful jargon now](https://simonwillison.net/2025/Sep/18/agents/) – Simon Willison’s September 2025 working definition, “An LLM agent runs tools in a loop to achieve a goal”, and where other definitions cause confusion." }, { "term": "Agile", "alsoCalled": [ "Agile software development", "Agile methods" ], "slug": "agile", "definition": "A way of working that delivers in small steps, learns from feedback and adapts its plans as understanding grows.", "category": "Operations", "related": [ "agentic-ai", "content-design", "content-designer", "content-operations", "content-team", "design-thinking", "devops", "lorem-ipsum", "sign-off", "slow-content", "user-testing", "workflow" ], "updated": "2026-10-16T13:18:00.000Z", "url": "https://contentlexicon.com/agile", "markdown": "https://contentlexicon.com/agile.md", "credit": "Agile, from the Content Lexicon by Contentious (https://contentlexicon.com/agile), CC BY 4.0", "body": "Instead of specifying everything up front and delivering it all at the end, an agile team works in short cycles, often of one to four weeks, and treats each one as a chance to learn. It values working results over exhaustive specification, people from different disciplines working together throughout and a pace the team can keep up. The term comes from software. In February 2001, 17 developers met at Snowbird in Utah and wrote the Manifesto for Agile Software Development: four values and 12 principles drawn from the lighter methods they were already using.\n\nIt helps to distinguish small-a agile from big-A Agile. The adjective names a quality of the work, and any team can have it. The noun names the industry that grew around the manifesto, with its frameworks, certifications, coaches and ceremonies. Scrum, with fixed-length sprints, a product owner and a scrum master, is the best known; kanban manages a continuous flow of work on a visible board, limiting how much is in progress at once. Frameworks can be useful scaffolding, but trouble starts when the method is installed and the habits never arrive. Dave Thomas, one of the manifesto's authors, argued in 2014 that the noun had become a marketing term, and Martin Fowler has called the result faux-agile: the vocabulary without the values.\n\n## Why it matters for content\n\nContent often arrives in agile teams late and in the wrong shape. A backlog item that reads \"add copy\" assumes words can be poured into boxes designed without them, and a sprint review that admires a prototype full of [lorem ipsum](https://contentlexicon.com/lorem-ipsum) is admiring something nobody can yet use. The manifesto's own test argues against this. If working software is the measure of progress, a page whose explanation hasn't been written isn't working. When [content designers](https://contentlexicon.com/content-designer) join research, write real words into early prototypes and test them alongside the interface, iteration covers the thing readers came for. Problems also surface while they're still cheap to fix, such as a heading that won't fit or a process that can't be explained plainly.\n\n[Content teams](https://contentlexicon.com/content-team) can borrow the habits for their own work, though they transfer unevenly. Editorial work already iterates through drafts, edits and corrections, and newsrooms have always replanned when events intervene. A visible board, a limit on how many pieces are in progress and a short monthly retrospective can show a team's real workload and stop everything being urgent at once. Some content can't be released half-finished, though: a medicines leaflet, a legal notice or a fundraising appeal has to be right when a reader acts on it, and \"we'll fix it next sprint\" is little comfort to them. Agility moves review earlier and makes it smaller, so a subject expert sees a rough draft in the first week rather than a finished page on deadline day. A project team that disbands at launch also leaves its content without an owner, which is one reason Fowler argues for organising teams around lasting products rather than projects.\n\n## How AI changes the picture\n\nMuch of agile's machinery was designed around scarce, slow human building time. Sprints, stand-ups and small batches helped teams manage that cost. When an agent can produce a working prototype or several drafts of a page in an afternoon, the constraint moves to deciding what's worth making and finding out whether it works for people. Content teams know this constraint well, because writing was rarely the slowest part of publishing. The loop of small steps and learning can now run much faster, but only as fast as a team can gather real feedback and make decisions. Ten generated variants remain ten guesses until readers have used them.\n\nAI also tests the manifesto's preferences. \"Individuals and interactions over processes and tools\" and \"working software over comprehensive documentation\" were weightings rather than prohibitions, and agentic work shifts the weight. The instructions, style guidance, [content models](https://contentlexicon.com/content-model) and review gates that agents work within become part of the product, and an agent can only follow standards someone has written down. Kent Beck, another of the manifesto's authors, describes \"augmented coding\" as keeping tests and careful design while AI does much of the typing. The content equivalent keeps editorial standards and evaluation in place while AI drafts, which gives people who can write those standards clearly a larger part in an agile team's work.\n\nThe principle most at risk is sustainable pace. The manifesto asks for a pace that sponsors, developers and users can maintain indefinitely, and cheap output makes that easy to forget. Every page published in a fast cycle still has to be maintained, and every change asks something of readers who had learned where things were. Agility includes knowing when not to ship.\n\n## Questions people ask\n\n**Is agile the opposite of waterfall?**\n\nThey're often presented that way. Waterfall runs in sequence, from requirements through design, build and testing to release, with each phase signed off before the next begins. Most organisations mix the two, fixing budgets and launch dates in advance and working iteratively inside them. Winston Royce's 1970 paper, usually cited as waterfall's origin, described the pure sequence as risky and argued for feedback between stages and an early trial version.\n\n**Does a content team need sprints?**\n\nOnly if its work divides into pieces that can be finished in a fortnight or so. A team redesigning a set of guidance pages may find sprints useful. A team handling a steady stream of requests, corrections and updates is often better served by a kanban board that shows what's waiting, what's in progress and where work gets stuck.\n\n**Where does a content designer fit in a Scrum team?**\n\nInside it. The Scrum Guide counts everyone who helps create each increment of the product as a developer, whatever their discipline. A content designer therefore belongs in planning, reviews and retrospectives rather than receiving tickets at the end. In practice that means joining user research, working with designers on prototypes and helping decide when a piece of work is done.\n\n**Can we be agile without adopting a framework?**\n\nYes. A small charity might meet once a month, look at how people used last month's pages, choose two changes and check the following month whether they helped. That's the loop Thomas describes: work out where you are, take a small step, adjust what you think in the light of what happened and go round again. It needs no certification. A framework becomes worth considering when several teams have to coordinate their work.\n\n## Learn more\n\n- [Agile is Dead (Long Live Agility)](https://pragdave.me/thoughts/active/2014-03-04-time-to-kill-agile.html) – Dave Thomas, one of the manifesto's authors, makes the 2014 case for treating agile as an adjective rather than a product, and reduces working with agility to a four-step loop.\n- [Aligning Content Work with Agile Processes](https://alistapart.com/article/aligning-content-work-with-agile-processes) – the content strategist Brendan Murray on what content teams already share with agile iteration, why early builds deserve real content rather than lorem ipsum and how stand-ups can bring content and development closer.\n- [Does AI Make the Agile Manifesto Obsolete?](https://www.infoq.com/news/2026/02/ai-agile-manifesto-debate) – InfoQ's 2026 report on the argument over agentic development, from the claim that agents are too fast for agile to Kent Beck's augmented coding and the view that agile is optimising a constraint that has moved." }, { "term": "AI governance", "alsoCalled": [], "slug": "ai-governance", "definition": "How an organisation oversees the risks, responsibilities and performance of the AI systems it develops, buys or uses.", "category": "Governance", "related": [ "agentic-ai", "ai-use-policy", "content-governance", "guardrails", "human-in-the-loop" ], "updated": "2026-10-10T17:00:00.000Z", "url": "https://contentlexicon.com/ai-governance", "markdown": "https://contentlexicon.com/ai-governance.md", "credit": "AI governance, from the Content Lexicon by Contentious (https://contentlexicon.com/ai-governance), CC BY 4.0", "body": "AI governance is the organisation-wide system for deciding how artificial intelligence may be chosen, built, used, checked and stopped. It joins policy to accountability: someone must understand each system, its purpose, its risks and who answers when it goes wrong.\n\nFor [content teams](https://contentlexicon.com/content-team), this is wider than an [AI use policy](https://contentlexicon.com/ai-use-policy). That policy might say whether a [writer](https://contentlexicon.com/writer) may paste unpublished copy into a chatbot. AI governance also covers the chatbot built into the [website](https://contentlexicon.com/website), the model answering customer questions and the automated agent updating records.\n\n## Why it matters for content\n\nContent is both an input to AI and one of its most visible outputs. Weak source material can produce misleading answers, while an AI feature can publish thousands of words before an editor has seen the first paragraph. Governance gives content people a place in decisions that might otherwise be treated as purely technical.\n\nGood governance names the system owner, the people affected, the evidence needed before release and the conditions for pausing it. It also keeps an inventory. If nobody can list the AI systems touching public content, oversight has already become a guessing game.\n\n## How AI governance changes the picture\n\nAI governance asks organisations to manage systems throughout their life, not merely approve a tool once. Models change, suppliers update services and source [content decays](https://contentlexicon.com/content-decay). Evaluation therefore needs to continue after launch, using examples that reflect real content, real audience needs and the harm a wrong answer could cause.\n\nRegulation adds duties according to the system and its risk. The EU AI Act uses a risk-based approach, while the NIST AI Risk Management Framework offers voluntary guidance for designing, using and evaluating trustworthy systems. Neither can be replaced by a committee with an impressive name and no authority.\n\n## Questions people ask\n\n**Is AI governance the same as an AI use policy?**\n\nNo. An AI use policy tells people how they may use AI. Governance is the wider system of ownership, risk management, evaluation, oversight and compliance around the AI itself.\n\n**Does this belong to the technology team?**\n\nNot alone. Technology, legal, information governance and security all matter, but content specialists understand the source material, publishing risks and audience consequences.\n\n**Do small organisations need it?**\n\nThey need a proportionate version. A register of tools, a named owner, clear prohibited uses and a review route may be enough for a small team. The paperwork should match the risk.\n\n**What should a content team ask first?**\n\nAsk which AI systems create, select, transform or deliver content, what sources they use and who can stop them. Those three questions expose a surprising amount.\n\n## Learn more\n\n- [NIST AI Risk Management Framework](https://www.nist.gov/itl/ai-risk-management-framework) – NIST's voluntary AI RMF 1.0 (2023), covering risks across AI design, use and evaluation; under revision as of October 2026.\n- [AI Act enters into force](https://commission.europa.eu/news-and-media/news/ai-act-enters-force-2024-08-01_en) – the European Commission's August 2024 introduction to the Act's risk-based categories, published when it entered into force. A historical introduction, not an up-to-date compliance timetable.\n- [OECD AI Principles overview](https://oecd.ai/en/ai-principles) – the intergovernmental standard adopted in 2019 and updated in 2024, with five values-based principles and five recommendations for policymakers." }, { "term": "AI slop", "alsoCalled": [ "Slop" ], "slug": "ai-slop", "definition": "Low-value AI-generated material produced with little care, often at speed and in large quantities.", "category": "AI and tooling", "related": [ "content-decay", "guardrails", "hallucination", "linting", "slow-content", "storytelling" ], "updated": "2026-10-11T06:25:00.000Z", "url": "https://contentlexicon.com/ai-slop", "markdown": "https://contentlexicon.com/ai-slop.md", "credit": "AI slop, from the Content Lexicon by Contentious (https://contentlexicon.com/ai-slop), CC BY 4.0", "body": "AI slop includes generic articles, fabricated testimonials, repetitive videos and synthetic images produced without much attention to their purpose or reliability. It can look polished: correct [grammar](https://contentlexicon.com/grammar), a convincing format and confident claims can conceal how little there is behind them.\n\nUsing AI doesn’t make content slop. A routine automated message can do its job well, while an elaborate article can offer almost nothing. Deliberately absurd entertainment also needs judging differently from material presented as factual evidence.\n\n## Why it matters for content\n\nSlop makes useful material harder to find and gives readers more work to do. A polished report that doesn’t connect its recommendations to evidence leaves the recipient to reconstruct the reasoning. The time saved in drafting may reappear as checking, correction or uncertainty elsewhere. Facts can be correct while the piece still adds little: repeating familiar advice isn’t the same as explaining how it applies.\n\nPublishing quotas and impression targets can reward another plausible item without requiring it to add much for the reader. Teams can appear productive while accumulating material that needs review and maintenance. A production target should leave room to reject a weak piece or commission the research it lacks.\n\n## How AI slop changes the picture\n\nCheap, repeatable content generation puts pressure on selection and moderation. A feed or [knowledge base](https://contentlexicon.com/knowledge-base) can accumulate many versions of the same thin account, each requiring someone to decide whether it adds anything. Publishing controls need to consider duplication and evidence as well as presentation; a fluent paragraph may pass a language check while failing an editorial one.\n\nSlop can also become source material for further generation. An assistant might summarise an unsupported report and remove the context that would have made its weakness visible. Where content feeds other systems, retain links to evidence and track derived versions. Correcting the source won’t necessarily update a detached answer or summary.\n\n## Questions people ask\n\n**Is all AI-generated content slop?**\n\nCarefully researched or usefully automated content can use AI without being slop. Low effort doesn’t settle quality, and taste complicates the label: absurd generated entertainment may give people pleasure or a shared joke. That doesn’t excuse material falsely presented as evidence.\n\n**How do you recognise it?**\n\nCheck its claims and sources, whether it explains anything specific and whether it delivers what it promises. Repetition and familiar phrasing can [prompt](https://contentlexicon.com/prompt) closer inspection, but they don’t establish AI authorship or poor quality on their own.\n\n**Can editing fix it?**\n\n[Editing](https://contentlexicon.com/editing) can clarify sound material. A draft built on unverified claims needs research; one without a worthwhile purpose may need a different brief or should not be published. Removing familiar AI phrases won’t supply missing knowledge.\n\n**How is slop different from hallucination?**\n\n[Hallucination](https://contentlexicon.com/hallucination) describes invented or unsupported output presented as factual. Slop is a wider judgement about quality and care. A piece can be slop without a single fabricated fact, while one hallucinated detail can undermine otherwise useful work.\n\n## Learn more\n\n- [Why Slop Matters](https://dl.acm.org/doi/10.1145/3786777) – an opinion paper arguing for the social and aesthetic value of some generated content, and examining why it is produced and consumed. A counterargument to blanket dismissal.\n- [What is AI slop?](https://theconversation.com/what-is-ai-slop-a-technologist-explains-this-new-and-largely-unwelcome-form-of-online-content-256554) – Adam Nemeroff’s account of generated text, images and music in the attention economy, emphasising misleading material and the pressures on creators and moderators.\n- [Slop is the new name for unwanted AI-generated content](https://simonwillison.net/2024/May/8/slop) – Simon Willison’s May 2024 post that helped popularise the term, comparing slop to spam: generated without care and pushed at people who didn’t ask for it." }, { "term": "AI use policy", "alsoCalled": [], "slug": "ai-use-policy", "definition": "The rules for how people in an organisation may use AI, including approved tools, prohibited material, review and disclosure.", "category": "Governance", "related": [ "ai-governance", "content-policy", "copyright-and-licensing", "guardrails", "hallucination", "human-in-the-loop" ], "updated": "2026-10-10T16:13:00.000Z", "url": "https://contentlexicon.com/ai-use-policy", "markdown": "https://contentlexicon.com/ai-use-policy.md", "credit": "AI use policy, from the Content Lexicon by Contentious (https://contentlexicon.com/ai-use-policy), CC BY 4.0", "body": "An AI use policy has to answer the questions people meet mid-task. Can I paste this interview transcript into a chatbot? Can I publish a headline a tool suggested? Do I need to say AI helped?\n\nThe useful version is short enough to consult during work and specific enough to settle a real decision. A ban without usable alternatives or a clear explanation can encourage unapproved use. A cheerful invitation to “use AI responsibly” leaves every difficult question unresolved.\n\n## Why it matters for content\n\n[Content teams](https://contentlexicon.com/content-team) handle unpublished plans, personal data, copyrighted material, brand guidance and public claims. They also work with tools that can produce fluent errors. A policy turns vague caution into practical boundaries: which content can enter which system, what can be generated and who checks it before publication.\n\nThe policy should distinguish risk. Generating ten headline options from public copy is different from asking a public chatbot to summarise confidential interview transcripts. Treating both tasks alike makes the rules either needlessly restrictive or dangerously loose.\n\n## How AI use policies change the picture\n\nThe policy sits above particular [prompts](https://contentlexicon.com/prompt) and [guardrails](https://contentlexicon.com/guardrails). It explains the organisation's position while tools, [workflows](https://contentlexicon.com/workflow) and training put that position into practice. Because services and laws change, it needs an owner, a [review date](https://contentlexicon.com/review-date) and a route for approving exceptions.\n\nFor content, human review must mean more than a person clicking approve. The reviewer needs time, subject knowledge and access to the source. At high volumes, a nominal [human in the loop](https://contentlexicon.com/human-in-the-loop) can become an expensive rubber stamp.\n\n## Questions people ask\n\n**What should an AI use policy include?**\n\nAt minimum: scope, approved and prohibited uses, data rules, copyright expectations, review responsibilities, disclosure, incident reporting and a review date.\n\n**Can people use ChatGPT and similar tools for work?**\n\nIt depends less on the tool than on the account. The same product can come as a free or personal subscription under consumer terms, or as a business service with a contract covering data retention, model training and admin controls. Decide which accounts are approved for which material: public copy might be fine almost anywhere, while confidential interview transcripts belong only in a service the organisation has assessed. Name the boundary, explain it and give people a route to ask about exceptions.\n\n**Do we have to disclose AI use?**\n\nThat depends on the context, law and effect on the audience. Set out when disclosure is required and what it needs to explain. Labelling a routine spelling check the same way as a synthetic interview can hide the distinction people need.\n\n**Is a policy enough?**\n\nNo. People need suitable tools, examples, training and a way to ask questions. A document nobody can apply is decorative governance.\n\n## Learn more\n\n- [Create an AI policy](https://www.ai.gov.au/staying-safe-and-responsible/essential-ai-practices/create-ai-policy) – practical guidance and a template from Australia's National AI Centre, covering acceptable uses, data, oversight and review.\n- [ICO guidance on AI and data protection](https://ico.org.uk/for-organisations/uk-gdpr-guidance-and-resources/artificial-intelligence/) – UK data-protection guidance on AI. Its legal context is jurisdiction-specific and the guidance is being reviewed following legislative change.\n- [The Guardian's approach to generative AI](https://www.theguardian.com/help/insideguardian/2023/jun/16/the-guardians-approach-to-generative-ai) – the Guardian's June 2023 statement of three broad principles from a cross-team working group: a published example of a newsroom policy." }, { "term": "Alt text", "alsoCalled": [ "Alternative text" ], "slug": "alt-text", "definition": "A short description attached to an image so people and systems that cannot see it still get the meaning.", "category": "Editorial craft", "related": [ "accessibility", "automated-testing", "metadata", "wcag" ], "updated": "2026-10-11T06:23:00.000Z", "url": "https://contentlexicon.com/alt-text", "markdown": "https://contentlexicon.com/alt-text.md", "credit": "Alt text, from the Content Lexicon by Contentious (https://contentlexicon.com/alt-text), CC BY 4.0", "body": "Alt text is short for alternative text. Screen readers use it for people who are blind or partially sighted, browsers can show it when an image fails to load, and search engines and AI systems use it to understand what an image is for. Good alt text describes the meaning of the image in context rather than listing visible objects.\n\nAsk what the reader would miss if the image disappeared. That's usually what the alt text needs to cover. Decorative images should be marked as decorative so they're skipped rather than described.\n\n## Why it matters for content\n\nAlt text makes image-based meaning available in words. That matters first for [accessibility](https://contentlexicon.com/accessibility), but it also improves the quality of search, indexing and machine use of content. A chart, infographic or screenshot may carry key information; if that information exists only in pixels, many readers and systems will miss it entirely. Writing alt text forces a team to decide what an image is doing on the page.\n\nImages that carry information need it most: charts, diagrams, product screenshots, photographs used as evidence, event images with visible text and graphics that summarise a point. It's especially important on high-traffic pages, campaign pages and service pages where missing visual information changes what a reader can understand or do.\n\n## How AI changes the picture\n\nSearch engines, image systems and generative tools all need text if they're to do anything sensible with an image. Alt text is written for people and should stay that way, but a missing description now costs you with machines as well.\n\nAI can also write the alt text. Automatic image description is useful for a first draft, or for making a start on a media library with thousands of undescribed images. But a model describes what's visible, not why the image is there. It can see a bar chart with five blue bars; it can't know that the point of the chart is the drop in the final year. Someone who knows the page still has to decide what the image is for.\n\n## Questions people ask\n\n**Do all images need alt text?**\n\nAll images need a decision. Informative and functional images need alt text. Decorative images should usually have empty alt text (alt=\"\") so screen readers skip them.\n\n**Is alt text the same as an image caption?**\n\nNo. A caption is visible to everyone. Alt text is primarily for people and systems that can't access the image visually.\n\n**Should alt text describe every detail?**\n\nUsually not. It should describe what matters in context. A portrait, a chart and a navigation icon all need different levels of description.\n\n**Does alt text help SEO?**\n\nSometimes, indirectly, because it helps systems understand an image. But writing it for search rather than for meaning usually produces bad alt text.\n\n## Learn more\n\n- [Alt Text as Poetry](https://alt-text-as-poetry.net/) – Bojana Coklyat and Shannon Finnegan's workbook reframes alt text as a creative practice, with writing exercises built around voice and belonging rather than compliance. A deliberately different register from the standards links.\n- [An alt Decision Tree](https://www.w3.org/WAI/tutorials/images/decision-tree/) – W3C's flowchart for deciding whether an image needs alt text and what kind.\n- [WebAIM: Alternative Text](https://webaim.org/techniques/alttext/) – a detailed explainer of what good alt text does." }, { "term": "API", "alsoCalled": [ "Application programming interface" ], "slug": "api", "definition": "Application programming interface: the agreed way one piece of software asks another for something, like a waiter between the diner and the kitchen.", "category": "AI and tooling", "related": [ "backend", "cms", "machine-readable" ], "updated": "2026-10-10T17:58:00.000Z", "url": "https://contentlexicon.com/api", "markdown": "https://contentlexicon.com/api.md", "credit": "API, from the Content Lexicon by Contentious (https://contentlexicon.com/api), CC BY 4.0", "body": "An API is an agreed format that lets one piece of software request something from another and get a predictable answer back. It sets out what can be asked for, how to phrase the request and what the reply will look like, so the systems on either side don't need to know how the other works inside.\n\nWeb APIs became common in the 2000s, when publishers and platforms began opening their content and data to outside developers.\n\nAn endpoint is the address a request goes to, such as one for articles and another for authors. Authentication, usually a key or token, controls who can ask and what they can see. Rate limits cap how many requests a system can make in a given time. In a [headless CMS](https://contentlexicon.com/headless-cms), the content API is how content leaves the system: the [website](https://contentlexicon.com/website), the app and every other channel request it from there.\n\n## Why it matters for content\n\nAPIs turn content from something fixed to one page into something other systems can use. A website can pull the current product description, an app can show the same help article and a support tool can check a draft against the one maintained version of the guidance. Every channel draws on the same source, so a correction made once reaches all of them.\n\nThat only works if the content behind the API is structured. An API passes on whatever it's given: if a \"summary\" field sometimes holds a summary and sometimes holds the opening paragraph, every system that requests it inherits the confusion. Field names, required fields and [content types](https://contentlexicon.com/content-type) become an agreement with other teams, and changing them can break things those teams have built. [Content model](https://contentlexicon.com/content-model) changes start to need the same care as code changes, with [versioning](https://contentlexicon.com/versioning) and notice.\n\nPublishing through an API also changes who controls the content. Once it leaves, the publisher no longer controls every place it appears. When the Guardian opened its content API in 2009, developers could store articles for 24 hours at most, and were warned that keeping them longer risked republishing a story withdrawn for legal reasons.\n\n## How AI changes the picture\n\nAssistants and agents reach content through APIs. A chatbot that answers from approved guidance, a writing tool that checks drafts against current rules and an agent that updates product records all depend on a controlled way to request and change content. [MCP](https://contentlexicon.com/mcp), the Model Context Protocol, now supported by many AI tools, gives models a standard way to discover those connections and decide when to use them.\n\nThat raises the stakes on what an API exposes. Draft, expired and approved content need to be distinguishable in the response, or a model will treat them as equally usable. Permissions matter more when the caller is software acting for a person, because an agent with write access can make a lot of changes very quickly. For [content teams](https://contentlexicon.com/content-team), \"can AI use this?\" usually means \"is there a structured, governed way for a system to reach it?\"\n\n## Questions people ask\n\n**Do content people need to understand APIs?**\n\nThey don't need to build them, but it helps to know what APIs make possible and what they ask of content. Being able to read API documentation well enough to see which fields exist and what they're called is useful when planning a content model or a migration.\n\n**Is an API the same as a website?**\n\nNo. A website is for people to use directly; an API is for software. Many organisations have both, and in a headless setup the website is one of the API's clients.\n\n**Does having an API make content reusable?**\n\nOnly if the content behind it is structured, consistent and maintained. An API can deliver a page-sized block of formatted text as easily as clean fields, and the systems receiving it will struggle to do much else with it.\n\n**What's the difference between an API and an integration?**\n\nAn API is the interface one system offers. An integration is a connection built with it, such as a [CMS](https://contentlexicon.com/cms) sending new articles to a newsletter tool. One API can support many integrations.\n\n## Learn more\n\n- [Introduction to web APIs](https://developer.mozilla.org/en-US/docs/Learn_web_development/Extensions/Client-side_APIs/Introduction) – MDN's beginner's guide to what APIs are and how they work, including the third-party APIs that let websites bring in maps, video and comments.\n- [A few notes on the Guardian Open Platform](https://simonwillison.net/2009/Mar/10/openplatform/) – Simon Willison's 2009 notes on the launch of the Guardian's content API, covering API keys, rate limits, filtering by tag and a 24-hour limit on storing articles.\n- [From guidance to infrastructure](https://voicetoneandstyle.com/learn/from-guidance-to-infrastructure) – Voice, Tone & Style on turning a style guide into something tools can query through an API." }, { "term": "Atomic content", "alsoCalled": [], "slug": "atomic-content", "definition": "Content broken into small, structured, reusable pieces that get assembled into pages rather than copied between them.", "category": "Structure and systems", "related": [ "content-model", "create-once-publish-everywhere", "dry", "single-source-of-truth", "structured-content" ], "updated": "2026-10-11T06:27:00.000Z", "url": "https://contentlexicon.com/atomic-content", "markdown": "https://contentlexicon.com/atomic-content.md", "credit": "Atomic content, from the Content Lexicon by Contentious (https://contentlexicon.com/atomic-content), CC BY 4.0", "body": "Atomic content is content broken into the smallest pieces that still make sense on their own, each stored and maintained once and assembled into pages, products and other outputs as needed. An atom might be a definition, a statistic with its source, a staff bio, a service description, a disclaimer or the short paragraph that explains what an organisation does. Each has one home, one owner and one current version, and everywhere that shows it pulls from that source.\n\nThe test for an atom is whether it can stand alone without losing meaning. A product's name, price and description might only make sense together, in which case they form one unit, and splitting them makes them harder to manage. Atomic content also overlaps with neighbouring ideas. [Structured content](https://contentlexicon.com/structured-content) is the wider practice of organising content into defined fields and relationships, and atoms are one result of it. Content reuse is what atoms make possible. A [component](https://contentlexicon.com/component) is a design element, such as a card or banner, that displays content, so the same atom might appear in several components.\n\nThe name borrows from Brad Frost's atomic design, a 2013 method for building interfaces from atoms, molecules and organisms. The content side has older roots in technical writing's single sourcing and in NPR's \"[create once, publish everywhere](https://contentlexicon.com/create-once-publish-everywhere)\" approach of the late 2000s, which kept stories separate from any one display so the same content could go to the [website](https://contentlexicon.com/website), iPods, radio displays and whatever came next.\n\n## Why it matters for content\n\nRepeated content is where inconsistency breeds. The same paragraph appears on a website, in a proposal, in a slide deck and on a campaign page. One version is edited, another is forgotten and soon nobody knows which is true. Working in atoms means a correction made once shows up everywhere, and it's clear who maintains what.\n\nAtoms also make the [content model](https://contentlexicon.com/content-model) visible. Deciding what counts as a unit forces questions that page-based publishing lets teams avoid: which facts change often, which need legal approval, which differ by region. The answers shape fields, [workflows](https://contentlexicon.com/workflow) and [review dates](https://contentlexicon.com/review-date).\n\nAtoms can lose context, though. An atom written to fit anywhere can end up fitting nowhere well, and a page assembled entirely from reusable pieces can read like a form. Most teams keep atoms for facts, definitions and approved statements, and write the connecting prose for each page.\n\n## How AI changes the picture\n\nSmaller, well-labelled pieces are easier for machines to find, interpret and recombine. A retrieval system working from a clear definition, an approved service description or a sourced statistic has less to misread than one working from a long page with mixed purposes, and it can cite the exact piece it used.\n\nGenerative tools also make near-duplicates cheap. Without a canonical source, a system can end up choosing between three slightly different charity registration numbers or four descriptions of the same service, with no way of telling which is current. Atomic content gives AI tools one version to point to, and gives people a single place to fix when a tool gets something wrong.\n\n## Questions people ask\n\n**Is atomic content the same as structured content?**\n\nThey overlap. Structured content is the wider practice of organising content into defined fields and relationships. Atomic content is about the reusable units that practice produces, and the discipline of keeping one maintained version of each.\n\n**Does everything need to be atomic?**\n\nNo, only the pieces where reuse, accuracy or consistency matter. A one-off feature article can stay whole, though it might still pull in a reusable definition or [call to action](https://contentlexicon.com/call-to-action).\n\n**How small should an atom be?**\n\nSmall enough to reuse and large enough to make sense alone. If a piece needs the text around it to be understood, it's too small. If people keep [editing](https://contentlexicon.com/editing) part of it for each new context, it's probably two pieces.\n\n**Who owns an atom?**\n\nWhoever is accountable for the fact it states, who may not be the person who first wrote it. A paragraph about fees might belong to finance even if the [content team](https://contentlexicon.com/content-team) maintains the wording. Recording that owner is part of what makes reuse safe.\n\n## Learn more\n\n- [Atomic Design Methodology](https://atomicdesign.bradfrost.com/chapter-2/) – Brad Frost on building interfaces from atoms, molecules and organisms, and why content and design can't be separated.\n- [A Content Model Is Not a Design System](https://alistapart.com/article/a-content-model-is-not-a-design-system/) – Mike Wills in A List Apart on naming content by meaning rather than layout and keeping related content together so every channel can reuse it.\n- [Atomic content](https://contentious.ltd/thoughts/atomic-content) – the case for designing small, reusable pieces that work for pages, products and machines." }, { "term": "Attention economy", "alsoCalled": [], "slug": "attention-economy", "definition": "An economic system in which people's limited attention is competed for, captured and turned into commercial or social value.", "category": "Foundations", "related": [ "ai-slop", "social-media" ], "updated": "2026-10-06T20:57:00.000Z", "url": "https://contentlexicon.com/attention-economy", "markdown": "https://contentlexicon.com/attention-economy.md", "credit": "Attention economy, from the Content Lexicon by Contentious (https://contentlexicon.com/attention-economy), CC BY 4.0", "body": "Platforms want people to keep scrolling, watching, clicking and reacting because attention can be converted into advertising revenue, data and influence. It isn't only a commercial trade: attention can also accumulate as visibility and reputation, which people exchange for standing, opportunities and reach.\n\nFor [content teams](https://contentlexicon.com/content-team), it isn't background. It shapes what kinds of content get rewarded, what rhythms feel normal and what pressures organisations absorb without always noticing.\n\n## Why it matters for content\n\nThe attention economy matters because it often rewards the wrong things. What gets attention isn't always what's useful, truthful or humane. It's often what's immediate, emotional, novel, provocative or addictive. Clickbait is the obvious pantomime villain here, but the same logic can shape perfectly respectable campaigns, headlines and [content plans](https://contentlexicon.com/content-plan) too. Outrage travels well. Nuance can be a harder sell.\n\nPurpose-driven organisations have to communicate inside channels designed to reward speed and stimulation, but they don't have to accept those values uncritically. A charity, university, public body or thoughtful business can decide that success isn't more impressions, more posts or more noise.\n\n## How AI changes the picture\n\nAI increases the supply of content dramatically, which makes attention feel even scarcer. If everyone can produce more posts, pages, summaries and campaigns, the competition for attention becomes louder and less rewarding for the people being competed over.\n\nThe same tools can tune headlines, thumbnails and send times for engagement, testing dozens of variants until one grabs. So the question to ask of a draft shifts from “will this perform?” to “does this deserve someone’s attention?”\n\nA reader-controlled assistant could change how attention is allocated rather than merely speeding up its consumption: screening incoming material, grouping updates into a single digest and helping someone choose a bounded reading session instead of an endless feed. Whether that happens depends on whose interests the assistant is designed to serve.\n\n## Questions people ask\n\n**Is the attention economy just social media?**\n\nNo. Social platforms are central to it, but the same pressures affect search, [email](https://contentlexicon.com/email), news, advertising and many digital products.\n\n**Is all attention bad?**\n\nAttention itself is necessary. The problem is the constant grabbing at it without giving enough value back.\n\n**How does this relate to slow content?**\n\n[Slow content](https://contentlexicon.com/slow-content) is one response: it values usefulness, substance and relationship over volume, interruption and constant performance.\n\n**Can organisations avoid the attention economy entirely?**\n\nRarely. But they can decide how much to adapt to it and where to resist it.\n\n## Learn more\n\n- [The Second Wave of Attention Economics](https://academic.oup.com/iwc/article/37/1/18/7733851) – Maxi Heitmayer's theoretical proposal that attention accumulates as reputation and can be exchanged for other forms of value, not only sold to advertisers.\n- [The Attention Economy](https://www.humanetech.com/youth/the-attention-economy) – the Center for Humane Technology's guide, written for young people, to how platforms profit from capturing attention.\n- [Slow Content](https://slowcontent.org/) – Contentious's case for depth and purpose over rapid production." }, { "term": "Audience segmentation", "alsoCalled": [], "slug": "audience-segmentation", "definition": "Dividing an audience into useful groups so content can meet different needs, behaviours or contexts.", "category": "Research and evidence", "related": [ "channel-strategy", "paddle-swim-dive", "persona", "user-needs" ], "updated": "2026-10-10T17:50:00.000Z", "url": "https://contentlexicon.com/audience-segmentation", "markdown": "https://contentlexicon.com/audience-segmentation.md", "credit": "Audience segmentation, from the Content Lexicon by Contentious (https://contentlexicon.com/audience-segmentation), CC BY 4.0", "body": "Instead of writing for “everyone”, which usually means writing for nobody in particular, segmentation asks who people are in relation to the thing you are helping them understand or do.\n\nSegments can be based on many things: needs, motivations, confidence, knowledge, behaviour, location, relationship to the organisation or how far they've got with a task. Good segmentation is practical: it helps teams decide what to say, what to leave out, what examples to use, what channels matter and how much explanation someone may need. Segments also overlap. The same person can be a donor, a volunteer and a worried parent in the same week, so segments work better as lenses than as boxes.\n\n## Why it matters for content\n\nAudience segmentation helps content become more specific without becoming chaotic. It gives teams a way to recognise that different people may need different levels of detail, different evidence, different [calls to action](https://contentlexicon.com/call-to-action) or different routes through the same material. It's one of the ways content moves from “we have published the information” to “the right people can use it”.\n\nSegmentation earns its keep when a single organisation serves several audience segments, when one subject has to work for people with different levels of knowledge or when a [channel strategy](https://contentlexicon.com/channel-strategy) depends on knowing who is where and why. It's also useful when content has become too generic to be helpful, or when every page seems to be trying to speak to five people at once.\n\n## How AI changes the picture\n\nAI makes segmentation more useful and riskier. It can help adapt content for different audience needs, reading contexts or levels of prior knowledge, but it needs clear audience distinctions to work from. If the segments are vague, stereotyped or based only on internal labels, AI can produce superficially tailored content that isn't meaningfully useful.\n\nA newcomer might choose an explanation of a report's terminology, while a specialist goes straight to its assumptions and methods. Someone else may want the same findings explained in relation to their role. AI can help vary the depth and route through shared source material; it needn't assign each person a permanent label. Start with expressed preferences rather than inferred stereotypes, seek consent where personal data is involved and keep the underlying evidence consistent.\n\n## Questions people ask\n\n**Is audience segmentation the same as personas?**\n\nNo. [Personas](https://contentlexicon.com/persona) are one way to express audience understanding. Segmentation is the underlying act of grouping people in ways that help decisions.\n\n**Should we segment by job title?**\n\nSometimes, but job title is often a weak shortcut. Needs, behaviour, confidence, context and task are usually more useful for content decisions.\n\n**Can one piece of content serve several segments?**\n\nYes, if it's structured carefully. Sometimes the answer is separate content; sometimes it's one page with clearer routes through it.\n\n**How many segments do we need?**\n\nAs few as are useful. Too many segments can become as unhelpful as none, especially if the team can't maintain different content for them.\n\n## Learn more\n\n- [Content Strategy for the Web](https://www.informit.com/store/content-strategy-for-the-web-9780321808301) – the second edition of Kristina Halvorson and Melissa Rach's book, a broad foundation for content strategy rather than a segmentation tutorial.\n- [Just-Right Personas: How to Choose the Scope of Your Personas](https://www.nngroup.com/articles/persona-scope/) – Kim Flaherty at Nielsen Norman Group on why the right way to divide an audience depends on the decision you're making.\n- [Explanations – Thinking Styles](https://indiyoung.com/explanations-thinking-styles/) – Indi Young on grouping people by how they think and what drives their decisions, rather than by demographics." }, { "term": "Audio", "alsoCalled": [], "slug": "audio", "definition": "Content people hear rather than read: podcasts, voice notes, interviews, narration, audio clips and spoken interfaces.", "category": "Editorial craft", "related": [ "accessibility", "channel-strategy" ], "updated": "2026-10-06T15:16:00.000Z", "url": "https://contentlexicon.com/audio", "markdown": "https://contentlexicon.com/audio.md", "credit": "Audio, from the Content Lexicon by Contentious (https://contentlexicon.com/audio), CC BY 4.0", "body": "Audio can be intimate, immediate and expressive in ways written content is not, because [voice](https://contentlexicon.com/voice) carries pace, emphasis, hesitation, warmth and authority.\n\nFor [content teams](https://contentlexicon.com/content-team), audio isn't “the same words, but with a microphone”. Spoken content has its own rhythm: sentences need room to breathe and the structure has to be audible. Listeners can't skim. Nor can they glance back up the page for a heading they missed, so signposting, pacing and repetition all work differently.\n\n## Why it matters for content\n\nAudio gives organisations another way to explain, teach, interview, reflect and build a relationship with an audience. It can be especially useful where human voice adds trust or texture: interviews, expert conversations, stories, oral histories, guided explanations, internal updates or material people may want to consume while doing something else.\n\nIt also brings responsibilities. Audio content needs planning, [editing](https://contentlexicon.com/editing), transcripts, summaries, ownership and a place in the wider [channel strategy](https://contentlexicon.com/channel-strategy). A 47-minute recording with no transcript, no useful notes and one point buried at minute 38 is an obstacle course with headphones.\n\n## How AI changes the picture\n\nSynthetic narration can offer another route through written source material, letting someone listen to a report or explanation they might otherwise struggle to make time for. AI can also help transcribe, summarise and edit recorded audio. The result still needs the rhythm and structure of spoken content, rather than a page read aloud without thought for the listener.\n\nListening can become responsive. An interactive audio experience can let someone interrupt to ask what an unfamiliar term means or request more detail before continuing. Google's December 2024 announcement of interactive Audio Overviews in NotebookLM demonstrated this direction: listeners could join an exchange with AI hosts to ask a question. It was an experimental feature, not evidence that every audio service can offer reliable conversation.\n\nRecorded human testimony and generated speech are different kinds of source, and listeners should be able to distinguish them. A real person's voice shouldn't be cloned without their consent, and synthetic voices need disclosure. Editorial checks and text alternatives remain part of the experience: someone should be able to read the material, check a claim and understand what they're hearing.\n\n## Questions people ask\n\n**Is audio mainly about podcasts?**\n\nPodcasts are one form. Audio also includes interviews, narration, audio articles, voice notes, recorded talks, guided experiences and voice interfaces.\n\n**Does audio need a transcript?**\n\nYes, for anything you publish: [WCAG](https://contentlexicon.com/wcag) expects a text alternative for recorded audio. Transcripts also support search, reuse, quotation and people who prefer to read.\n\n**Should every article have an audio version?**\n\nNot necessarily. It depends whether listening adds value for the audience and whether the organisation can maintain the format properly.\n\n**Can AI make audio content from written content?**\n\nTechnically yes, but that doesn't automatically make it good audio. Spoken content still needs rhythm, structure, editing and a reason to exist.\n\n## Learn more\n\n- [Making audio and video media accessible](https://www.w3.org/WAI/media/av/) – W3C guidance on transcripts, captions and accessible media.\n- [Campfire tales: The essentials of writing for radio](https://www.npr.org/sections/npr-training/2025/05/31/g-s1-65875/campfire-tales-the-essentials-of-writing-for-radio) – NPR Training on short sentences, pacing and letting the recorded voices do the work.\n- [NotebookLM gets a new look, audio interactivity and a premium version](https://blog.google/innovation-and-ai/models-and-research/google-labs/notebooklm-new-features-december-2024) – Google's December 2024 announcement shows listeners joining an Audio Overview to ask its AI hosts for explanation, with the feature's experimental limitations made explicit." }, { "term": "Author experience", "alsoCalled": [], "slug": "author-experience", "definition": "What it is like to create, edit, approve and maintain content inside an organisation’s systems.", "category": "Operations", "related": [ "cms", "content-editor", "content-operating-model", "content-operations", "decentralised-governance", "hub-and-spoke", "workflow" ], "updated": "2026-10-07T13:53:00.000Z", "url": "https://contentlexicon.com/author-experience", "markdown": "https://contentlexicon.com/author-experience.md", "credit": "Author experience, from the Content Lexicon by Contentious (https://contentlexicon.com/author-experience), CC BY 4.0", "body": "Author experience covers the [CMS](https://contentlexicon.com/cms), [workflows](https://contentlexicon.com/workflow), templates, guidance, permissions, previews, governance and all the small moments where a person either gets useful support or ends up hunting through a shared drive for last year's template.\n\nIt matters because every public content experience has an internal author experience behind it. If the people making content have to fight the tools, guess the rules, copy from old pages, wait for unclear approvals or publish without confidence, the reader eventually feels the consequences.\n\n## Why it matters for content\n\nAuthor experience affects content quality because content quality isn't produced by words alone. It's produced by people working inside systems. Those systems shape what authors can see, what they can change, what they understand, what they remember to check and how confidently they can publish. A good author experience makes structure clear, guidance available, ownership visible and common tasks simple. It reduces unnecessary decisions and supports the ones that need judgement.\n\nThis becomes especially important when content has to be maintained over time. Launching a [website](https://contentlexicon.com/website) with neat pages is one thing. Keeping those pages accurate, findable and current is quite another. If the [editing](https://contentlexicon.com/editing) process makes simple updates feel risky or mysterious, people avoid making them. Content then decays: old messages linger, broken links survive, [accessibility](https://contentlexicon.com/accessibility) slips and nobody is entirely sure which version to trust.\n\nAuthor experience also has a governance role. It turns rules into practical support. A [content model](https://contentlexicon.com/content-model), [style guide](https://contentlexicon.com/style-guide), workflow or approval process only helps if people can use it while doing the work. If guidance lives in one place, templates in another, previews somewhere else and ownership in someone’s head, authors are being asked to perform archaeology before breakfast. Better author experience brings the important things together at the point of need.\n\n## How AI changes the picture\n\nAI could make the author's working environment a place to explore decisions as well as enter content. An author might ask why a field exists, find the current source behind a claim or compare possible structures before choosing one. Help at the point of work could reduce hunting through guidance and leave more attention for editorial judgement. These are possible designs, not promises about what every CMS already provides.\n\nThe environment should let the author understand and challenge the assistance. If a tool suggests moving a qualification earlier, it should show the passage and explain the effect on the reader, leaving the author free to keep, adapt or reject the change. Comparing alternatives can build confidence without making the tool the final arbiter of good writing.\n\nSources, review and permissions still need to be clear. Which material is canonical? What must a person check, and where should uncertainty be flagged? Which standards apply, and who can approve publication? Those answers belong where the author is working, alongside the draft and its sources, so using AI doesn't require a separate expedition through the organisation's rules.\n\n## Questions people ask\n\n**Is author experience just CMS usability?**\n\nCMS usability is part of it, but author experience is broader. It includes the whole practical environment around content work: fields, templates, permissions, previews, workflow, guidance, governance, support, training and the messy handoffs between people. A usable CMS can still produce a poor author experience if the rules are unclear or the process is full of avoidable friction.\n\n**Why should readers care about what authors experience?**\n\nBecause readers inherit the results. If authors can't find the right guidance, preview the right layout, understand the required structure or update old content confidently, the reader is more likely to meet confusing, inconsistent or outdated content. Internal pain has external consequences.\n\n**Can AI fix a poor author experience?**\n\nNot by itself. AI can help authors draft, summarise, classify or check content, but it can't compensate for unclear ownership, weak governance or a system that makes the right action hard to find. In fact, AI may expose those weaknesses faster. A beautifully worded draft still needs accurate source material, sensible structure and someone with the judgement to decide whether it should exist.\n\n**When does author experience matter most?**\n\nWhenever content is complex, regulated, spread across teams or maintained over time. Organisations with many contributors, frequent updates, several channels or high-risk information feel it most. The more people involved in content, the more the system needs to help them make good decisions without turning every update into a small trial.\n\n## Learn more\n\n- [Designing Content Authoring Experiences](https://authoringexperience.com/) – Greg Dunlap's book and companion site on designing CMS interfaces and workflows around the people who use them.\n- [Designing Authoring Experiences That Editors Don’t Hate](https://www.lullabot.com/articles/designing-content-authoring-experiences-editors-dont-hate) – Greg Dunlap explores editorial interfaces, including dashboards and guidance adapted to authors’ experience.\n- [Embark: Dynamic documents for making plans](https://www.inkandswitch.com/embark) – Ink & Switch's travel-planning research prototype combines freeform notes, structured data and interactive views, suggesting how tools might support thinking within a document. It's adjacent research, not a CMS or an AI product." }, { "term": "Automated testing", "alsoCalled": [ "Test automation" ], "slug": "automated-testing", "definition": "Repeatable checks that software runs automatically to flag defined problems in code or content.", "category": "AI and tooling", "related": [ "content-health-check", "continuous-integration", "copy-editing", "human-in-the-loop", "linting", "readability", "system-one-models" ], "updated": "2026-10-06T20:54:00.000Z", "url": "https://contentlexicon.com/automated-testing", "markdown": "https://contentlexicon.com/automated-testing.md", "credit": "Automated testing, from the Content Lexicon by Contentious (https://contentlexicon.com/automated-testing), CC BY 4.0", "body": "Without automated tests, mistakes get found by a confused reader, an editor with a sinking feeling or a client with a screenshot. Developers use automated tests to check whether software still behaves as expected. [Content teams](https://contentlexicon.com/content-team) can borrow the idea.\n\nNot every content question can or should be automated. Whether an argument is persuasive, a metaphor is working or a page treats people with enough care usually needs human judgement. But many checks don't. Are all links working? Does each image have appropriate [alt text](https://contentlexicon.com/alt-text) or get marked as decorative? Is the page owner recorded? Is the [review date](https://contentlexicon.com/review-date) in the future? Does every statistic link to a source? Has a required field been left blank?\n\n## Why it matters for content\n\nAutomated testing matters because content quality isn't only a matter of taste. Some things can be checked consistently: broken links, missing [metadata](https://contentlexicon.com/metadata), duplicated headings, empty required fields, invalid dates, banned terms, reading age thresholds, [accessibility](https://contentlexicon.com/accessibility) basics and whether a page still matches agreed structural rules.\n\nThe value is partly speed and partly consistency. A machine can run the same checks every time content changes without becoming tired, distracted or weirdly invested in the semicolon debate. That frees editors and [content designers](https://contentlexicon.com/content-designer) to spend more time on judgement, meaning and usefulness.\n\nAutomated testing becomes especially useful when content is large, structured, regularly updated or business-critical. A small brochure site may survive with a manual checklist. A large content estate, a [knowledge base](https://contentlexicon.com/knowledge-base), a product catalogue or an AI source library needs repeatable checks, because otherwise decay is inevitable.\n\n## How AI changes the picture\n\nOne help article can now feed a chatbot answer, a summary [email](https://contentlexicon.com/email) and four translations in an afternoon. If that article carries a broken link or a product name that was retired in the spring, the mistake turns up in all of them at once. Teams that generate, summarise and repurpose content with AI need checks that run as fast as the production does, on the routine things that shouldn't be left to memory.\n\nAI can help with more nuanced checks too: comparing a draft against a [style guide](https://contentlexicon.com/style-guide), spotting unsupported claims, identifying likely duplication, checking whether a summary preserves the meaning of the source, or flagging where generated content has drifted from approved terminology. These checks aren't perfect and shouldn't pretend to be. They're useful first passes, not a replacement for editorial judgement.\n\nThere's also a source-quality issue. AI systems increasingly rely on content as input: policies, help articles, product data, [prompts](https://contentlexicon.com/prompt), examples and internal guidance. Automated testing can help keep those sources usable by checking freshness, completeness, links, metadata and structure before a chatbot starts quoting last year's refund policy.\n\n## Questions people ask\n\n**Is automated testing only for developers?**\n\nDevelopers gave us the language and many of the tools, but the principle is wider: define what should be true, then check it repeatedly. Content teams can apply the same principle to content quality, structure, accessibility, governance and maintenance.\n\n**Can content quality be automated?**\n\nSome of it can. A test can tell you whether a page has a review date, whether a link is broken or whether an image is missing alt text. It can't reliably tell you whether the argument is humane, persuasive or strategically right. Automate the checks that don't need judgement, so people have more attention for the ones that do.\n\n**Does automated testing replace a content review?**\n\nIt makes review better. Automated checks catch predictable problems before a human reviewer sees the work, which means the review can focus on substance, clarity, risk and usefulness instead of being dragged into preventable housekeeping.\n\n**Where should a team start?**\n\nStart with the five checks that are easiest to define and most annoying to repeat, such as broken links, missing alt text, missing owners, overdue review dates and empty required fields. If a check starts an argument every time it runs, it's probably not your first test.\n\n## Learn more\n\n- [Vale](https://docs.vale.sh/) – a prose linter that checks writing against your own style rules and can fail a build when a rule is broken.\n- [Introduction to automated testing](https://developer.mozilla.org/en-US/docs/Learn_web_development/Extensions/Testing/Automated_testing) – MDN on automating cross-browser tests with task runners and browser-testing services, assuming HTML, CSS and JavaScript knowledge.\n- [Continuous Documentation Deployment](https://docs.readthedocs.com/platform/stable/continuous-deployment.html) – Read the Docs on automatic documentation builds, pull-request previews and versioning alongside software releases." }, { "term": "Backend", "alsoCalled": [ "Back end" ], "slug": "backend", "definition": "Everything behind the scenes that stores, shapes and supplies the content people meet in a website, app or other channel.", "category": "Structure and systems", "related": [ "api", "cms", "content-model", "content-operating-model", "database", "frontend", "headless-cms", "metadata", "structured-content", "workflow" ], "updated": "2026-10-11T06:23:00.000Z", "url": "https://contentlexicon.com/backend", "markdown": "https://contentlexicon.com/backend.md", "credit": "Backend, from the Content Lexicon by Contentious (https://contentlexicon.com/backend), CC BY 4.0", "body": "In software, the backend is the server side of a [website](https://contentlexicon.com/website) or app: the [databases](https://contentlexicon.com/database), [APIs](https://contentlexicon.com/api), business logic, authentication and integrations that store information and decide what gets sent to the screen. A service directory might keep opening hours, locations and eligibility rules in one system, then supply them to a website, a mobile app and a chatbot. The [frontend](https://contentlexicon.com/frontend) is where people read and act on that information.\n\nFor content, the backend reaches well beyond the server. Behind any page sit the [content model](https://contentlexicon.com/content-model) and [metadata](https://contentlexicon.com/metadata) that give it shape, the [taxonomy](https://contentlexicon.com/taxonomy) that places it, the [workflow](https://contentlexicon.com/workflow) and approvals it passes through, the [style guide](https://contentlexicon.com/style-guide) and source knowledge it draws on and the people accountable for keeping it right. Ann Rockley made a similar division in 2016, separating frontend [content strategy](https://contentlexicon.com/content-strategy), which concerns audience, needs and the content itself, from backend strategy, which concerns structure, modularity, metadata, repositories and the rules for assembling content automatically.\n\n## Why it matters for content\n\nMany problems readers notice on the frontend begin in the backend. An out-of-date price on a page is rarely a writing failure. More often nobody owned the figure, or its [review date](https://contentlexicon.com/review-date) lived in a spreadsheet the [CMS](https://contentlexicon.com/cms) couldn't see, or it had been typed into six pages instead of stored once. Correcting the page deals with one instance; storing the number once, with an owner and a review date, deals with the cause.\n\nSeen this way, a large part of a [content operating model](https://contentlexicon.com/content-operating-model) is content backend. In the Contentious model, a content layer answers seven practical questions on top of two organisational foundations: an essence layer of purpose, character and relationships, and a knowledge and opinion layer of frames, data and beliefs. Those foundations steer content without appearing on any page as such. Several of the content layer's answers sit behind the scenes too: where content lives and how it connects to everything else; when it's reviewed and retired; who decides and who's accountable; and the processes, guidelines, systems and tools that keep the work running. Designing all of that is backend work, even if nobody involved writes a line of server code.\n\nRockley's argument for coordination applies across the whole of it. Decisions about what people need can produce content that serves them well but has to be rebuilt by hand for every channel and every update. Structural decisions made without that knowledge can produce tidy systems holding nothing anyone wants to read. That's why content people need a say in backend choices. If a warning has its own field, the system can validate it, reuse it and show it where it's needed; buried in a general Body field, it's much harder to manage. Rockley puts workflow governance in the overlap between the two kinds of strategy, and deciding who creates, reviews, changes and publishes content does belong to both.\n\n## How AI changes the picture\n\nAI brings the backend closer to the reader. A [knowledge-base](https://contentlexicon.com/knowledge-base) assistant answers from the sources, fields and search services it can reach, not from the polished page. Separate fields for status and eligibility can help it pick the right material, provided its retrieval rules use them and permissions are enforced; storing an owner or a date doesn't make an answer trustworthy by itself. Content that once only had to look right on a page now has to make sense when it's pulled apart, recombined and quoted out of context.\n\nThe wider backend changes role too. [Voice](https://contentlexicon.com/voice) guides, briefs, rubrics, approval rules and audience research used to be reference documents for people. Increasingly they become instructions, evaluation criteria and [guardrails](https://contentlexicon.com/guardrails) for agents that draft, check or route content, which makes their gaps visible quickly. A style guide that never settled how to describe a sensitive service will produce inconsistent drafts as fast as the agent can write them. Instructions describe the behaviour wanted, while permissions, validation and approval gates enforce it, so both need designing as deliberately as the prose.\n\nAI can also help maintain the backend. It can compare records, suggest classifications, propose metadata and flag possible duplicates: a team could use it to find conflicting opening hours across a service directory, then check each suggestion against the service itself. Missing facts still need collecting from the people who hold them; no model or content structure supplies evidence by itself.\n\n## Questions people ask\n\n**Do content people need to understand backend technology?**\n\nEnough to ask good questions. You don't need to write server code, but it helps to know where content is stored, which fields exist, how it's reused and what the system makes easy or awkward. Much of the content backend is yours already: models, taxonomies, workflows and style guides are content work.\n\n**Where does the CMS fit?**\n\nA CMS is one part of the backend, usually the part authors see. Around it there may be databases, APIs, search indexes, asset libraries, translation systems and integrations with other services, as well as the non-technical backend of ownership, rules and guidance. In a headless setup, the CMS stores and manages content while separate frontends display it.\n\n**Who looks after the content backend?**\n\nUsually several people, which is part of the challenge. Developers run the servers, [content designers](https://contentlexicon.com/content-designer) or strategists shape models and taxonomies, editors keep the style guide, subject experts own the facts and someone in [content operations](https://contentlexicon.com/content-operations) keeps the workflows running. Rockley suggests looking for backend skills in less obvious places: technical communicators, learning designers and document managers have long worked with modular, reusable content and rich metadata.\n\n**Where does the backend end and the frontend begin?**\n\nThe line moves with the question. A heading level is chosen in the content model and felt on the page; audience research shapes both what's written and how it's structured. Treat the distinction as a way of seeing two kinds of decision rather than a wall between two teams.\n\n## Learn more\n\n- [Why You Need Two Types of Content Strategist](https://contentmarketinginstitute.com/content-marketing-strategy/why-you-need-two-types-of-content-strategist) – Ann Rockley's case for pairing frontend strategy with \"intelligent content\" strategy behind the scenes, with the questions each asks and where to find backend skills inside an organisation.\n- [Introduction to the server side](https://developer.mozilla.org/en-US/docs/Learn_web_development/Extensions/Server-side/First_steps/Introduction) – MDN's plain-language account of what server-side code does and how it differs from code running in the browser; the technical sense of backend.\n- [The content operating model in the AI age](https://contentious.ltd/resources/the-content-operating-model-in-the-ai-age) – the Contentious model of organisational foundations and a content layer, much of which sits behind the interface." }, { "term": "Branch", "alsoCalled": [], "slug": "branch", "definition": "A separate line of work for developing related changes before they're reviewed and brought into the main version or abandoned.", "category": "Operations", "related": [ "a-b-testing", "git", "versioning" ], "updated": "2026-10-07T13:53:00.000Z", "url": "https://contentlexicon.com/branch", "markdown": "https://contentlexicon.com/branch.md", "credit": "Branch, from the Content Lexicon by Contentious (https://contentlexicon.com/branch), CC BY 4.0", "body": "Developers use branches to try work in parallel: one person fixes a bug, another tests a new feature and the main version carries on until the changes are ready to merge.\n\nContent work often needs the same idea. A team may want to rewrite guidance, test a new structure, prepare a policy update or experiment with AI-generated variants without breaking the content that people are already using. In a controlled publishing [workflow](https://contentlexicon.com/workflow), branching keeps proposed changes separate from the live version until they're ready. The draft can be questioned, changed, reviewed and, if necessary, shown the door.\n\n## Why it matters for content\n\nBranching matters because content change is rarely as simple as [editing](https://contentlexicon.com/editing) one page. A new service description may affect navigation, help text, [emails](https://contentlexicon.com/email), [PDFs](https://contentlexicon.com/pdf), product data, support scripts and internal guidance. If all those changes happen directly in the live estate, the team has very little room to test the shape of the work before readers meet it.\n\nA branch gives people somewhere safer to do messy thinking. With review and publishing controls in place, it helps protect live content while making collaboration easier: reviewers can see the proposed change as a coherent version, not as scattered comments across a dozen documents. It also helps teams compare before and after, decide what should be merged and keep a record of why the change was made.\n\nFor [content operations](https://contentlexicon.com/content-operations), branching is especially useful during migrations, redesigns, [structured-content](https://contentlexicon.com/structured-content) changes and major rewrites. These are the moments when people need freedom to try something properly without treating the public service as the rehearsal room.\n\n## How AI changes the picture\n\nAI has made experiments cheap. A team can ask an AI tool to propose alternative structures, rewrite a set of pages into a new pattern or classify existing content against a model. That can be helpful, but it belongs somewhere separate from the live estate until the work has been checked.\n\nBranching also gives AI-assisted work a clearer review boundary. Instead of asking “what did the tool change?” after the fact, a branch lets people inspect a proposed set of changes before they become authoritative. That matters because generated content can be plausible, inconsistent or wrong in small ways that are hard to spot one paragraph at a time.\n\nAI doesn't remove the need for controlled change; it increases it. Teams need places where change can be explored, tested and reviewed before it reaches users, source systems or other machines that may reuse it.\n\n## Questions people ask\n\n**Is a branch the same as a draft?**\n\nNot quite. A draft is usually one piece of unfinished content. A branch keeps related changes together on a separate line of work, which might cover a page, feature, content set or codebase.\n\n**Do content teams need Git to use branches?**\n\n[Git](https://contentlexicon.com/git) is one technical implementation of branching. Teams can apply the working principle through [staging](https://contentlexicon.com/staging) sites, [CMS](https://contentlexicon.com/cms) versions, duplicated Notion pages, sandbox [databases](https://contentlexicon.com/database) or controlled drafts, though those tools don't necessarily provide Git-style branches or merging. Git or no Git, everyone needs to know which version is live and which is experimental.\n\n**What happens when a branch is finished?**\n\nIt's reviewed and either merged into the main version, revised further or abandoned. Abandoning a branch isn't failure; it's one of the reasons branches exist.\n\n**What can go wrong?**\n\nBranches become dangerous when nobody knows which version is current, when the branch drifts too far from the live version or when changes are copied across by hand without checking. Branching needs naming, ownership and a habit of closing loops.\n\n## Learn more\n\n- [Branches in a Nutshell](https://git-scm.com/book/en/v2/Git-Branching-Branches-in-a-Nutshell) – the official Git book’s visual explanation of branches.\n- [Branches](https://docs.github.com/en/pull-requests/reference/branches) – GitHub guidance on using branches with collaborative review.\n- [Docs as Code](https://www.writethedocs.org/guide/docs-as-code/) – Write the Docs on documentation teams working with developers' tools, from version control to review before anything merges." }, { "term": "Business case", "alsoCalled": [], "slug": "business-case", "definition": "The argument for why a piece of work is worth doing, including the value, cost, risk and expected outcome.", "category": "Foundations", "related": [ "content-operating-model", "content-strategy" ], "updated": "2026-10-06T20:57:00.000Z", "url": "https://contentlexicon.com/business-case", "markdown": "https://contentlexicon.com/business-case.md", "credit": "Business case, from the Content Lexicon by Contentious (https://contentlexicon.com/business-case), CC BY 4.0", "body": "A business case lets a decision be made on more than enthusiasm and a hopeful spreadsheet. For content work, a business case might support a new [website](https://contentlexicon.com/website), a [content audit](https://contentlexicon.com/content-audit), a [style guide](https://contentlexicon.com/style-guide), a governance review, a [CMS](https://contentlexicon.com/cms) migration or an AI-assisted content programme.\n\nA good business case connects content work to organisational priorities. “We need a [content model](https://contentlexicon.com/content-model)” may be true, but “we need a content model so product information can be reused accurately across the website, sales material and support tools” is easier to understand and much harder to dismiss.\n\n## Why it matters for content\n\nContent work is often undervalued because its costs and benefits are poorly described. Ask for “some words for the new site” and the request gets budgeted like stationery. A business case shows what the work is for instead: fewer calls to the helpline about the same confusing form, a CMS licence the organisation is already paying for put to proper use, an editor's Fridays no longer spent fixing the same broken pages.\n\nYou'll want one when the work needs budget, senior attention or cooperation from other teams. It also helps when content problems are being felt somewhere else: in the support queue, in a sales team rewriting product descriptions for every pitch, or in staff time lost to the same recurring mess.\n\n## How AI changes the picture\n\nAI changes the business case for content because it changes both the opportunity and the risk. Better content foundations can make AI-assisted work more reliable: clearer sources, stronger guidance, better [metadata](https://contentlexicon.com/metadata) and more structured material. Weak content foundations can make AI cheaper to use and more expensive to live with.\n\nSo a business case for AI-era content has to count more than the tooling. The licence is the easy line. The harder ones are the editor who reviews what the tool drafts, the owner who keeps source pages current and the answer that goes wrong. A support assistant quoting an eligibility rule that changed last April generates complaints, corrections and staff time that never appear in the software budget.\n\n## Questions people ask\n\n**Is a business case just about money?**\n\nMoney is part of it, but a decision-maker also wants to know what happens if nothing changes, what else you considered and what will be different afterwards.\n\n**Do small content projects need a business case?**\n\nNot always formally, but even small projects benefit from a clear reason for doing the work and a shared view of what success looks like.\n\n**What makes a content business case persuasive?**\n\nA clear problem, evidence that it matters, a practical proposed response and a link to organisational goals people already care about.\n\n**Can the benefit of content work be measured?**\n\nOften, but not always in one neat number. Useful evidence might be analytics, user research, support tickets tagged to a confusing page, [accessibility](https://contentlexicon.com/accessibility) findings or hours of staff time saved.\n\n## Learn more\n\n- [Business case basics](https://www.apm.org.uk/resources/what-is-project-management/what-is-a-business-case) – APM on comparing benefits, costs and risks across options, including doing nothing, and revisiting the case at decision gates. The explanation is public; templates and additional resources are member-only.\n- [Content Strategy 101](https://www.nngroup.com/articles/content-strategy/) – useful grounding for the organisational value behind content investment.\n- [How to make the case for content investment](https://contentmaturity.com/learn/make-the-case-for-content) – how quantified capability gaps, business outcomes and colleagues’ own words can turn advocacy into evidence." }, { "term": "Call to action", "alsoCalled": [ "CTA" ], "slug": "call-to-action", "definition": "The specific next step content invites the reader to take.", "category": "Editorial craft", "related": [ "content-design", "copywriting", "microcopy" ], "updated": "2026-10-11T06:23:00.000Z", "url": "https://contentlexicon.com/call-to-action", "markdown": "https://contentlexicon.com/call-to-action.md", "credit": "Call to action, from the Content Lexicon by Contentious (https://contentlexicon.com/call-to-action), CC BY 4.0", "body": "A call to action (CTA for short) might invite someone to check eligibility, book a place or read the next article. Its wording should tell them what the next step actually is.\n\nTiming matters too. An invitation to book is premature if the page hasn't explained the price or who the event is for. A good CTA follows the reader's decision rather than trying to hurry it.\n\n## Why it matters for content\n\nCalls to action are central to making content actionable. When a page is meant to support an action, leaving the next step unclear can prevent it from doing its job. An essay may instead be complete when it gives someone a new way to think. This matters in service content, fundraising, product journeys, support material, event pages and campaign content – anywhere the reader is meant to move from understanding to action.\n\nBad CTAs are usually vague, premature or selfish. “Learn more” tells the reader very little. “Submit” is a bureaucratic shrug. “Get started today” can be fine, unless the page has not yet explained what “started” involves or why anyone should be getting. The words need to match the task, the reader’s state of mind and the actual thing that will happen.\n\nThe best CTAs aren't decoration added at the end. They are part of the content’s argument. If the desired action is unclear, the page probably has a deeper purpose problem. If the action is clear but the CTA is weak, the content may fail at the final inch.\n\n## How AI changes the picture\n\nAI can produce many CTA variants very quickly, which is useful for exploration and testing. It can also check whether button labels, links and headings are consistent across a journey. Across a service, that comparison can reveal labels that promise different destinations for the same action.\n\nThe risk is that AI writes CTAs that sound smooth but ignore context. A model may suggest “Start your journey” when the task is to renew a licence, report a bereavement or check eligibility for support. That isn't a [tone](https://contentlexicon.com/tone) issue; it's a failure of judgement. CTAs have to respect what the user is doing and what may be at stake.\n\nAI also makes calls to action matter beyond the page. Generated summaries, chat assistants and answer engines may recommend next steps from your content. If the source page is vague about what to do, the machine may be vague too, or infer a step the source never authorised.\n\n## Questions people ask\n\n**Is a CTA always a button?**\n\nNo. A CTA can be a button, link, form [prompt](https://contentlexicon.com/prompt), phone number, [email](https://contentlexicon.com/email) instruction, next article or any clear invitation to act. The format should follow the task, not the other way round.\n\n**How many CTAs should a page have?**\n\nFor a straightforward booking page, one primary action may be enough, with a secondary link for access questions. A service page supporting different circumstances may need several routes. Make their priority clear and test whether people can choose the right one.\n\n**What makes a CTA effective?**\n\nThe label, timing and destination should agree. ‘See prices’ needs to reveal prices, not first require an email address. Test whether people understand the action and whether it delivers what it promises.\n\n**Should we avoid “learn more”?**\n\nSometimes it's defensible, but more specific link text almost always helps: “Read the eligibility guidance”, “See ticket prices” or “Compare the plans” gives the reader an actual clue.\n\n## Learn more\n\n- [“Learn more” is not enough](https://uxcontent.com/learn-more-is-not-enough-a-case-for-descriptive-cta/) – the case for labels that describe the destination or action.\n- [Let’s do this: how to write better calls to action](https://medium.com/deliveroo-design/lets-do-this-how-to-write-better-calls-to-action-ea534768b599) – A Deliveroo design account connecting action verbs with movement and outcomes; an interface case, not current product documentation.\n- [Advocates not parrots](https://contentious.ltd/thoughts/advocates-not-parrots) – why meaningful engagement asks more of content than another button click or petition signature." }, { "term": "Card sorting", "alsoCalled": [], "slug": "card-sorting", "definition": "A research method where people group and label items to reveal how they think about them.", "category": "Research and evidence", "related": [ "design-thinking", "findability", "information-architecture", "taxonomy", "user-needs", "user-testing" ], "updated": "2026-10-11T06:23:00.000Z", "url": "https://contentlexicon.com/card-sorting", "markdown": "https://contentlexicon.com/card-sorting.md", "credit": "Card sorting, from the Content Lexicon by Contentious (https://contentlexicon.com/card-sorting), CC BY 4.0", "body": "A card sort shows how people think about information: what they expect to find together and what language they naturally use for it.\n\nIn content and [user experience](https://contentlexicon.com/ux) (UX) work, card sorting is often used to shape navigation, [information architecture](https://contentlexicon.com/information-architecture), [taxonomy](https://contentlexicon.com/taxonomy) and category labels. Participants might be given cards representing pages, topics, services or concepts, then asked to organise them into groups. In an open card sort they create the groups themselves. In a closed card sort they sort items into existing categories.\n\n## Why it matters for content\n\nCard sorting matters because organisations are very good at organising content around themselves. Teams, departments, funding streams, internal acronyms and legacy structures all creep into navigation and labels. Users arrive with their own mental models. Card sorting helps reveal the difference before the structure is set in concrete.\n\nThe value is evidence: a card sort shows patterns, surprises and disagreements. It can reveal that two services people inside the organisation treat as separate are understood as one thing by users, or that an internal label has no useful meaning outside the building.\n\nCard sorting is also useful in [voice](https://contentlexicon.com/voice) and [tone](https://contentlexicon.com/tone) work, where sorting words, examples or qualities can help a group move from abstract preference to shared language. People often find it easier to compare and arrange examples than to answer a grand question like “what should we sound like?”, which tends to summon adjectives in their least useful form.\n\n## How AI changes the picture\n\nAI can make card sorting faster to prepare and easier to analyse. It can turn a [content inventory](https://contentlexicon.com/content-inventory) into candidate cards, cluster open-sort results, spot recurring labels and summarise where participants agreed or diverged. Used carefully, that gives researchers more time to interpret what the patterns mean.\n\nBut AI shouldn't replace the research itself. A card sort exists to show how real people make sense of information, not how a model predicts they might. AI can help with the paperwork around the method; it can't stand in for the moment when a participant hesitates, moves a card twice and explains what felt wrong.\n\nThe outputs travel further now, too. The groupings and labels a card sort produces can feed [content models](https://contentlexicon.com/content-model), [metadata](https://contentlexicon.com/metadata), search and AI retrieval as well as navigation, so a good sort can help people and machines find the same things.\n\n## Questions people ask\n\n**Is card sorting only for navigation?**\n\nNavigation is a common use, but card sorting can also help with taxonomy, content grouping, terminology, voice and tone, product categories and any situation where shared meaning matters.\n\n**What's the difference between open and closed card sorting?**\n\nIn an open sort, participants create and name their own groups. In a closed sort, they place items into categories you already provide. Open sorts are better for discovery; closed sorts are better for testing an existing structure.\n\n**How many people do you need?**\n\nFewer than you might fear. Nielsen Norman Group suggests at least 15 participants for a qualitative sort, and 30 to 50 if you want numbers you can generalise from. A small, well-chosen group can reveal useful patterns, especially alongside other evidence.\n\n**Does card sorting give you the final structure?**\n\nIt gives you evidence for one. You still need judgement, content knowledge, business context and sometimes the courage to ignore one very enthusiastic participant’s category called “miscellaneous vibes”.\n\n## Learn more\n\n- [Card Sorting: Uncover Users’ Mental Models](https://www.nngroup.com/articles/card-sorting-definition/) – Nielsen Norman Group guidance on running and interpreting card sorts.\n- [Card Sorting: A Definitive Guide](https://boxesandarrows.com/card-sorting-a-definitive-guide/) – Donna Spencer and Todd Warfel's guide to preparing, running and analysing a sort.\n- [Tree Testing: Evaluate Menu Labels and Categories](https://www.nngroup.com/articles/tree-testing/) – a useful companion method for testing whether a proposed structure works." }, { "term": "Centralised governance", "alsoCalled": [ "Centralised model", "Centralised publishing" ], "slug": "centralised-governance", "definition": "A governance model in which one central team plans, creates, approves and publishes most of an organisation's content.", "category": "Governance", "related": [ "content-governance", "content-team", "decentralised-governance", "head-of-content", "hub-and-spoke", "sign-off", "silo", "style-guide", "subject-matter-expert" ], "updated": "2026-10-14T08:13:00.000Z", "url": "https://contentlexicon.com/centralised-governance", "markdown": "https://contentlexicon.com/centralised-governance.md", "credit": "Centralised governance, from the Content Lexicon by Contentious (https://contentlexicon.com/centralised-governance), CC BY 4.0", "body": "Centralised governance puts the authority to publish, and most of the work of producing content, in one place. A single team, usually in communications, marketing or digital, owns the [website](https://contentlexicon.com/website) and the other main channels. Colleagues elsewhere bring requests, information and approvals, and the central team plans, writes, edits and publishes. In a small organisation that team might be one person with an inbox; in a large one it can run like an in-house agency, with editors, designers and specialists in different formats.\n\nIt sits at one end of a spectrum, with [decentralised governance](https://contentlexicon.com/decentralised-governance) at the other and [hub and spoke](https://contentlexicon.com/hub-and-spoke) in between. Centralisation can also be partial: an organisation may centralise its platform, its standards or its [sign-off](https://contentlexicon.com/sign-off) without centralising who writes. The UK government's move to [GOV.UK](http://GOV.UK) in the early 2010s, which closed hundreds of departmental and agency websites, is often cited as centralisation on a grand scale. Yet most of the pages on the new site are still written and maintained by other departments, while the central team looks after the few thousand that most citizens and businesses need.\n\n## Why it matters for content\n\nThe appeal is coherence. When one team sees the whole estate, it can spot duplication, keep one [voice](https://contentlexicon.com/voice) and make sure every page meets the same standards for [accessibility](https://contentlexicon.com/accessibility), accuracy and structure. Accountability is easy to trace: when something is wrong, everyone knows whose job it is to fix it. Specialists get colleagues who share their craft, people to review their work and a career path that doesn't depend on being the only content person in a department.\n\nThe costs come from distance and volume. Every request joins a queue, so a small correction can wait behind a campaign, and colleagues who want something published quickly learn to route round the team with a separate [microsite](https://contentlexicon.com/microsite), a [PDF](https://contentlexicon.com/pdf) on a shared drive or a social account nobody mentioned. The central team also knows less about each subject than the people who work on it, so drafts travel back and forth for checking, and content comes to be treated as the final polish rather than part of how a service is designed. Lisa Welchman makes the case against a web team that decides and runs everything: it becomes a production [silo](https://contentlexicon.com/silo), cut off from the business it's meant to serve and hard to size properly.\n\nCentralisation tends to suit organisations with a modest volume of content, serious legal or reputational risk or an estate that has grown beyond anyone's control. Many centralise for a period to restore order after years of uncoordinated publishing, then hand more work back to other teams once standards and ownership are settled.\n\n## How AI changes the picture\n\nAI tools ease the central team's oldest problem, which is capacity. Drafting from a structured brief, summarising a subject expert's notes, checking a page against the [style guide](https://contentlexicon.com/style-guide) and flagging pages due for review all shorten the time each request takes, so a small team can handle more without lowering its standards. The bottleneck moves rather than disappears. Someone still has to decide what's worth publishing, confirm the facts with the people who know them and answer for the result.\n\nThey also make the central team easier to bypass. When anyone in the organisation can produce a plausible page, briefing or chatbot answer in minutes, requests that once came to the team can go round it altogether. The team’s authority comes to depend less on being the only people able to write and more on what it provides to everyone else: approved sources, tested instructions and a publishing route that colleagues trust.\n\n## Questions people ask\n\n**Does centralisation mean nobody else can publish?**\n\nUsually it means nobody else publishes without the central team. Some organisations give a few trusted colleagues [editing](https://contentlexicon.com/editing) rights for news or events, while the central team keeps control of structure, navigation and the most visited pages. What matters is where the final decisions sit.\n\n**How do you stop a central team becoming a bottleneck?**\n\nTriage requests by value and risk, publish the criteria and say no clearly when something doesn't need a page. Templates, self-service routes for routine updates and a named contact for each part of the organisation all help. If the queue keeps growing anyway, the organisation may have outgrown the model.\n\n**When does it make sense to centralise?**\n\nWhen content has sprawled beyond repair, when mistakes carry legal or safety consequences or when the volume is small enough for one team to manage well. It's a harder case for organisations with many specialist services that change often.\n\n**What does the rest of the organisation do?**\n\nIt stays responsible for what's true. Subject experts supply knowledge, check accuracy and tell the central team when something changes. Centralisation works badly when colleagues assume that handing over the writing also hands over responsibility for the facts.\n\n## Learn more\n\n- [300+ websites to just 1 in 15 months](https://insidegovuk.blog.gov.uk/2014/12/19/300-websites-to-just-1-in-15-months/) – Mark Hazelby's 2014 account of moving more than 300 government organisations onto one website; the comments add a pointed critique and a reply explaining that most of the 150,000 pages stayed with the teams that wrote them.\n- [Managing Chaos: frequently asked questions](https://rosenfeldmedia.com/books/managing-chaos-frequently-asked-questions/) – short answers from Lisa Welchman's book on digital governance, including why letting one web team decide and run everything creates a production silo." }, { "term": "Channel strategy", "alsoCalled": [], "slug": "channel-strategy", "definition": "A plan for where content should appear, what each channel is for and how the channels work together.", "category": "Foundations", "related": [ "audience-segmentation", "audio", "content-marketing", "content-strategy", "create-once-publish-everywhere", "email", "message-architecture", "microsite", "social-media", "website", "workflow" ], "updated": "2026-10-10T17:52:00.000Z", "url": "https://contentlexicon.com/channel-strategy", "markdown": "https://contentlexicon.com/channel-strategy.md", "credit": "Channel strategy, from the Content Lexicon by Contentious (https://contentlexicon.com/channel-strategy), CC BY 4.0", "body": "In practice, channel strategy decides what belongs on the [website](https://contentlexicon.com/website), in [email](https://contentlexicon.com/email), on [social media](https://contentlexicon.com/social-media), in search, in events, in print, in [audio](https://contentlexicon.com/audio), in video or in whatever platform everyone has just become strangely excited about.\n\nAn organisation rarely needs to be everywhere, which is often just being exhausted in public. A useful channel strategy asks which audience segments are in which places, what they need there, what kind of relationship the channel supports and what the organisation can realistically maintain.\n\n## Why it matters for content\n\nChannel strategy stops content being treated as one message pushed everywhere. Different channels do different (but usually connected) jobs. A website may be the stable source of truth, email may nurture an existing relationship, social media may support visibility or conversation and a report may hold depth and evidence. The strategy should make those roles explicit so content isn't copied and pasted across every surface.\n\nIt matters most when a team is publishing across several places, when effort is spread thinly or when content is being made for channels out of habit rather than because they serve a clear role. A good channel strategy protects attention, both for the audience and the team.\n\n## How AI changes the picture\n\nAI makes it much easier to adapt one idea for many channels, which is useful and dangerous. Without a plan, you get the same weak thing in six formats. With one, AI can help reshape content for the job each channel is meant to do: a summary for email, a canonical page for search, a short post for discovery or a script for audio.\n\nAI is also a route through which people encounter content. Someone might meet a research finding in an assistant's answer, visit the website to check the evidence behind it and return to an email series for continuing interpretation. A useful strategy decides what each of those encounters contributes and how the source stays reachable. This is an illustrative design, not a claim that every assistant will send traffic back.\n\nFast repurposing needs firmer governance. When one announcement can become twelve posts before lunch, teams need clearer rules about which channel holds the source, what can be adapted, what must stay consistent and what needs human review before it goes out.\n\n## Questions people ask\n\n**Is channel strategy the same as content strategy?**\n\n[Content strategy](https://contentlexicon.com/content-strategy) is broader. Channel strategy is the part that decides where content should live, travel and be adapted.\n\n**Is this just distribution?**\n\nDistribution is getting content out. Channel strategy decides which channels matter, what role each one plays and how success should be judged.\n\n**Do we need to be on every channel?**\n\nNo. A good channel strategy often gives a team permission to do fewer channels better.\n\n**Should the same content appear everywhere?**\n\nUsually not in the same form. The underlying message may stay consistent, but the format, timing, depth and [call to action](https://contentlexicon.com/call-to-action) should fit the channel.\n\n## Learn more\n\n- [Optimizing for Context in the Omnichannel User Experience](https://www.nngroup.com/articles/context-specific-cross-channel/) – Nielsen Norman Group on shaping each channel around what people do there, rather than offering the same experience everywhere.\n- [Consistency in the Omnichannel Experience](https://www.nngroup.com/articles/omnichannel-consistency/) – the other half of the balance: keeping functionality, data and presentation consistent across channels.\n- [Overview and key findings of the 2026 Digital News Report](https://reutersinstitute.politics.ox.ac.uk/digital-news-report/2026/dnr-executive-summary) – the Reuters Institute’s June 2026 finding that social media and video networks have overtaken publishers’ own websites and apps as the most used source of online news, with AI chatbots growing from a small base." }, { "term": "CMS", "alsoCalled": [ "Content management system" ], "slug": "cms", "definition": "The software used to store, organise and publish content.", "category": "Structure and systems", "related": [ "author-experience", "content-editor", "content-model", "headless-cms", "microsite", "rollback", "technical-debt", "versioning", "website", "workflow" ], "updated": "2026-10-06T20:57:00.000Z", "url": "https://contentlexicon.com/cms", "markdown": "https://contentlexicon.com/cms.md", "credit": "CMS, from the Content Lexicon by Contentious (https://contentlexicon.com/cms), CC BY 4.0", "body": "CMS stands for content management system, and it's where templates, fields, [metadata](https://contentlexicon.com/metadata), permissions, [workflows](https://contentlexicon.com/workflow) and publishing decisions become concrete. WordPress, Drupal, Webflow, Sanity and Contentful are all CMSs, though they work in very different ways and suit very different teams.\n\nFor [content teams](https://contentlexicon.com/content-team), the CMS is more than a container. It shapes what kinds of content are easy to create, how reusable the content is, how structured it becomes and how safe the publishing process feels. Anyone who has pasted the same opening hours into forty pages, knowing they'll change again at Christmas, has met a CMS's opinions first-hand.\n\n## Why it matters for content\n\nA good CMS supports good [content operations](https://contentlexicon.com/content-operations). It helps teams manage structure, workflow, review, permissions and publishing in a way that fits the content rather than fighting it. A poor CMS (or a poorly implemented one) makes good habits harder: metadata gets skipped, structure gets bypassed and content is forced into shapes it was never meant to take.\n\nChoosing the product is only part of the decision. Content teams live with the choices made during configuration: which fields exist, what they are called, which are required, how preview works and whether the workflow matches the way publishing really happens. Two organisations can use the same software and give their editors completely different experiences.\n\n[Author experience](https://contentlexicon.com/author-experience) deserves the same attention as the public interface. Clear field names, useful guidance and sensible defaults make the right action easier. A cluttered form with fifty unexplained fields trains people to work around the system, usually by pasting everything into the one box they understand.\n\nThe difference shows wherever content is maintained over time rather than posted once and forgotten, and wherever several people need to create or manage content together.\n\n## How AI changes the picture\n\nConnect an AI tool to a CMS and you find out quickly how it was built. Systems with clear structure, useful metadata and reusable fields are easier to connect to assistants, retrieval tools and automated checks. Systems built around copying formatted text from page to page make that much harder.\n\nSo treat the CMS as a source of governed, machine-usable content rather than a place to paste finished words. The better the underlying structure, the more reliably AI can work with it.\n\nThat structure also points beyond the [website](https://contentlexicon.com/website). A publisher, for example, could supply approved explanations, author details and corrections to its own site and to a conversational reading service, with the CMS managing the source and its changes and the service providing the interaction. Not every CMS offers such a reader-facing service, but governed content is what would make one possible.\n\nAn [API](https://contentlexicon.com/api) does not make a CMS AI-ready by itself. The content behind it still needs meaningful fields, stable identifiers, clear relationships and a reliable way to tell current material from retired material. Otherwise the connection gives the AI access to the same duplication and ambiguity editors already deal with.\n\n## Questions people ask\n\n**Is a CMS just a website editor?**\n\nIt usually includes one, but it also manages structure, metadata, workflow, permissions and publishing.\n\n**Do content people need to understand the CMS deeply?**\n\nNot deeply, usually. But they need enough to understand how it shapes content, even if someone else configures the details.\n\n**Is a headless CMS the same thing?**\n\nNot quite. A [headless CMS](https://contentlexicon.com/headless-cms) is one kind of CMS: it manages content but leaves presentation to other systems, delivering content through APIs.\n\n**Can the wrong CMS damage content quality?**\n\nYes. If the system makes structure, review or reuse hard, quality tends to suffer no matter how good the intentions are.\n\n## Learn more\n\n- [CMS requirements](https://webproject.guide/cms-requirements/) – The Web Project Guide on working out what a CMS needs to do: features, infrastructure, cost and ecosystem.\n- [Choosing a CMS](https://webproject.guide/select-cms/) – the next chapter: vendor research, demonstrations and proofs of concept before committing to one.\n- [Karen McGrane on Content Management](https://theinformed.life/2024/10/20/episode-151-karen-mcgrane/) – Karen McGrane considers content separated from its containers, the consequences of AI remixing and how to judge whether content still serves its purpose." }, { "term": "Competitor analysis", "alsoCalled": [ "Competitive analysis" ], "slug": "competitor-analysis", "definition": "A review of similar organisations and useful comparators to understand the field, find gaps and identify practices to emulate.", "category": "Research and evidence", "related": [ "content-audit", "content-marketing", "content-strategy" ], "updated": "2026-10-11T06:23:00.000Z", "url": "https://contentlexicon.com/competitor-analysis", "markdown": "https://contentlexicon.com/competitor-analysis.md", "credit": "Competitor analysis, from the Content Lexicon by Contentious (https://contentlexicon.com/competitor-analysis), CC BY 4.0", "body": "Read side by side, similar organisations' content shows the common messages, dominant claims, useful gaps, repeated weaknesses and places where everyone appears to have copied the same sentence from the same haunted brochure.\n\nFor content work, competitor analysis isn't about imitation. Knowing what's already out there helps an organisation make better choices about positioning, evidence, [tone](https://contentlexicon.com/tone), structure, topics and service explanations. In charities, public services, education and purpose-driven organisations, the comparison set may include peers, partners, alternatives or organisations shaping expectations in the same space.\n\nComparator analysis is related, but slightly different. A competitor is usually a similar organisation offering something comparable to a similar audience, whereas a comparator doesn't need to be a usual suspect or even work in the same field. It's an organisation, product or service doing something well that you could learn from or adapt. A charity might study a retailer’s returns process, a museum’s wayfinding, a government service’s form design or a software company’s explanation of a complicated feature. The work is in examining that practice in context and asking whether the underlying choice could work for you.\n\n## Why it matters for content\n\nCompetitor analysis helps [content teams](https://contentlexicon.com/content-team) see what their audience may already have heard, where the organisation sounds indistinct and where there may be a more useful point of view to own. It can reveal generic promises, unsupported claims, missing explanations, tone problems and common patterns to adopt carefully or avoid deliberately.\n\nComparator analysis adds a different source of insight. It stops a team from treating the current sector as the only place to look. A similar organisation may show you the language your audience expects, while a well-designed service elsewhere may show you how to make a process clearer, reduce anxiety, explain a choice or help someone take action. Competitor analysis tells you about the field you're in. Comparator analysis can help you imagine a better way of working within it.\n\nBoth approaches matter when an organisation is repositioning, redesigning a [website](https://contentlexicon.com/website), launching a product or service, entering a crowded space or trying to explain why it's different. They also support practical editorial decisions: what proof to provide, what language to avoid, which topics need depth, what experience to improve and what the organisation can say with more credibility than others.\n\n## How AI changes the picture\n\nAI can summarise pages, cluster themes, compare sign-up flows and spot repeated claims, but it can also flatten distinctive differences into vague patterns if the question is poorly framed. That makes competitor and comparator analysis easier to start and easier to do badly. The analysis still needs human judgement about relevance, evidence, positioning and what the organisation should do next.\n\nAI can also widen the field of comparison. Instead of looking only at three familiar competitors, a team can ask it to find examples of organisations that explain complex services clearly, handle sensitive decisions well or make important information easy to act on. That can be useful for discovering comparators, provided the examples are checked properly and the team studies the actual source material rather than trusting a summary.\n\nMeanwhile, more of what competitors publish is being generated too, and it tends to converge. So ask less “what are others publishing?” and more “where is everyone starting to sound the same, and what can we learn from organisations that make this experience more specific, more useful or more true?”\n\n## Questions people ask\n\n**Is comparator analysis just copying another organisation?**\n\nOnly if it's done badly. The real work is understanding the principle behind a practice, then deciding whether it fits your audience, purpose, constraints and [voice](https://contentlexicon.com/voice). A different context may contain a useful lesson, but it rarely supplies a ready-made answer.\n\n**Who counts as a competitor?**\n\nNot only commercial competitors. It can include peers, alternatives, partners, substitutes and anyone shaping audience expectations in the same space.\n\n**What should we look at?**\n\nLook at messages, topics, structure, evidence, tone, [calls to action](https://contentlexicon.com/call-to-action), terminology, channels, key tasks and the questions organisations do or don't answer. For comparators, look especially closely at the specific practice you want to understand.\n\n**Can AI do competitor or comparator analysis for us?**\n\nIt can help gather, compare and summarise material, and it may suggest useful comparators. The conclusions still need strategic and editorial judgement, along with a check of the original evidence.\n\n## Learn more\n\n- [Competitive Usability Evaluations: Definition](https://www.nngroup.com/articles/competitive-usability-evaluations/) – Nielsen Norman Group on comparing your experience with others', including organisations only loosely related to yours.\n- [The Ultimate Guide to a Competitive Content Analysis (+ Template)](https://www.semrush.com/blog/competitive-content-analysis/) – a practical method for comparing themes, formats, quality and performance, leaning on Semrush's own tools." }, { "term": "Component", "alsoCalled": [], "slug": "component", "definition": "A reusable part – of content, design or code – made once and used in many places.", "category": "Structure and systems", "related": [ "atomic-content", "content-pattern", "design-system", "separation-of-concerns", "structured-content", "wireframe" ], "updated": "2026-10-09T10:28:00.000Z", "url": "https://contentlexicon.com/component", "markdown": "https://contentlexicon.com/component.md", "credit": "Component, from the Content Lexicon by Contentious (https://contentlexicon.com/component), CC BY 4.0", "body": "A component is a self-contained, reusable part with an agreed purpose, made once and used in many places. Standard parts made manufacturing faster and repairs simpler long before software borrowed the principle for buttons, cards, alerts and navigation.\n\nComponents can be parts of an interface, pieces of code, units of content or all three; what makes them components is that they can be combined and reused rather than recreated each time.\n\n## Why it matters for content\n\nInterface components shape expression. A warning banner needs a consequence and a next step; a story card needs a headline, a summary and somewhere to go next. Design them without content thinking and they may look tidy but fail when real words arrive, producing truncated meaning, awkward workarounds or an elegant box full of nonsense.\n\nContent components govern what is said rather than how it looks. A product description, disclaimer or eligibility rule can be stored with its own fields and [metadata](https://contentlexicon.com/metadata), then reused across pages, [emails](https://contentlexicon.com/email), apps and chatbot answers. When a disclaimer exists as one maintained component rather than forty separate copies, a single correction can reach every place that uses it. This supports [create once, publish everywhere](https://contentlexicon.com/create-once-publish-everywhere) and a [single source of truth](https://contentlexicon.com/single-source-of-truth).\n\nThe craft lies in choosing the right size. Topic-based documentation reuses whole topics; [atomic content](https://contentlexicon.com/atomic-content) works with smaller pieces. Make components too large and reuse becomes difficult; make them too small and they lose the context that makes them meaningful.\n\nComponents also give teams a shared vocabulary: “use the campaign card” is clearer than “make that bit near the top with the blue box and the link”. Agreed names and guidance make discussion and testing easier, while leaving room for the variation the content needs.\n\n## How AI changes the picture\n\nAI makes components more important because reusable structures give machines something dependable to work with. If a product card has separate fields for name, summary, price, eligibility and [call to action](https://contentlexicon.com/call-to-action), an AI system can check, summarise or repurpose that information more reliably than if it's buried in one block of prose. Content components go further: an assistant that answers questions by drawing on approved, owned components – the current eligibility rule, the agreed wording of a refund policy – has much less room to improvise than one paraphrasing whatever pages it happens to find.\n\nAI can also help teams test components against real content. It can generate edge cases, flag missing fields, compare examples, identify where component guidance is unclear and spot when authors are using a component for the wrong job. That's helpful, provided the results are reviewed by people who understand both the content and the product.\n\nA team could use generated test content to expose whether an alert still makes sense when its next step is long or unavailable. These are test cases, not evidence that the live component is accessible. The risk is filling components without the operational context. A generated alert may sound polished while missing the actual consequence. A generated card may obey the character count while saying nothing useful. Defining the [content standards](https://contentlexicon.com/content-standards) alongside the component’s fields is the safest way to ensure healthy components.\n\n## Questions people ask\n\n**Is a component the same as a template?**\n\nNo. A component is a reusable part within a system. A template is a larger arrangement of components and content for a particular purpose.\n\n**Is a content component the same as a content type?**\n\nNot quite. A [content type](https://contentlexicon.com/content-type) models a whole kind of item, such as an event, a product or a case study. A content component is a reusable part that may appear within many types – the same [accessibility](https://contentlexicon.com/accessibility) statement might sit in every event and every venue page.\n\n**How does a component relate to a content pattern?**\n\nThey work at different levels. A [content pattern](https://contentlexicon.com/content-pattern) describes how to handle a recurring task, such as turning down an application, and often brings several components together to do it. The component is reused as it stands, while the pattern is reused as a way of thinking, so the result looks different each time.\n\n**Who owns component content?**\n\nUsually several people. Designers, [content designers](https://contentlexicon.com/content-designer), developers and product owners all affect how a component works. Someone still needs to own the words, rules and examples, so decisions don't disappear between disciplines.\n\n**What makes a good component?**\n\nA clear purpose, defined fields, useful guidance, accessible behaviour, good examples and limits that match real content. If every use needs an exception, the component is either too narrow or pretending to be universal.\n\n**Can components make content too rigid?**\n\nThey can if badly designed. Good components create useful consistency while leaving room for the variation the content needs. With content components in particular, the risk is chopping material so finely that it reads like a set of parts rather than something written for a person.\n\n## Learn more\n\n- [Adapting ourselves to adaptive content](https://karenmcgrane.com/talks/adapting-ourselves-to-adaptive-content/) – Karen McGrane's influential talk makes the case for content as well-structured, reusable chunks rather than page-shaped blobs, ready for channels nobody has invented yet.\n- [Components](https://designsystem.digital.gov/components/overview/) – the US Web Design System’s catalogue shows how reusable interface components are defined and documented.\n- [Guidelines for creating web platform compatible components](https://www.w3.org/2001/tag/doc/webcomponents-design-guidelines) – A 2018 W3C TAG finding about browser custom elements, including native behaviour and keyboard interaction; technical Web Components guidance, not a standard for every content module." }, { "term": "Content", "alsoCalled": [], "slug": "content", "definition": "The words, images, numbers, sounds and interactions an organisation uses to communicate, in every format and on every channel, internal and external.", "category": "Foundations", "related": [ "content-audit", "content-design", "content-ecosystem-map", "content-governance", "content-inventory", "content-marketing", "content-strategy", "knowledge-base", "microcopy", "social-media", "structured-content", "website" ], "updated": "2026-10-11T06:23:00.000Z", "url": "https://contentlexicon.com/content", "markdown": "https://contentlexicon.com/content.md", "credit": "Content, from the Content Lexicon by Contentious (https://contentlexicon.com/content), CC BY 4.0", "body": "Content covers far more than most people picture when they hear the word. It includes articles and web pages, but also the label on a button, the question on a form, the error message, the confirmation [email](https://contentlexicon.com/email) and the letter that follows it. It includes photographs, illustrations, diagrams, charts, maps, tables of figures, video, [audio](https://contentlexicon.com/audio) and the transcript beside them. It includes the help centre, the annual report, the product description, the terms and conditions and the answer a chatbot gives. And it includes what an organisation says to its own people: the intranet, the policies, the training and the guidance a new starter needs on their first day.\n\nIf someone reads, watches, hears or uses it to understand something or get something done, it's content. The [website](https://contentlexicon.com/website), app, [CMS](https://contentlexicon.com/cms) or printed leaflet is the content container. Format and channel describe what shape content takes and where it goes.\n\nIn everyday use the word has narrowed. Online, content often means posts and videos made for social platforms, the output of \"content creators\" or the articles and assets a marketing team publishes. All of that is content, and it's the most visible (and often the loudest) part. It's also a small part. Most of what people rely on from an organisation is neither social nor marketing.\n\n## Why the definition matters\n\nHow an organisation defines content decides what it looks after. If content means campaigns and social posts, that's where the budget, the skills and the editorial care go. The forms, letters, help pages, policies, product information and staff guidance are left to whoever happens to write them, with no owner, standards or [review date](https://contentlexicon.com/review-date). Yet that's often where people succeed or fail: a confusing letter becomes a phone call, an unclear form question becomes a rejected application and an out-of-date policy on the intranet becomes a mistake someone makes in good faith.\n\nThe full definition makes the work visible. A [content inventory](https://contentlexicon.com/content-inventory), audit or strategy that counts only the marketing site misses most of the estate, along with the cost of maintaining it. Seeing everything shows how the pieces depend on each other: the product description draws on the specification, the help article repeats the policy and the chart in the annual report comes from the same figures as the press release. When one changes, the others need to follow. It also shows what content is for. Content can inform, educate, persuade, entertain or support, and an organisation that thinks of it as promotion tends to stop at persuasion.\n\nInternal content deserves the same care. Staff are an audience with needs, deadlines and limited patience, and the expertise held in internal documents is often the raw material for everything published outside.\n\n## How AI changes the picture\n\nAI turns the broad definition into a practical one. Assistants, search tools and agents built on an organisation's material read whatever they're given: the published pages, but also the [PDFs](https://contentlexicon.com/pdf) nobody has opened in years, the old policy still on the intranet, the support transcripts and the spreadsheet of product data. A retrieval system can't tell material someone thought of as content from material that was only ever lying around. Whatever it draws on shapes the answers people receive, so anything it can read needs an owner, a date and a reason to exist. A narrow definition leaves that material unmanaged.\n\nContent is also increasingly assembled at the moment someone asks. A chatbot's answer, a generated summary or a personalised email may never have existed as a page anyone wrote or approved in that form. It's still content, and the organisation is still responsible for it. Some of the work moves from writing each piece towards looking after the sources, structure and rules that answers are made from.\n\nThe line between content and data blurs as well. A model can turn a table into a paragraph, a paragraph into structured fields or a set of figures into a chart, so the numbers in a [database](https://contentlexicon.com/database) and the prose that explains them become two forms of the same material. Organisations that keep them in separate systems, owned by separate teams, may find that separation harder to defend.\n\n## Questions people ask\n\n**Isn't content mostly social media and video?**\n\nThose are channels and formats, and for some organisations they're important ones. A charity's content also includes its fundraising pages, the guidance its helpline staff use and the impact report its trustees [sign off](https://contentlexicon.com/sign-off). A bank's includes the app, the statements and the letter explaining a change of terms. Social posts are often the most visible content an organisation has, and rarely the largest share of it.\n\n**Does internal content count?**\n\nYes. The intranet, policies, procedures, training materials and [knowledge base](https://contentlexicon.com/knowledge-base) serve an audience of staff, volunteers or members, and they can be clear or confusing, current or out of date like anything else. They also feed external content: the answer a customer gets is often only as good as the guidance given to the person or system answering.\n\n**What's the difference between content and data?**\n\nData is recorded so it can be stored, processed and counted; content is shaped so people can understand and use it. They overlap constantly. A price is data, and the sentence explaining what it includes is content. A chart turns data into content through its choice of figures, labels and caption. [Structured content](https://contentlexicon.com/structured-content) sits deliberately between the two.\n\n**Why do so many people dislike the word?**\n\nIt can sound like packing material: something to fill a space, measured by volume rather than value. Critics have long objected that it flattens writing, journalism and art into interchangeable stuff. The defence is practical. Digital channels carry every form of communication, so they need one word that covers all of it. Used with care, it names the whole of what an organisation communicates, which is what a word this broad is for.\n\n## Learn more\n\n- [Content](https://mitpress.mit.edu/9780262543286/content/) – Kate Eichhorn's MIT Press introduction to the narrower sense of the word, traced from the early internet to social media, with the Instagram egg, a photo that spread for the sake of spreading, as its emblem.\n- [Thingy words](https://www.kmworld.com/Articles/Columns/Perspective-on-Knowledge/Thingy-words-97014.aspx) – David Weinberger's column on why the internet needs words as abstract as \"content\", and what they cost: he agrees with Doc Searls that it makes the web sound one-way, and would rather say \"stuff\".\n- [The content manifesto](https://contentious.ltd/the-content-manifesto) – Contentious's principles, including that content is everywhere, from button text and videos to privacy statements." }, { "term": "Content audit", "alsoCalled": [], "slug": "content-audit", "definition": "A structured review of what content exists, how good it is and what should happen to it next.", "category": "Operations", "related": [ "competitor-analysis", "content", "content-debt", "content-decay", "content-health-check", "content-inventory", "content-plan", "content-strategy", "decentralised-governance", "gap-analysis", "retention-and-archiving", "rot-content", "technical-debt" ], "updated": "2026-10-07T13:53:00.000Z", "url": "https://contentlexicon.com/content-audit", "markdown": "https://contentlexicon.com/content-audit.md", "credit": "Content audit, from the Content Lexicon by Contentious (https://contentlexicon.com/content-audit), CC BY 4.0", "body": "A content audit takes the pages you have and asks of each one whether the facts are still right, whether anyone reads it, who owns it and whether it does what people came for.\n\nIt's more than a spreadsheet of URLs, though spreadsheets often feature. The point is not counting pages for the joy of counting pages. It's about building enough evidence to decide, page by page and section by section, what to keep, fix, merge or retire.\n\n## Why it matters for content\n\nA content audit gives teams a clearer picture of their content estate than memory or anecdote ever can. It turns up the three pages explaining the same fee in three different ways, the guidance nobody has touched since 2019 and the section whose owner left two years ago. That makes it useful both for immediate clean-up work and for larger decisions about strategy, architecture and operations.\n\nAudits earn their keep when an organisation has inherited a large or messy body of content, when a redesign or migration is planned, when AI or search is exposing content weaknesses, or when nobody can confidently answer basic questions such as what exists, what matters and what should go.\n\n## How AI changes the picture\n\nAI expands what counts as the content estate. It's no longer only pages on a [website](https://contentlexicon.com/website). Teams may also need to audit [prompts](https://contentlexicon.com/prompt), [knowledge bases](https://contentlexicon.com/knowledge-base), retrieved source sets, generated outputs and the documents that assistants rely on. Weak source material now has more ways to cause trouble, because it can feed summaries, chatbots and automated drafting as well as human readers.\n\nAI can also help with parts of the audit, such as classification or first-pass summarising, but it doesn't remove the need for editorial judgement. If anything, it increases the value of knowing which content is accurate, current and trustworthy enough to become machine input.\n\n## Questions people ask\n\n**Is a content audit the same as a content inventory?**\n\nNo. An inventory is mainly about what exists. An audit goes further and evaluates quality, usefulness and what should happen next.\n\n**Do we have to audit every page?**\n\nNot always. Sometimes a representative sample or a high-priority subset is enough, depending on the decision you're trying to make.\n\n**Is a content audit just about performance data?**\n\nPerformance can be one input, but a page can perform well and still be inaccurate, inaccessible or structurally weak.\n\n**When should we do a content audit?**\n\nOften before a redesign, migration, governance reset, AI rollout or major [content strategy](https://contentlexicon.com/content-strategy) decision. And whenever the state of the content is more assumed than known.\n\n## Learn more\n\n- [Content Inventory and Auditing 101](https://www.nngroup.com/articles/content-audits/) – scope, criteria and prioritisation without pretending everything must be audited at once.\n- [What is a content audit and how to do it](https://wgcontent.com/blog/how-to-conduct-a-successful-content-audit/) – a practical process joining crawl data, analytics and editorial judgement.\n- [Content audit vs content maturity assessment: what's the difference?](https://contentmaturity.com/learn/content-audit-vs-content-maturity) – a crisp distinction between examining the content itself and examining the organisation that produces it." }, { "term": "Content brief", "alsoCalled": [], "slug": "content-brief", "definition": "The document that tells a writer – human or machine – who a piece is for, what it's meant to do, what to draw on and how it should sound, before they start.", "category": "Operations", "related": [ "content-plan", "context-engineering", "editing", "persona", "skill", "user-needs" ], "updated": "2026-10-07T13:53:00.000Z", "url": "https://contentlexicon.com/content-brief", "markdown": "https://contentlexicon.com/content-brief.md", "credit": "Content brief, from the Content Lexicon by Contentious (https://contentlexicon.com/content-brief), CC BY 4.0", "body": "Beyond audience, purpose, sources and sound, a good content brief says what message and evidence the piece must carry, what form it should take, what constraints apply, who approves it and how anyone will know whether it worked.\n\nA brief isn't the same as a request. A request says \"we need a page about the new service\". A brief turns that into decisions: which readers, arriving with which questions, needing which facts, in which format, by which route to publication. Its real value is that it makes hidden assumptions visible before they become expensive, because fluent content aimed at the wrong reader is still the wrong content.\n\n## Why it matters for content\n\nWithout a brief, the [writer](https://contentlexicon.com/writer) has to guess. They guess how much the reader already knows, which claims the organisation is willing to stand behind, which source material is authoritative and what \"good\" looks like. Late drafts, repeated review rounds and unanswered reader questions can all begin with an inadequate brief. A brief front-loads the thinking so the [editing](https://contentlexicon.com/editing) conversation is about craft rather than intent.\n\nBriefs pull the most weight where several people touch the same piece, where [subject-matter experts](https://contentlexicon.com/subject-matter-expert) supply the substance but not the shape, where freelancers or agencies are involved, or where content has to meet [accessibility](https://contentlexicon.com/accessibility), legal or [plain-language](https://contentlexicon.com/plain-language) standards. They also matter when the stakes are high and the audience is unfamiliar: the further the writer sits from the reader, the more the brief has to carry. A good brief is short and specific. Some of the worst briefs are enormous, because containing every possible thought is an elegant way of avoiding the hard decisions.\n\n## How AI changes the picture\n\nAI doesn't change what a brief is. It changes how much rests on it. A generative tool will produce plausible words from almost nothing. Given a vague brief, a model will default to the most average version of the topic, sound confident about it and obediently follow whichever pattern it has seen most often. Judgement doesn't enter only at the brief, either: it continues through research, writing and review. A thin brief now comes back as a finished-looking draft within seconds, and the decisions nobody made are much harder to spot inside polished prose than in a half-written page.\n\nAn assistant can also work in the other direction. It could identify an unexplained claim or a missing reader question, ask for clarification and compare possible treatments against the agreed purpose. People still settle the claim and choose the approach.\n\nBriefs are also becoming [machine-readable](https://contentlexicon.com/machine-readable) inputs rather than internal paperwork. Audience, purpose, sources, [tone](https://contentlexicon.com/tone) rules, terminology and things to avoid are the context an AI tool needs, which means a brief written properly can be reused as [prompt](https://contentlexicon.com/prompt) material, retrieval instructions or a checking standard. Briefing is shifting from a nice habit to part of the content infrastructure. What you'd once have explained verbally to a colleague now needs to exist in writing, with the authoritative sources named rather than assumed.\n\n## Questions people ask\n\n**Is a brief the same as a content plan?**\n\nNo. A plan decides what content is needed and when. A brief covers one piece and defines what it has to do.\n\n**Do small teams need briefs?**\n\nYes, though lighter ones. A one-page brief with audience, purpose, key message, sources and approver is enough, and it saves more time than it takes.\n\n**Can AI write the brief?**\n\nIt can draft a structure and ask sensible questions, but the decisions in a brief – who matters, what we're claiming, which source is authoritative – are editorial and organisational judgements. Handing those to a tool defeats the purpose.\n\n**What if the brief turns out to be wrong?**\n\nIn some ways it has done its job. A brief you can disagree with is far more useful than an unspoken assumption nobody can see until publication.\n\n## Learn more\n\n- [Best practices for creating a strategic content brief](https://www.siteimprove.com/blog/content-briefs-best-practices/) – useful prompts for objectives, audience, narrative and process.\n- [The Discipline of Content Strategy](https://alistapart.com/article/thedisciplineofcontentstrategy/) – Kristina Halvorson's 2008 essay on why content needs planning and governance, which places an individual commission in its wider system." }, { "term": "Content calendar", "alsoCalled": [ "Editorial calendar" ], "slug": "content-calendar", "definition": "A schedule showing what content is planned, when it will be created or published and who is responsible.", "category": "Operations", "related": [ "content-lifecycle", "content-owner", "content-plan", "email", "workflow" ], "updated": "2026-10-11T06:27:00.000Z", "url": "https://contentlexicon.com/content-calendar", "markdown": "https://contentlexicon.com/content-calendar.md", "credit": "Content calendar, from the Content Lexicon by Contentious (https://contentlexicon.com/content-calendar), CC BY 4.0", "body": "A content calendar might include articles, [emails](https://contentlexicon.com/email), social posts, reports, campaigns, product updates, events, seasonal moments and [review dates](https://contentlexicon.com/review-date).\n\nAt its simplest, a content calendar helps a team see what's coming. That visibility matters: it prevents everyone discovering on Tuesday that three newsletters, a campaign launch and a chief executive blog are all apparently happening on Thursday. A useful calendar makes the work visible early enough for people to make sensible decisions about timing, effort and ownership.\n\n## Why it matters for content\n\nA content calendar gives content work a shared rhythm. It connects each piece to a purpose, an audience and the people doing the work. That makes it easier to spot gaps, manage competing demands and give important work enough time for research, writing, design, review, [accessibility](https://contentlexicon.com/accessibility) checks and publication. Without that visibility, [content teams](https://contentlexicon.com/content-team) can spend their time responding to whichever request arrived most recently, while strategic work slides off the table.\n\nThe calendar also supports better judgement about quantity and timing. It can show whether several pieces are aimed at the same audience, whether a campaign is asking too much of the same people or whether a supposedly regular series has become a ritual nobody has questioned in years. It's particularly useful when content depends on other work or external events – a product release, a policy decision, a research finding, a client approval or a legal review. Recording those relationships turns a list of dates into a planning tool.\n\nThe danger is that the calendar becomes a machine for producing content because a slot exists. Regular publishing helps only when each piece has a job. A weekly post that nobody needs isn't strategy; it's a small recurring tax on everyone’s will to live. A good calendar gives enough structure for consistency while leaving room to cancel, combine, delay or replace content when circumstances change. It doesn't pretend the future has signed a contract.\n\n## How AI changes the picture\n\nGenerative tools can turn out drafts, variations and ideas fast enough to fill every empty slot in a schedule by lunchtime. If the calendar only records publishing slots, it may encourage a team to produce more content without making any of it more useful. So it needs to record why each item deserves to exist, for whom and why now, as well as when it's due.\n\nAI can help with parts of the planning process. It might group planned items by audience or theme, identify repeated topics, suggest missing information, summarise dependencies or turn a well-maintained calendar into a draft briefing view. It can also help compare planned content with source material and flag where an idea appears to have no clear evidence behind it. Those uses help because they support human decisions rather than pretending to make them.\n\nThe quality of the output will depend on the quality of the calendar. Clear fields for purpose, audience, channel, owner, status, source, review date and dependencies give both people and machines something meaningful to work with. An AI tool can't do much with entries such as “social content” or “thought leadership piece”, because there's nothing in them to work from. Ownership and approval points still matter too, particularly when AI is involved in drafting or publishing content at scale.\n\n## Questions people ask\n\n**How is a content calendar different from a content strategy?**\n\nA strategy explains what you're trying to achieve, who you're helping and what choices you'll make. A calendar shows how planned work is being organised over time. The calendar should express the strategy, not replace it.\n\n**Does every piece of content need a fixed publication date?**\n\nNot always. Some work needs a firm date, such as an event announcement or product update. Other work may be better managed through a priority, target window or review point. False precision makes planning look tidy while hiding uncertainty.\n\n**How much detail should a calendar contain?**\n\nEnough to support decisions and handovers, but not so much that maintaining it becomes a second content job. The fields listed above, kept current, are usually more useful than a small novel about every draft.\n\n**Can AI manage a content calendar for us?**\n\nAI can help organise, classify and question the information in a calendar. It shouldn't be left to decide what deserves attention, what is accurate or what should be published without appropriate human judgement and review.\n\n## Learn more\n\n- [7 Steps to a More Strategic Editorial Calendar](https://contentmarketinginstitute.com/content-optimization/7-steps-to-a-more-strategic-editorial-calendar) – capacity, content mix and audience value before coloured boxes.\n- [Hearst starts with 'why' as journalists asked to produce fewer articles](https://pressgazette.co.uk/publishers/hearst-starts-with-why-as-journalists-asked-to-produce-fewer-articles/) – Press Gazette reports how Hearst UK requires a stated purpose before any article is pitched, and found traffic rising as volume fell. A documented example of the entry's central argument, from July 2026.\n- [Your Editorial Calendar Is Not Your Content Marketing Strategy](https://contentmarketinginstitute.com/content-distribution-promotion/your-editorial-calendar-is-not-your-content-marketing-strategy) – Michele Linn’s argument that a calendar is tactical execution and needs a strategy behind it: who you’re helping, how you help differently and how you’ll know it’s working." }, { "term": "Content costs", "alsoCalled": [], "slug": "content-costs", "definition": "The full cost of creating, reviewing, publishing, maintaining and retiring content over time.", "category": "Operations", "related": [ "content-debt", "content-maintenance", "technical-debt" ], "updated": "2026-10-08T16:39:00.000Z", "url": "https://contentlexicon.com/content-costs", "markdown": "https://contentlexicon.com/content-costs.md", "credit": "Content costs, from the Content Lexicon by Contentious (https://contentlexicon.com/content-costs), CC BY 4.0", "body": "The visible fee for producing the first version is often only a modest part of the total.\n\nCosts include money, time and attention. A short page may involve several specialists and approval rounds. Once published, it occupies a content system, appears in search and creates an expectation that the organisation will keep it accurate. There's no such thing as a free page, though plenty have escaped without an owner.\n\n## Why it matters for content\n\nUnderstanding cost helps organisations make better decisions about what deserves to exist. If teams count only production, publishing more appears cheap while maintenance accumulates invisibly. The result is often duplicated, outdated or low-value content that makes useful material harder to find.\n\nDifferent content carries different continuing costs. A campaign page may have a short life and clear retirement date. Regulatory guidance may need frequent expert review. Translated content multiplies update work across languages. [Structured content](https://contentlexicon.com/structured-content) may cost more to model initially but reduce duplication later.\n\nCost isn't an argument for producing as little as possible. It's an argument for investing deliberately. Useful content can reduce support demand, prevent errors, improve services and build trust. The question is whether its value justifies the resources needed to keep it healthy.\n\n## How AI changes the picture\n\nAI lowers the apparent cost of drafting and adaptation. This can make valuable work possible, but it can also encourage organisations to create far more content than they can govern. Cheap production moves the constraint elsewhere: source quality, review, risk, maintenance and the reader’s finite attention.\n\nAI systems also create new content costs. [Prompts](https://contentlexicon.com/prompt), instructions, knowledge sources, evaluation cases and generated responses need design and upkeep. Models and tools may change, requiring retesting. A [workflow](https://contentlexicon.com/workflow) that saves writing time but demands extensive correction or creates harmful errors may be costly in less obvious ways.\n\nUsed carefully, AI can reduce costs by identifying duplication, suggesting updates, supporting translation and checking routine quality criteria. Savings should be assessed across the lifecycle rather than announced after the first fast draft.\n\n## Questions people ask\n\n**How do we calculate the cost of one piece of content?**\n\nChoose a period, such as the next year, and estimate creation, approval and expected updates within it. Add the relevant share of technology, translation and specialist review costs. A rough range is a useful starting point; avoid charging a whole shared platform to one page or counting the same staff time twice.\n\n**Is more content always more expensive?**\n\nAdditional independently maintained items usually add work. Reusing one approved record in several views can be cheaper than keeping several copies, although building and maintaining that reuse has a cost too.\n\n**Does deleting content save money?**\n\nIt can, if the content has little value and can be removed safely. Retirement itself needs care: redirects, records, dependencies and [user needs](https://contentlexicon.com/user-needs) may remain.\n\n**Will AI make content much cheaper?**\n\nAI can reduce production costs and make research, quality control and maintenance faster and more efficient. Human judgement and accountability still matter, and savings can be diluted if teams use the extra capacity simply to produce more.\n\n## Learn more\n\n- [Establishing true cost of ownership of your content operations](https://www.arcastgroup.com/insights/establishing-true-cost-of-ownership-of-your-content-operations) – Arcast looks beyond software fees to the costs of training, support, maintenance and eventual retirement, helping you see what a content operation needs over its whole life.\n- [Content operations: your cost-saving and scaling opportunity](https://www.thinklikeapublisher.com/content-operations-marketings-most-overlooked-cost-saving-and-scaling-opportunity/) – A marketing practitioner explores how clearer ownership and better workflows can reduce duplicated work and slow approvals, making room for growth without simply adding more production." }, { "term": "Content debt", "alsoCalled": [], "slug": "content-debt", "definition": "The future work created by publishing content without planning to maintain, reuse or retire it.", "category": "Operations", "related": [ "content-audit", "content-costs", "content-decay", "content-maintenance", "content-owner", "rot-content", "single-source-of-truth", "technical-debt" ], "updated": "2026-10-15T15:42:00.000Z", "url": "https://contentlexicon.com/content-debt", "markdown": "https://contentlexicon.com/content-debt.md", "credit": "Content debt, from the Content Lexicon by Contentious (https://contentlexicon.com/content-debt), CC BY 4.0", "body": "Content debt is the future work an organisation creates when it publishes without planning to maintain, reuse or retire what it publishes. Outdated pages, duplicated facts, inaccessible [PDFs](https://contentlexicon.com/pdf), unclear ownership and pages written for one channel that now have to serve several all count. None of it stops a [website](https://contentlexicon.com/website) working today, but each makes the next change slower, riskier and more expensive.\n\nThe term borrows from [technical debt](https://contentlexicon.com/technical-debt), Ward Cunningham's 1992 metaphor for the cost of shipping software quickly and not going back to rework it. The comparison holds up well. Some debt is a sensible choice: launching a service page before the [content model](https://contentlexicon.com/content-model) is perfect may be the right call. Like a loan, though, it charges interest, paid in repeated work, confused readers and lost trust. Trouble begins when nobody records the shortcut, understands the interest or decides whether to pay it back.\n\nContent debt is a broader idea than [content decay](https://contentlexicon.com/content-decay), the gradual loss of accuracy as the world moves on. Decay is one source of debt. Other debt is created on the day of publication: a second page that duplicates an approved source, a PDF nobody can read with a screen reader, a campaign [microsite](https://contentlexicon.com/microsite) with no owner and no end date.\n\n## Why it matters for content\n\nContent debt usually shows up as repeated work, often in other parts of the organisation. Readers find the wrong version, support teams explain confusing guidance and editors make the same change in twenty places. Each problem looks small, while the combined effect damages trust and makes change expensive. Robert Rose puts it well: every published asset carries a maintenance obligation, a small recurring promise that it will stay true, current and findable.\n\nA debt register turns that invisible irritation into manageable work. Each entry records the item, its consequence, an owner, the risk and the likely effort. One might read: \"Opening hours are typed into 40 [branch](https://contentlexicon.com/branch) pages instead of pulled from the branch [database](https://contentlexicon.com/database). Consequence: every change takes two days and pages disagree. Owner: digital team. Fix: move the field into the content model.\"\n\nSome debt should be repaid quickly, some accepted for now and some cleared altogether by deleting content nobody needs. When [GOV.UK](http://GOV.UK) [content designers](https://contentlexicon.com/content-designer) audited 7,396 transport pages with the Department for Transport, they concluded that \"do we need to remove this page?\" should be the first question, because a yes makes the others unnecessary.\n\nThe opening hours example also shows where content debt meets technical debt. The problem sits in the pages, but the fix sits in the content model. A register that covers both saves an argument about whose problem it is before anyone fixes it.\n\nPrevention matters too. Clear ownership, structured reuse, [review dates](https://contentlexicon.com/review-date) and retirement plans reduce avoidable debt, and they work best as part of everyday publishing rather than an annual clean-up.\n\n## How AI changes the picture\n\nAI can find content debt at a scale no team could manage by hand. A well-instructed model can read an entire estate in hours, picking out duplicates, outdated references, contradictory claims and inconsistent terminology, and it can show which problems turn up most often in user questions or generated answers. It can draft the routine fixes too. People still decide what's true now, which version is authoritative and what can go.\n\nAI can also increase debt by making production cheap. More generated pages, summaries and variants mean more material to review, each with its own maintenance obligation. Weak source content becomes especially costly when assistants reuse it, because one outdated fact can surface in thousands of answers to people who never see the page it came from. [Prompts](https://contentlexicon.com/prompt), system instructions and retrieval sources are content too, and they belong in the same register as pages and documents.\n\n## Questions people ask\n\n**Is all old content debt?**\n\nNo. Age alone doesn't make content debt. A stable reference can stay useful for years, while a newly published page creates debt immediately if it duplicates an approved source that will now need two updates.\n\n**Should we fix the oldest items first?**\n\nPrioritise by consequence, frequency and future work. A small error in critical guidance may matter more than a large archive nobody uses.\n\n**Is content debt the same as content decay?**\n\nNo. Decay is the loss of usefulness over time. Debt also includes deliberate shortcuts, structural problems and neglected maintenance that create future cost. Decay that nobody deals with becomes debt.\n\n**Can AI clear the backlog?**\n\nIt can find most of it and handle a good share of the routine fixes. Someone still needs to be accountable for what's authoritative, what's safe to change and what's no longer needed.\n\n## Learn more\n\n- [How To Face Your Content Debt and Get Buy-In for Fixing It](https://contentmarketinginstitute.com/content-operations/content-debt-fix) – Robert Rose's 2026 column on the maintenance obligation every asset carries, with a blunt \"default-dead\" rule for teams whose debt is too large to audit: treat anything older than two years as wrong until a named owner reverifies it.\n- [What we learned from a large scale content audit](https://insidegovuk.blog.gov.uk/2018/02/23/what-we-learned-from-a-large-scale-content-audit/) – Mark Green and Steph Wilson's first-hand account of auditing 7,396 transport pages on [GOV.UK](http://GOV.UK), including what they'd do differently next time." }, { "term": "Content decay", "alsoCalled": [], "slug": "content-decay", "definition": "The gradual loss of accuracy, relevance and usefulness as content ages.", "category": "Operations", "related": [ "content-audit", "content-debt", "content-lifecycle", "content-maintenance", "evergreen-content", "review-date", "rot-content", "technical-debt" ], "updated": "2026-10-11T06:24:00.000Z", "url": "https://contentlexicon.com/content-decay", "markdown": "https://contentlexicon.com/content-decay.md", "credit": "Content decay, from the Content Lexicon by Contentious (https://contentlexicon.com/content-decay), CC BY 4.0", "body": "Content decay happens when published content stays the same while the facts, services and language it describes move on. A page that was right when it went live becomes less right over time: a fee goes up, a form is replaced, a team is renamed, a link stops working or the words people use for a topic change.\n\nThat sets decay apart from content that was wrong from the start, which is a quality problem at publication rather than a maintenance one. It also explains why decay is easy to miss. A decayed page can still attract traffic, look polished and pass automated checks while directing people towards an expired process or a retired service. Link rot, where links break as pages move or disappear, is one of the most measurable forms. Pew Research Center found that 38% of web pages that existed in 2013 were no longer accessible a decade later.\n\n## Why it matters for content\n\nDecayed content damages trust and creates work elsewhere. People contact support, make poor decisions or repeat outdated claims. Search engines and internal systems may keep presenting the page because nobody has marked it as unreliable.\n\nSpotting decay means looking at accuracy as well as traffic. Falling visits can point to a page that no longer matches what people search for, but some of the most harmful decay sits on popular pages. Better clues include support contacts that mention the page, internal searches that land on it and then search again, feedback saying the instructions didn't work, and changes to the policies, products or prices it depends on. Recording those dependencies lets a team check the right pages when something changes.\n\nDifferent content decays at different rates. Prices, people and regulations are volatile; an explanation of a method may stay useful for years. A [review date](https://contentlexicon.com/review-date) set according to that risk catches decay before readers do, and a [ROT content](https://contentlexicon.com/rot-content) audit (redundant, outdated, trivial) finds what slipped through. Some content has finished its useful life, and retiring it is as much a fix as updating it.\n\n## How AI changes the picture\n\nComparing a page with an updated source can help a team find changed requirements or conflicting claims. Broken-link monitoring needs ordinary automation rather than AI; judging whether a changed sentence affects a service needs more context. Flags should point to the evidence so someone can check them.\n\nAI also carries decay further. An assistant may reuse stale guidance without sending readers to the original page, so the error reaches people who never see the page or its date. Review triggers need to cover the content supplied to assistants as well as the pages people visit. Dates and canonical markers can help a system choose between versions when it uses them, but they don't establish that the claim is still true.\n\n## Questions people ask\n\n**Is content decay the same as falling search traffic?**\n\nNo. Falling traffic can have many causes, including changes to search engines that have nothing to do with the page. A popular page can still contain an expired requirement, so check accuracy and usefulness separately from visits.\n\n**Does adding a new date make content fresh?**\n\nA review date should mean someone checked the content against the relevant evidence. Update it if something needs changing, and record a completed review even if the wording is still right. Changing a publication date without doing either is misleading.\n\n**Which content should we check first?**\n\nHigh-use, high-risk and fast-changing material, plus content that feeds services or AI systems. A page that tells people how to apply, what they'll pay or what they're entitled to usually deserves attention before a background explainer.\n\n**Can evergreen content decay?**\n\nYes, though more slowly. Even durable ideas need occasional checks for accuracy, examples, links and the terms people now use.\n\n## Learn more\n\n- [Web content has a maintenance problem](https://gerrymcgovern.com/web-content-has-a-maintenance-problem/) – Gerry McGovern’s polemic about the labour of editing, moderation and deletion, challenging claims that cheap generation means cheap content.\n- [When Online Content Disappears](https://www.pewresearch.org/data-labs/2024/05/17/when-online-content-disappears) – Pew Research Center's 2024 study of link rot, finding that 38% of web pages from 2013 had gone a decade later and that around one in five government pages contained at least one broken link.\n- [How to update old content](https://moz.com/blog/how-to-update-content) – Ann Smarty's search-focused guide to diagnosing and refreshing older articles; practical source and link checks sit alongside marketing tactics." }, { "term": "Content design", "alsoCalled": [], "slug": "content-design", "definition": "Designing content around user needs, tasks and evidence.", "category": "Editorial craft", "related": [ "agile", "content", "content-designer", "content-pattern", "conversational-design", "frontend", "microcopy", "paddle-swim-dive", "pair-writing", "user-needs", "ux", "ux-writer", "wireframe" ], "updated": "2026-10-08T15:13:00.000Z", "url": "https://contentlexicon.com/content-design", "markdown": "https://contentlexicon.com/content-design.md", "credit": "Content design, from the Content Lexicon by Contentious (https://contentlexicon.com/content-design), CC BY 4.0", "body": "Content design starts with evidence – user research, search data, support enquiries, analytics – and uses that evidence to decide what content should exist, in what form, in what order and in what words.\n\nIt's easy to mistake content design for writing with a smarter job title. The distinction matters: a [content designer](https://contentlexicon.com/content-designer) might conclude that the best answer to a problem is a table, a single sentence, a redesigned form or no content at all. Writing is one of the tools available, not the purpose of the work. Content designers usually lead it, and in product teams [UX writers](https://contentlexicon.com/ux-writer) often share it, with researchers, service designers and subject experts closely involved.\n\n## Why it matters for content\n\nBad content isn't necessarily badly written. It can be content that answers a question nobody asked, in a place nobody looks, in a shape nobody can use. Content design attacks that problem at the root by insisting that the need comes before the draft. It tends to produce less content, better placed and easier to maintain, which is usually cheaper as well as kinder.\n\nIt does its most valuable work in transactional and service contexts, where someone is trying to apply for something, understand an entitlement, make a decision or complete a task.\n\n## How AI changes the picture\n\nAI is very good at producing plausible content quickly, which makes the design question more important. If nobody has established what a page is for and who needs it, AI will happily generate three versions of the wrong thing, all fluent and none necessary.\n\nAI helps with the middle of the process. It can cluster support enquiries, summarise research notes, suggest alternative structures, draft variants for testing and check [readability](https://contentlexicon.com/readability) at scale. What it can't do is decide which [user need](https://contentlexicon.com/user-needs) deserves to exist as content, or take responsibility for the consequences when the answer is wrong. Content design is mostly judgement, and judgement is the part that stays human.\n\nThere's also a newer design consideration: content designed for tasks now needs to survive being summarised by machines. Clear structure, [plain language](https://contentlexicon.com/plain-language), accurate headings and self-contained answers all help an AI system represent the content faithfully rather than inventively.\n\n## Questions people ask\n\n**Is content design just a fashionable name for copywriting?**\n\nThe two overlap, and content designers borrow plenty from the copywriter's craft. The difference is where the work starts: content design first decides what should exist at all, in what format and sequence, and only then writes it.\n\n**Do we need content design if we already have UX designers?**\n\nUsually yes, and ideally the two work together. Interfaces are mostly language, and language left to the end of a design process tends to arrive as an apology.\n\n**Does content design only apply to government and service sites?**\n\nIn the UK it was popularised by government digital teams, but the underlying discipline applies anywhere people come to content with a purpose – charities, universities, membership bodies, museums, commercial sites.\n\n**Where does the evidence come from if we have no research budget?**\n\nSearch queries, site search, support tickets, sales questions and the things staff explain repeatedly on the phone. Most organisations are sitting on more evidence than they think.\n\n## Learn more\n\n- [Content Design, second edition](https://contentdesign.london/shop/content-design-by-sarah-winters-and-rachel-edwards) – the practical book from Sarah Winters and Rachel Edwards on research, planning and delivery.\n- [Why Content Design Matters: An Interview with Sarah Richards](https://medium.com/ux-and-front-end-interviews/why-content-design-matters-an-interview-with-sarah-richards-a9ad0de3c37a) – an interview with Sarah Richards (now Sarah Winters), who set up content design at GDS, about designing around needs rather than polishing requested pages.\n- [The depleted bookshop](https://contentious.ltd/thoughts/the-depleted-bookshop) – a plea for digital content that can move, delight and surprise as well as inform." }, { "term": "Content designer", "alsoCalled": [], "slug": "content-designer", "definition": "A role that uses evidence about people's needs to decide what content should exist and to design it so they can understand and act on it.", "category": "Roles", "related": [ "agile", "content-design", "content-editor", "content-strategist", "content-team", "subject-matter-expert", "ux-writer", "writer" ], "updated": "2026-10-11T06:24:00.000Z", "url": "https://contentlexicon.com/content-designer", "markdown": "https://contentlexicon.com/content-designer.md", "credit": "Content designer, from the Content Lexicon by Contentious (https://contentlexicon.com/content-designer), CC BY 4.0", "body": "A content designer works out what people need to know or do, then designs the content that gets them there: deciding whether something should be a page, a table, a form or nothing at all, and writing, testing and refining it. Content designers usually work in multidisciplinary teams with user researchers, service and interaction designers, developers and [subject matter experts](https://contentlexicon.com/subject-matter-expert). They're accountable for whether the content works for the people using it, not only whether it reads well.\n\nThe title spread from UK government digital teams in the early 2010s and is now common in product companies, charities and public services in many countries. It overlaps with [UX writer](https://contentlexicon.com/ux-writer), and many organisations use the two interchangeably. Levels are often well defined: the UK government framework runs from associate content designer to [head of content](https://contentlexicon.com/head-of-content) design, with responsibility widening from individual pieces of content to whole services and teams.\n\n## Why an organisation needs a content designer\n\nOrganisations without content designers tend to publish what they're asked to publish. A department requests a page, someone writes it and the site gains one more thing to maintain, whether or not anyone needed it. A content designer changes the starting point. They ask what problem the content is meant to solve, look at the evidence and push back when the answer is a different format, a simpler process or no content at all. That challenge is the most valuable part of the job, and the part most easily lost when the role is treated as a writing service.\n\nThe need is clearest wherever people come to content to get something done: applying, paying, booking, checking whether they're eligible or recovering when something goes wrong. Support calls about things the [website](https://contentlexicon.com/website) should have explained are often the first sign. The role depends on joining work early, with access to user research and enough standing to influence the service as well as the words. Brought in at the end to tidy up the copy, even an excellent content designer becomes an expensive proofreader.\n\n## How AI changes the role\n\nContent designers have moved quickly into AI work, partly because so much of it is language. In product teams they write the instructions that shape how assistants respond, design what happens when an assistant can't help and test generated answers against what users need. In services and publishing they increasingly design for two kinds of reader: people, and the systems that retrieve, summarise and answer from their content. The UK government's 2026 update to the content designer skills added content management and architecture as the one skill unique to the role, framed broadly enough to cover reuse anywhere from paper letters to AI interfaces.\n\nThe risk for the role lies in a misunderstanding: the idea that anyone with an assistant can now do [content design](https://contentlexicon.com/content-design), or that one designer with AI can do the work of several. Andrew Tipp argues that this mistakes what the job is, because research, judgement and the relationships needed to change a service don't compress the way drafting does. A job description might now include designing for AI-generated or AI-retrieved content, evaluating responses against [user needs](https://contentlexicon.com/user-needs) and structuring content for reuse, alongside research, testing and [accessibility](https://contentlexicon.com/accessibility).\n\n## Questions people ask\n\n**What do the different levels of content designer do?**\n\nJunior content designers usually work on defined pieces within a team, with their work reviewed. Mid-level designers take responsibility for whole journeys or products and work directly with research and policy colleagues. Senior and lead designers set standards, review others' work and shape services across teams, and a head of content design builds the profession within an organisation. Boundaries vary, so describe the scope of decisions rather than relying on the title.\n\n**Where should content designers sit?**\n\nUsually in the teams building the products and services they work on, with a central lead or community that keeps standards consistent. Embedded designers know the service; a central lead stops each team inventing its own conventions.\n\n**How many content designers do we need?**\n\nIt depends on how many services or products are changing at once and how much content they involve. In an organisation with several product or service teams, embedding a content designer in each is a sensible starting point; a small charity might begin with one designer working across its most-used journeys. Watch for designers spread so thinly that they only ever review other people's copy.\n\n**What background makes a good content designer?**\n\nJournalism, [editing](https://contentlexicon.com/editing), [copywriting](https://contentlexicon.com/copywriting), policy and research roles all lead into it. What matters most is comfort with evidence: someone who will happily abandon a well-written draft because testing shows it isn't working.\n\n## Learn more\n\n- [Content designer](https://ddat-capability-framework.service.gov.uk/role/content-designer) – the UK government's description of six content designer levels, from associate to head of content design; one employer's framework, but a detailed starting point for job descriptions.\n- [Updating the skills content designers need in government](https://designnotes.blog.gov.uk/2026/06/09/updating-the-skills-content-designers-need-in-government/) – an account by content design leads from MOJ, DWP, GDS and HMRC of how the role's skills were rewritten, including the new skill in content management and architecture.\n- [The future of content design in an AI world](https://uxdesign.cc/the-future-of-content-design-in-an-ai-world-bb58ba26455c) – Andrew Tipp's essay arguing that the real threat to content designers is the belief that fewer of them are needed, or that anyone with an assistant can do the job." }, { "term": "Content ecosystem map", "alsoCalled": [], "slug": "content-ecosystem-map", "definition": "A picture of where content lives and how it travels between those places.", "category": "Research and evidence", "related": [ "content", "content-inventory", "content-strategy", "information-architecture", "silo", "systems-thinking" ], "updated": "2026-10-10T21:07:00.000Z", "url": "https://contentlexicon.com/content-ecosystem-map", "markdown": "https://contentlexicon.com/content-ecosystem-map.md", "credit": "Content ecosystem map, from the Content Lexicon by Contentious (https://contentlexicon.com/content-ecosystem-map), CC BY 4.0", "body": "A map can show how a research institute's findings move from data to a publication to teaching material, press briefings and partner sites. It describes the places and connections that matter, rather than listing individual pages.\n\nAgreed boundaries help reduce the risk of sprawling over-complication. A map for a [website](https://contentlexicon.com/website) migration may focus on systems and update flows; one for a publishing strategy may also show teams and outside relationships. Missing connections can be discoveries, not evidence that the exercise has failed.\n\n## Why it matters for content\n\nContent rarely lives in one tidy place. A service update may appear on a website, in [email](https://contentlexicon.com/email) templates, in help articles, in chatbot knowledge, in staff guidance and in a [PDF](https://contentlexicon.com/pdf) that partners still distribute. A content ecosystem map helps teams understand where information travels, where duplication happens and where responsibility gets lost.\n\nIt's especially useful during [content strategy](https://contentlexicon.com/content-strategy), transformation, migration and governance work. Before changing [content operations](https://contentlexicon.com/content-operations), introducing AI, redesigning a site or improving quality, a team needs to understand the current landscape. The map shows dependencies, risks, ownership gaps and the places where one change may have consequences somewhere else.\n\n## How AI changes the picture\n\nAI agents add new routes through a content ecosystem. An agent may draw on a [knowledge base](https://contentlexicon.com/knowledge-base), supply answers in a customer portal and pass questions to a support team. Mapping those connections helps people see where an answer came from and which systems depend on a changed source.\n\nA team could then compare intended flows with observed use: does the agent retrieve the approved instructions or an older partner copy? The map helps frame that investigation; it doesn't itself mark sources as trustworthy or enforce exclusions. Include the selection and approval mechanisms where they matter.\n\n## Questions people ask\n\n**Is this the same as a content inventory?**\n\nNo. A [content inventory](https://contentlexicon.com/content-inventory) usually lists individual pieces of content, often page by page or item by item. A content ecosystem map shows places, systems, channels and relationships. The inventory tells you what exists. The map helps you see how the environment works.\n\n**How detailed should the map be?**\n\nDetailed enough to support the decision in front of you. A high-level map may be enough for strategic planning. A migration, governance review or AI-readiness project may need more detail about ownership, source systems, update flows and dependencies.\n\n**Who should be involved?**\n\nContent, digital, marketing, comms, product, support, sales, IT and operational teams may all know about different parts of the ecosystem. The map is often most useful because it brings these partial views together.\n\n**What makes a good ecosystem map?**\n\nA good map makes the relationships between its boxes clear. It should identify where content originates, where it's reused, where duplication exists, who owns what and which systems or channels depend on each other.\n\n## Learn more\n\n- [An Introduction to Content Ecosystem Maps](https://www.braintraffic.com/blog/an-introduction-to-content-ecosystem-maps) – Scott Kubie’s process for choosing boundaries, labelling relationships and checking the map with others; also explains why a map is not an inventory or workflow.\n- [How to Use a Content Ecosystem Map](https://www.braintraffic.com/blog/how-to-use-a-content-ecosystem-map) – Scott Kubie’s follow-up on what to do with a finished map: building shared understanding, choosing which problems to tackle first and updating it as channels and roles change." }, { "term": "Content editor", "alsoCalled": [ "Web content editor", "Web editor" ], "slug": "content-editor", "definition": "A role that edits, publishes and maintains an organisation's content so that each piece meets its standards and fits with the rest.", "category": "Roles", "related": [ "author-experience", "cms", "content-designer", "content-manager", "content-team", "copy-editing", "editing", "senior-content-editor", "sign-off", "writer" ], "updated": "2026-10-12T17:06:00.000Z", "url": "https://contentlexicon.com/content-editor", "markdown": "https://contentlexicon.com/content-editor.md", "credit": "Content editor, from the Content Lexicon by Contentious (https://contentlexicon.com/content-editor), CC BY 4.0", "body": "A content editor shapes, checks and publishes an organisation's content and keeps it in good order once it's live. In a typical week that might mean [editing](https://contentlexicon.com/editing) drafts from colleagues across the organisation, writing and updating pages themselves, building them in the [CMS](https://contentlexicon.com/cms) with the right [metadata](https://contentlexicon.com/metadata), links and images, checking [accessibility](https://contentlexicon.com/accessibility) and retiring pages that have outlived their purpose.\n\nEditing is the practice at the centre of the job; the role adds publishing, upkeep and a working knowledge of the systems involved. A good content editor has a clear sense of structure and hierarchy: what the reader needs first, which details belong together and what can be cut. Their remit includes reshaping content as well as polishing it; [proofreading](https://contentlexicon.com/proofreading) is the final check for errors.\n\nThe title stretches a long way. In publishing and journalism it can mean a commissioning or desk editor; in many charities, universities and public bodies it means the person who looks after the [website](https://contentlexicon.com/website). Web editor and digital editor often describe the same job, and some employers advertise [content design](https://contentlexicon.com/content-design) roles under it. A content editor usually works within priorities and standards set by a [content manager](https://contentlexicon.com/content-manager) or [senior content editor](https://contentlexicon.com/senior-content-editor), alongside [writers](https://contentlexicon.com/writer), [content designers](https://contentlexicon.com/content-designer) and subject experts.\n\nMany CMSs also have an “editor” permission level, which is an access right rather than a job, so describe the remit you need rather than trusting the title.\n\n## Why an organisation needs a content editor\n\nWhen many people can publish, nobody sees the whole. Departments add pages for their own purposes, terminology drifts, three versions of the same guidance appear and old pages linger because deleting them is nobody's job. A content editor reads across those boundaries: they notice the duplicate before it goes live, keep a consistent [voice](https://contentlexicon.com/voice) across contributions and know which pages depend on each other.\n\nCollaboration with [subject matter experts](https://contentlexicon.com/subject-matter-expert) is central to the role. Through questions, shared drafts and [pair writing](https://contentlexicon.com/pair-writing), a content editor helps colleagues turn specialist knowledge into content readers can use. Subject experts can contribute their knowledge without having to master the CMS, the [style guide](https://contentlexicon.com/style-guide) and accessibility requirements.\n\nThe role needs clear authority to send work back, query claims and decline to publish what doesn't meet the standard, with a route to escalate to a senior editor or [content owner](https://contentlexicon.com/content-owner). It can't resolve conflicting priorities between departments or make up for a missing strategy, and one content editor reviewing every page for a large organisation will soon become a bottleneck. Scale the role to the volume, or pair it with good guidance and training for contributors.\n\n## How AI changes the role\n\nMuch of what reaches a content editor has now been drafted or polished with AI, often without anyone saying so. That moves the job towards verification. A fluent submission no longer signals that the contributor understood the subject, so content editors need to be ready to ask where a claim came from, and need the authority to send back a page whose source nobody can name. A job description might now include checking AI-assisted drafts against source material and applying the organisation's [AI use policy](https://contentlexicon.com/ai-use-policy) to submissions.\n\nThe CMS is changing too. Many platforms now offer generated summaries, suggested [alt text](https://contentlexicon.com/alt-text), automatic tagging and translation, so some of what appears on a page was written by the system rather than a person. Someone has to decide which of those features to switch on, review their output and answer for the result, and the content editor is usually closest to it. Eileen Webb argued in 2014 that field labels and help text belong inside the CMS, where authors need them; where a platform's assistants draw on the same guidance, that writing now steers machines as well as people. A content editor who can write it clearly, test what it produces and explain problems to developers is worth more than one who only operates the interface.\n\nRoutine formatting and tagging may take less of the week, but accountability for what gets published stays where it was.\n\n## Questions people ask\n\n**Is a content editor the same as a content designer?**\n\nThe titles overlap and some employers use them interchangeably. Content design usually starts from research into [user needs](https://contentlexicon.com/user-needs) and decides what content should exist and in what form, often within a product or service team. A content editor more often works on material that colleagues have drafted, and on publishing and maintaining it. Plenty of people do both; the job description should say which matters most.\n\n**Should our content editor also write?**\n\nMost content editors write or rewrite a good share of what they publish, and writing keeps their editing practical. If writing takes over, though, the reviewing and upkeep that justified the role tend to slip.\n\n**Where should a content editor sit?**\n\nSomewhere they can see across the organisation's publishing, commonly a central digital, communications or [content team](https://contentlexicon.com/content-team) reporting to a content manager or senior content editor. Close working relationships with subject experts and developers matter as much as the reporting line.\n\n**What should we look for when hiring?**\n\nStrong editing and writing, a clear sense of structure and hierarchy, comfort with a CMS and the ability to work closely with subject matter experts, plus the tact to tell a senior colleague that their page isn't ready. Journalism, publishing, communications and web production all lead into the role. A short editing test using one of your own pages will tell you more than a portfolio of polished articles; ask candidates to explain changes to its structure and emphasis as well as its wording.\n\n## Learn more\n\n- [Training the CMS](https://alistapart.com/article/training-the-cms/) – Eileen Webb's A List Apart article on putting editorial guidance into CMS field labels and help text; written for the people who set up a CMS, it's also a revealing picture of what content editors need from one.\n- [Web content editor](https://nationalcareers.service.gov.uk/job-profiles/web-content-editor) – the National Careers Service job profile for England, a useful snapshot of the role's typical tasks, entry routes and progression, and of how far the title overlaps with content designer and digital editor." }, { "term": "Content governance", "alsoCalled": [], "slug": "content-governance", "definition": "The ownership, rules and decision-making that keep content accurate, consistent and maintained over time.", "category": "Governance", "related": [ "ai-governance", "centralised-governance", "content", "content-operating-model", "content-operations", "content-owner", "content-plan", "content-policy", "decentralised-governance", "editorial-board", "head-of-content", "hub-and-spoke", "human-in-the-loop", "single-source-of-truth", "stakeholder-mapping", "style-guide", "workflow" ], "updated": "2026-10-07T14:44:00.000Z", "url": "https://contentlexicon.com/content-governance", "markdown": "https://contentlexicon.com/content-governance.md", "credit": "Content governance, from the Content Lexicon by Contentious (https://contentlexicon.com/content-governance), CC BY 4.0", "body": "Governance describes how authority and accountability are shared out: who decides, on what basis and who answers for the result. Corporate governance does this for a company’s board and management; data governance does it for information such as customer records. Content governance applies the same thinking to the words, images, pages and messages an organisation publishes.\n\nIt usually includes named [content owners](https://contentlexicon.com/content-owner) for pages or topics, the standards content should meet (a [style guide](https://contentlexicon.com/style-guide), [content principles](https://contentlexicon.com/content-principles), [accessibility](https://contentlexicon.com/accessibility) requirements), permissions to create, approve and publish, a schedule for review and a way of settling disagreements and exceptions. It covers the whole life of a piece of content, from whether it should exist at all to when it should be updated, archived or removed.\n\n## Why it matters for content\n\nWithout governance, content quality depends on memory, goodwill and whichever diligent person happens to be in the room. With it, teams know who owns a page, who can approve a change, how often material should be reviewed, which standards apply and what happens when something is disputed or outdated. The need grows wherever content is shared across teams, published in several places, subject to regulation, translated, reused or kept for years. Governance is what stops content becoming a pile of reasonable exceptions.\n\nIt works best when it’s built into ordinary work: an owner field in the [CMS](https://contentlexicon.com/cms), a [review date](https://contentlexicon.com/review-date) that produces a reminder, a clear route for exceptions and someone with authority to settle them. A policy document can record the arrangement, but it can’t carry it out.\n\nProportion matters. A small team may need one named owner, a shared standard and a monthly review of risky content. A regulated service spread across several teams will need firmer permissions, evidence of approval and a clearer escalation route. Copying a large organisation’s system into a small team adds ceremony without much benefit, while running a regulated service on informal arrangements creates risk.\n\n## How AI changes the picture\n\nAI increases the number of ways content can be created, edited, summarised and repurposed. That makes governance more important because the questions multiply: which source is canonical, which instructions should the assistant follow, who can approve AI-generated changes, what should be reviewed by a person and what should never be automated.\n\nGovernance needs to be reflected in the tools and [workflows](https://contentlexicon.com/workflow) an assistant actually uses. A written rule that only approved pages may be published is useful, but publishing permissions and review steps must support it.\n\n## Questions people ask\n\n**How is governance different from content strategy?**\n\nStrategy sets direction and priorities. Governance deals with ownership, rules and decision-making so that the work can be carried out consistently.\n\n**Is governance the same as workflow?**\n\nNot quite. Workflow is the route content takes through creation, review and maintenance. Governance is the set of roles, rules and responsibilities that shape that route.\n\n**Who should own content governance?**\n\nUsually someone with enough authority to settle disputes across teams, supported by owners for individual pages or topics who know the subject. The right arrangement depends on how much content there is and how risky it is, not on a set number of people.\n\n**Does governance always slow things down?**\n\nPoorly designed governance can add delays. Clear ownership and proportionate checks can reduce repeated decisions and disputes, though some reviews take time because the consequences warrant it.\n\n## Learn more\n\n- [Content governance: let’s make it work](https://www.braintraffic.com/blog/content-governance-lets-make-it-work) – Brain Traffic on recognising content responsibilities in people’s time, resources and performance expectations after launch.\n- [Content Governance 101](https://www.contensis.com/community/blog/content-governance-101) – Robert Mills’s practical sequence for roles, standards, review and maintenance, followed by a short Contensis and Insytful product section.\n- [Why your content problems are organisational, not editorial](https://contentmaturity.com/learn/bakery-not-bread) – the bakery-not-bread argument for fixing ownership, standards and systems rather than repeatedly polishing symptoms." }, { "term": "Content health check", "alsoCalled": [], "slug": "content-health-check", "definition": "A recurring, criteria-based assessment of how well content serves its purpose and what needs attention.", "category": "Research and evidence", "related": [ "accessibility", "automated-testing", "content-audit", "content-maintenance", "content-standards", "kpis", "readability" ], "updated": "2026-10-11T06:24:00.000Z", "url": "https://contentlexicon.com/content-health-check", "markdown": "https://contentlexicon.com/content-health-check.md", "credit": "Content health check, from the Content Lexicon by Contentious (https://contentlexicon.com/content-health-check), CC BY 4.0", "body": "A content health check looks at live content – a [website](https://contentlexicon.com/website), a help centre, an intranet or the [knowledge base](https://contentlexicon.com/knowledge-base) behind an assistant – and rates it against criteria such as accuracy, clarity, [accessibility](https://contentlexicon.com/accessibility) and usefulness. Each round records ratings alongside the findings that explain them. Because the criteria and usually the sample stay the same, one round can be compared with the last, so a team sees whether things are getting better or worse rather than a single snapshot.\n\nThe name borrows from medicine. Like a routine check-up, it's preventive and comparative: it looks for early signs of trouble against a known baseline instead of waiting for a crisis. Like screening, it's better at showing where to look more closely than at explaining causes. The practice belongs to the same family as the qualitative [content audit](https://contentlexicon.com/content-audit), the heuristic review and the content scorecard. What sets it apart is repetition with consistent criteria.\n\n## Why it matters for content\n\nContent rarely fails all at once. A price changes, a linked form moves or a policy is revised, and the page explaining it stays as it was. A page may pass a spell check yet describe a service that no longer exists. A health check brings different kinds of evidence together so a team can decide whether to correct, investigate, merge or retire something, instead of treating every problem as a rewrite.\n\nChecks can work page by page or follow the tasks people come to do, such as applying for a grant or cancelling a subscription. Following a task exposes problems that sit between pages, like a missing step or a dead end, which a page-by-page review can miss. Repeated results help teams maintain content, prioritise and test whether improvements last.\n\n## How AI changes the picture\n\nRecurring checks used to be limited by the hours available for reading. Language models can apply written criteria across thousands of pages, which makes it practical to check a whole estate regularly rather than a sample once a year. They can also compare content with a source of truth, matching support answers against a revised procedure and highlighting the passages that now disagree. A useful flag shows the evidence behind it, so a reviewer can see why a page scored badly and judge whether the assessment holds.\n\nWhat needs checking is changing too. Assistants and AI search draw on help articles, policies and knowledge bases to answer questions, so a weak source may affect many answers that nobody reads individually. Health checks increasingly need to cover that source material and, where possible, samples of the answers built from it.\n\nAn automated checker is an instrument, and instruments need calibrating. Test it against examples where the right answer is known, and be clear about criteria it can't assess from the text alone: whether content is useful may depend on research, service data or what people do next. Automating a judgement doesn't establish that the criterion or the result is sound.\n\n## Questions people ask\n\n**How is a health check different from a content audit?**\n\nA health check is a kind of audit, focused on quality and repeated over time. A full content audit often starts with an inventory of everything that exists and ends with decisions about what to keep, merge or remove, typically before a redesign or migration. A health check keeps watching how well it's doing.\n\n**Which criteria should we use?**\n\nChoose criteria that reflect what your content is for and that reviewers can apply consistently. Most frameworks cover accuracy, usefulness, clarity, accessibility, structure and currency, with additions for [voice](https://contentlexicon.com/voice), brand or regulation where they matter. Published frameworks and scorecard templates save starting from scratch. Whichever you use, write down what each rating means, so that two reviewers looking at the same page would reach a similar score.\n\n**Does a high score prove that content works?**\n\nIt shows that content meets the criteria as assessed. Whether it helps people still needs evidence from use, such as task completion, search behaviour, support contacts or research with readers. A health check is diagnostic; testing with real people is how you confirm the diagnosis.\n\n**Who should run it?**\n\nWhoever is responsible for keeping live content useful, with subject experts checking accuracy. A small team might review its ten most important pages each quarter; a large publisher might combine automated assessment across the estate with expert review of a risk-based sample. A second reviewer, or a periodic comparison of scores between colleagues, helps keep individual bias in check.\n\n## Learn more\n\n- [How to evaluate content using a content scorecard](https://contentstrategyinc.com/how-to-use-a-content-scorecard/) – Content Strategy Inc.'s method for scoring content by walking through user scenarios, with the caution that expert review is diagnostic and still needs testing with real customers.\n- [Taming the content monster](https://medium.com/designing-atlassian/taming-the-content-monster-e8be5f937e72) – Samantha Campbell's account of scoring a 282-page Atlassian content hub against criteria adapted from Nielsen's usability heuristics, which found that many of the most visited pages scored medium to low.\n- [Content Health Check](https://contenthealthcheck.com/) – Contentious's tool for recurring, AI-assisted checks of live content against the 15 criteria of its Content Health Framework." }, { "term": "Content inventory", "alsoCalled": [], "slug": "content-inventory", "definition": "A structured list of the content you have and the basic facts about each item.", "category": "Research and evidence", "related": [ "content", "content-audit", "content-ecosystem-map", "content-owner", "content-plan", "database", "gap-analysis", "knowledge-base", "metadata", "pdf", "rot-content" ], "updated": "2026-10-11T06:24:00.000Z", "url": "https://contentlexicon.com/content-inventory", "markdown": "https://contentlexicon.com/content-inventory.md", "credit": "Content inventory, from the Content Lexicon by Contentious (https://contentlexicon.com/content-inventory), CC BY 4.0", "body": "The facts recorded for each item in a content inventory might include URL, title, type, location, owner, date, status or channel. The inventory answers the question \"what have we got?\" before anyone moves on to \"is it any good?\"\n\nIt isn't yet an evaluation, which is why it often sits underneath a [content audit](https://contentlexicon.com/content-audit), migration, governance project or redesign.\n\n## Why it matters for content\n\nA content inventory makes the content estate visible. It shows how much exists, where it lives, what kinds of content are present and how the estate is distributed across channels or sections. That visibility is useful on its own, and it also provides the raw material for audits, migrations, maintenance plans and governance work.\n\nThe need is sharpest when an organisation has inherited a large body of content, when several systems are involved, when nobody's quite sure what exists, or when a team is about to make decisions that depend on the shape of the estate.\n\n## How AI changes the picture\n\nAn inventory of AI source material needs facts that a list of page URLs won't capture: which version is approved, who can access it, where the original lives and when it's reviewed. [Prompt](https://contentlexicon.com/prompt) libraries and [knowledge bases](https://contentlexicon.com/knowledge-base) belong in the inventory when they influence the system's output.\n\nCrawlers and AI tools can compile or enrich an inventory far faster than someone with a spreadsheet and a free week: [Content Health Check](https://contenthealthcheck.com), for one, builds an inventory of the pages it can crawl before scoring them. That doesn't remove the need for an agreed structure. The inventory still needs fields, ownership and a clear purpose if it's to be useful rather than merely large. Plenty of inventories are built with great care for a migration and never opened again once it's done.\n\n## Questions people ask\n\n**What's the difference between an inventory and an audit?**\n\nAn inventory lists what exists. An audit goes further and evaluates quality, usefulness and what should happen next.\n\n**Do we need to include every file?**\n\nNot always. It depends on the decision the inventory is meant to support. Some inventories are exhaustive; others are deliberately scoped.\n\n**Is a sitemap enough?**\n\nRarely. A sitemap lists URLs; an inventory adds the facts you need to act on them, such as type, owner, status and channel.\n\n**Can a content inventory be useful after the project ends?**\n\nYes, if it's maintained. An up-to-date inventory becomes an operational tool rather than a one-off artefact.\n\n## Learn more\n\n- [Content Inventory and Auditing 101](https://www.nngroup.com/articles/content-audits) – a practical distinction between recording what exists and evaluating it.\n- [The UX Cookbook content inventory](https://theuxcookbook.com/content-inventories) – Emma Gould's spreadsheet method, illustrated with the University of Arizona library's inventories.\n- [Content inventory, audit and management](https://web.wsu.edu/web-training/resources/website-maintenance/content-inventory-audit-management/) – Washington State University's inventory and audit workflow, with a spreadsheet template and fields for documents and multimedia." }, { "term": "Content lifecycle", "alsoCalled": [ "Content life cycle" ], "slug": "content-lifecycle", "definition": "The stages content passes through, from idea to retirement.", "category": "Operations", "related": [ "content-calendar", "content-decay", "content-maintenance", "content-operations", "content-owner", "evergreen-content", "retention-and-archiving", "review-date", "workflow" ], "updated": "2026-10-11T06:24:00.000Z", "url": "https://contentlexicon.com/content-lifecycle", "markdown": "https://contentlexicon.com/content-lifecycle.md", "credit": "Content lifecycle, from the Content Lexicon by Contentious (https://contentlexicon.com/content-lifecycle), CC BY 4.0", "body": "A campaign page may be planned, drafted, reviewed and published over a few weeks, then kept as a record after the campaign ends. A service guide may instead cycle through review and revision for years. The stages and their order depend on the content.\n\nPlanning the later stages means deciding what will trigger a review, who can approve a change and when the item should stop being presented as current. Publication is a milestone, not the end of that responsibility.\n\n## Why it matters for content\n\nThinking in lifecycles changes what counts as finished. If the plan stops at publication, it can leave later changes and accountability unaddressed. If the lifecycle includes maintenance, review and retirement, then accuracy, relevance and usefulness become ongoing responsibilities with owners and dates attached.\n\nIt also makes cost visible. A page isn't a fixed expense paid once at creation; it carries a small recurring bill for every review, correction, translation and migration it survives. Budgeting for those stages helps a team decide what it can sustain.\n\nThe lifecycle matters most before a big project – a new site, a campaign, a [content type](https://contentlexicon.com/content-type) – because that's when it's easy to design review points in, and painfully expensive to add them later.\n\n## How AI changes the picture\n\nFaster drafting can increase the work arriving at review, but AI can also support later stages. It can compare a guide with a revised source, prepare a change summary or suggest overlapping items to merge. Routine age and link checks may need no AI at all.\n\nAn assistant introduces another place where retired content may still be used. Inaccurate guidance was already harmful before AI; retrieval can spread it through additional conversations. A retirement plan should therefore consider search indexes and connected knowledge sources as well as the public page, without deleting records that must be retained.\n\n## Questions people ask\n\n**Do small teams need a formal lifecycle?**\n\nThey need the review and retirement stages more than the ceremony. Start with an owner, a review trigger and a decision about what happens when the need ends. A date helps only if someone acts on it; high-risk material may need event-driven review too.\n\n**Is the lifecycle the same as a workflow?**\n\nNo. A [workflow](https://contentlexicon.com/workflow) describes how a particular piece of work moves between people and states. The lifecycle is the longer arc of the content itself, including the years after publication.\n\n**When should content be retired rather than updated?**\n\nWhen the need it met has gone, when it duplicates something better, or when keeping it accurate costs more than the value it delivers.\n\n**Who owns the later stages?**\n\nSomebody should, by name. Lifecycle stages without an owner are aspirations, not stages.\n\n## Learn more\n\n- [The 7 stages of enterprise content lifecycle management](https://www.techtarget.com/searchcontentmanagement/tip/7-key-stages-of-enterprise-content-lifecycle-management) – Jonathan Bordoli’s enterprise and records-management model, including categorisation, reuse and removal.\n- [What Is a Content Lifecycle?](https://review.content-science.com/what-is-content-lifecycle/) – Content Science Review’s definition of the stages from planning and creation through maintenance and reuse to retirement.\n- [Why a recurring content health check beats a one-off audit](https://contenthealthcheck.com/learn/beyond-the-snapshot) – how a baseline and trend turn quality checking from an occasional project into part of the content lifecycle." }, { "term": "Content maintenance", "alsoCalled": [], "slug": "content-maintenance", "definition": "The ongoing work of keeping content accurate, relevant and healthy after publication.", "category": "Operations", "related": [ "content-costs", "content-debt", "content-decay", "content-lifecycle", "content-operations", "content-owner", "continuous-integration", "editing", "review-date", "rot-content", "technical-debt" ], "updated": "2026-10-11T06:24:00.000Z", "url": "https://contentlexicon.com/content-maintenance", "markdown": "https://contentlexicon.com/content-maintenance.md", "credit": "Content maintenance, from the Content Lexicon by Contentious (https://contentlexicon.com/content-maintenance), CC BY 4.0", "body": "Content maintenance is the work of keeping published content accurate, relevant and usable for as long as it stays live. It starts from the premise that publication creates a responsibility rather than completing one. Some of that work is planned, such as an annual look at a guidance page. Some is prompted by an event: a policy change, a new product, a rebrand, a complaint or a broken link.\n\nWhatever sets it off, the work itself usually takes one of four forms. Corrections fix errors as they're found. Reviews check content against current evidence, ideally at intervals set by risk and recorded with a [review date](https://contentlexicon.com/review-date) and an owner. Consolidation merges duplicates and overlaps so there's one canonical version. Retirement removes or archives content that has done its job, with redirects so links and bookmarks still lead somewhere useful.\n\n## Why it matters for content\n\nMaintenance protects people from outdated instructions and organisations from avoidable risk. It also improves [findability](https://contentlexicon.com/findability), because removing duplicates and old copies stops current content being buried beneath historical versions.\n\nA mature approach assigns owners, review triggers and sensible cycles. High-risk or volatile material needs closer attention than stable background information, and reviews prompted by policy, product or personnel changes often catch more than annual reminders. Setting those intervals and owners is covered under Review date.\n\nMaintenance also needs time and recognition. If everyone’s time and effort gets spent on new publication, the organisation rewards growth and leaves decay to accumulate. Budgeting for upkeep, and reporting it alongside new content, makes the work visible to the people who fund it.\n\n## How AI changes the picture\n\nAutomated checks can find broken links or lapsed review dates. AI can compare content with supplied sources, flag possible contradictions and suggest which items need attention first. It needs evidence of a change, though: plausible wording alone can't establish whether a fact is out of date.\n\nA flagged item still needs a responsible owner. A system may notice that a service promise no longer matches approved guidance, but consequential changes need someone with the evidence and authority to approve them.\n\nAI also creates new material to maintain: [prompts](https://contentlexicon.com/prompt), knowledge sources, retrieval indexes and generated answers. These belong in the same lifecycle, with owners, review and retirement, rather than piling up indefinitely. An assistant drawing on a retired page will keep repeating it until someone removes it from the index as well as the [website](https://contentlexicon.com/website).\n\n## Questions people ask\n\n**How often should content be reviewed?**\n\nAccording to risk, rate of change and use. A page about fees might need checking every few months and whenever the fees change, while a background explainer might manage with a yearly look. Combine scheduled checks with triggers for when the underlying facts or services change.\n\n**Who owns maintenance?**\n\nA named role with subject authority and access to editorial support. \"The team\" is rarely a usable owner, because nobody in particular is prompted to act.\n\n**Should low-traffic content be deleted?**\n\nInvestigate its purpose first. Low use may point to poor findability, a rare but critical need or content that no longer earns its place, and each calls for a different response.\n\n**Can AI maintain content automatically?**\n\nIt can handle checks and low-risk suggestions, such as fixing a broken link to a page that has moved. Accountability and high-impact updates still need people.\n\n## Learn more\n\n- [Web content has a maintenance problem](https://gerrymcgovern.com/web-content-has-a-maintenance-problem/) – Gerry McGovern’s critique of publishing cultures and automation promises that leave upkeep unfunded.\n- [Hail the maintainers](https://aeon.co/essays/innovation-is-overvalued-maintenance-often-matters-more) – Andrew Russell and Lee Vinsel's essay on the overlooked work of keeping things going. Its wider argument about valuing upkeep over novelty helps explain why content maintenance struggles for recognition and funding.\n- [Create less crud](https://contentious.ltd/thoughts/create-less-crud/) – the case for clearing out abandoned, duplicated and decaying content instead of endlessly adding more." }, { "term": "Content manager", "alsoCalled": [], "slug": "content-manager", "definition": "A role responsible for the day-to-day running of an organisation's content, from planning and production to quality and upkeep.", "category": "Roles", "related": [ "cms", "content-editor", "content-operations-manager", "content-owner", "content-strategist", "content-team", "head-of-content", "senior-content-editor", "workflow" ], "updated": "2026-10-11T06:24:00.000Z", "url": "https://contentlexicon.com/content-manager", "markdown": "https://contentlexicon.com/content-manager.md", "credit": "Content manager, from the Content Lexicon by Contentious (https://contentlexicon.com/content-manager), CC BY 4.0", "body": "A content manager is accountable for what an organisation publishes from week to week: what gets planned and commissioned, whether it meets the standard when it goes out and whether it's still right a year later. The job usually involves managing people as well as pages. That might mean a small team of [writers](https://contentlexicon.com/writer) and editors, or the many contributors elsewhere in the organisation who publish occasionally and need guidance, deadlines and a second pair of eyes.\n\nThe title stretches a long way. It can describe one person running a charity [website](https://contentlexicon.com/website) alongside newsletters and [social media](https://contentlexicon.com/social-media), or someone leading a team of editors across a large reference portfolio. Seniority follows the remit, so agree what the person will own and decide before comparing salaries with similarly named jobs elsewhere.\n\n## Why an organisation needs a content manager\n\nOrganisations need content managers because content work crosses teams. Someone needs to help people coordinate what they’re creating, share knowledge and make sure the pieces work together. That also includes looking after content over time: reviewing it, updating it and retiring it when it’s no longer useful. A content manager can make that work visible and help clarify ownership, but the responsibility is shared across the organisation.\n\nThe need tends to become obvious at a recognisable moment: when publishing spreads beyond one team, when a website outgrows the person who built it or when an agency finishes a project and nobody inside can say what happens next. Agencies can produce and advise, but somebody in the organisation has to own the estate between projects. To do that, the role needs authority that contributors recognise: to set standards they follow, to turn down a page that duplicates another and to retire one that's past its use. A content manager can't make up for a strategy nobody has agreed or a workload nobody has resourced. Nor should they approve every change personally; delegating routine decisions stops them becoming the queue everyone waits in.\n\n## How AI changes the role\n\nThe balance of the job moves from producing and checking content towards running the arrangements that produce it. Where assistants draft, adapt or check material, someone has to decide which tasks they take on, maintain the guidance, examples and approved sources they draw from, and notice when a policy change or a new style rule leaves their instructions out of date. Without a separate operations role, much of that falls to the content manager. Where there is one, the content manager decides what assistants should produce and to what standard, and the operations manager builds and runs the arrangements that make it happen. Reviewing becomes a matter of patterns as well as pages: whether an assistant keeps misreading a particular kind of request, say, and what needs to change so it stops.\n\nJob descriptions are starting to catch up. Employers increasingly mention AI familiarity, and some now list [workflow](https://contentlexicon.com/workflow) automation among a content manager's responsibilities. A useful description might add evaluating AI output against the organisation's standards, keeping sources and guidance current and agreeing where automated steps need a human check. The person still needs strong editorial judgement, because they're answerable for a system that produces content as well as for the content itself.\n\nCheap production also gives more weight to a long-standing part of the role: deciding what shouldn't be published. When another page costs almost nothing to draft, a content manager's authority to say no, merge two pages or retire an old one can do more for readers than any amount of extra output.\n\n## Questions people ask\n\n**Where does a content manager sit among similar roles?**\n\nClose to several. A [content strategist](https://contentlexicon.com/content-strategist) sets direction and designs how content should work; a content manager runs it week by week. A [senior content editor](https://contentlexicon.com/senior-content-editor)'s authority centres on editorial direction, commissioning and standards, and many organisations treat the two titles as one job. A [content operations manager](https://contentlexicon.com/content-operations-manager) looks after the workflows, tools and routines everyone depends on, which smaller teams usually fold into the content manager's remit.\n\n**Do we need a full-time content manager?**\n\nNot always. Volume, risk and the number of people publishing decide it. A small organisation with one website and a handful of contributors might combine the role with communications work. Signs that it needs dedicated time include nobody knowing roughly how much content exists, reviews that are always overdue and pages that contradict each other.\n\n**Who should a content manager report to?**\n\nUsually whoever leads communications, marketing or digital, depending on where content decisions are made. Choose the line that gives the role backing across teams, since much of its work involves asking colleagues in other departments to change how they publish.\n\n**What background makes a good content manager?**\n\nJournalism, [editing](https://contentlexicon.com/editing), communications, [content design](https://contentlexicon.com/content-design) and digital project work all lead into it. The combination matters more than the route: sound editorial judgement and enough comfort with systems to work alongside developers, along with the patience to persuade contributors who don't report to you.\n\n## Learn more\n\n- [We Analyzed 8,000 Content Marketing Job Listings](https://www.semrush.com/blog/content-marketing-job-market-study) – Semrush's 2026 study of US job ads, which finds employers asking less for writing alone and more for ownership, analytics and workflow automation; a snapshot of one market and one sector, not a picture of every content role.\n- [Lisa Welchman: content, AI, and digital governance](https://contentandai.com/lisa-welchman/) – a Content and AI podcast conversation in which the digital governance specialist recalls that maintaining Cisco's product pages made her, as a junior employee, one of the most powerful people in its ecosystem, and argues that the people closest to new technology have more agency than they think." }, { "term": "Content marketing", "alsoCalled": [], "slug": "content-marketing", "definition": "Creating and sharing content to build an audience relationship in support of marketing goals.", "category": "Foundations", "related": [ "channel-strategy", "competitor-analysis", "content", "content-strategy", "copywriting", "email", "social-media", "storytelling" ], "updated": "2026-10-10T17:53:00.000Z", "url": "https://contentlexicon.com/content-marketing", "markdown": "https://contentlexicon.com/content-marketing.md", "credit": "Content marketing, from the Content Lexicon by Contentious (https://contentlexicon.com/content-marketing), CC BY 4.0", "body": "Content marketing includes articles, guides, newsletters, podcasts, videos, social posts and the downloadable report that asks for an [email](https://contentlexicon.com/email) address before revealing page two.\n\nThe stated promise is value before promotion. At its best, content marketing helps people understand a subject and builds trust over time. At its worst, it treats every question as an opportunity to herd someone into a funnel, producing a great deal of material that serves the publishing target more than the reader.\n\n## Why it matters for content\n\nContent marketing has shaped much of the modern web. It funds useful explanations and original research, but it also rewards volume, search traffic and constant publication. When the measure is how many leads enter the system, teams can lose sight of whether the content is accurate, distinctive or needed.\n\nThe strongest work begins with audience needs and organisational knowledge. It offers something worth a person’s attention, then makes the commercial relationship clear. Useful content may support marketing as a consequence of being useful: it demonstrates expertise, creates familiarity and gives people a reason to return or recommend the organisation.\n\nThis demands restraint. Not every product announcement requires a “thought leadership” article and not every keyword deserves a page. Marketing goals should sit alongside editorial standards, evidence, [accessibility](https://contentlexicon.com/accessibility) and maintenance. Content that wins a click but wastes someone’s time has achieved rather less than the dashboard suggests.\n\n## How AI changes the picture\n\nA small specialist team can use AI to adapt a well-researched publication into formats it couldn't otherwise support. A gardening supplier could turn its own seasonal trials into short explanations for different growing conditions, with the underlying evidence available to readers. That adds something more useful than several versions of a generic planting guide.\n\nThe same tools can produce invented expertise or personalisation that feels intrusive. Setting clear boundaries around claims and data, and judging each format by what readers gain from it, keeps both risks in check. As discovery moves into AI-generated answers, a publication may influence decisions without attracting a visit; clicks alone won't describe that relationship.\n\n## Questions people ask\n\n**Is all organisational content content marketing?**\n\nNo. Service information, policies, support content and product instructions primarily help people complete tasks. They may affect reputation or sales, but that doesn't make marketing their main purpose.\n\n**What makes content marketing ethical?**\n\nMake the source and commercial aim clear, respect consent and avoid misleading claims. If a report asks for contact details, explain what happens to those details before someone submits them.\n\n**Does useful content still need a call to action?**\n\nNot always. A relevant next step can help, but the content should still deliver on its own promise. Withholding the useful part until after conversion is a different sort of transaction.\n\n**Can AI write our content marketing?**\n\nIt can assist with research, drafting and adaptation. People still need to supply evidence, judgement, distinctive ideas and responsibility for what the organisation says.\n\n## Learn more\n\n- [What is content marketing?](https://contentmarketinginstitute.com/what-is-content-marketing) – CMI’s definition connects audience value with marketing outcomes; its benefit claims express the practice’s own case rather than independent proof.\n- [John Deere's 'The Furrow' - World's Oldest Content Marketing](https://contently.com/2013/10/03/the-story-behind-the-furrow-2/) – Contently’s history of the farming magazine John Deere, which launched in 1895 and reached a circulation of 4 million by 1912.\n- [Content marketing maturity vs content maturity](https://contentmaturity.com/learn/content-marketing-maturity-vs-content-maturity) – why a capable marketing operation is only one part of organisation-wide content maturity." }, { "term": "Content maturity", "alsoCalled": [], "slug": "content-maturity", "definition": "How capable an organisation is at making, managing and improving content.", "category": "Foundations", "related": [ "content-governance", "content-health-check", "content-operating-model", "content-strategy" ], "updated": "2026-10-08T16:39:00.000Z", "url": "https://contentlexicon.com/content-maturity", "markdown": "https://contentlexicon.com/content-maturity.md", "credit": "Content maturity, from the Content Lexicon by Contentious (https://contentlexicon.com/content-maturity), CC BY 4.0", "body": "Content maturity describes the system behind the work rather than any single piece of it: strategy, culture, day-to-day operations, the content itself and the infrastructure underneath. Two organisations can publish equally strong material today. The mature one will probably still be doing so in three years, at lower cost and with less heroic effort, because good work is built into how it operates.\n\nThe idea borrows from maturity models, which began in software engineering. In the late 1980s, Watts Humphrey's Capability Maturity Model described how organisations move from improvised, hero-dependent work to processes that are defined and steadily improved. Applied to content, it asks whether good results come from a system or from heroism. It covers everything an organisation produces, from reports and help articles to forms and policies, as well as marketing.\n\n## Why it matters for content\n\nContent problems often survive rewrites because their causes sit elsewhere. An audit may reveal outdated pages; a maturity lens asks why they became outdated and how to stop it happening again. Doing content well is closer to running a bakery than baking a loaf. A gifted baker can rescue one batch, but the shop depends on working ovens, reliable suppliers, a sensible rota and recipes everyone can find. In the same way, stale pages usually point to a missing review rhythm, and a queue of drafts awaiting [sign-off](https://contentlexicon.com/sign-off) is a [workflow](https://contentlexicon.com/workflow) problem. Neither is solved at the keyboard.\n\nGood content can also hide a fragile system. A skilled team in an immature organisation still produces excellent work, but through late nights, persistent chasing and knowledge nobody has written down. Nothing looks wrong until a key person leaves, the volume doubles or a new channel arrives.\n\nMaturity changes the economics, too. Where nobody owns content, the same work gets paid for repeatedly: a page decays, embarrasses someone and is rewritten from scratch. Capabilities depend on one another, so a good [CMS](https://contentlexicon.com/cms) can't compensate for absent ownership. Even so, maturity isn't a race towards bureaucracy. The right level depends on scale, risk and ambition.\n\n## How AI changes the picture\n\nAI amplifies whatever operating model it meets. With documented standards, clear ownership and a review step before publication, it can help a team draft, adapt and maintain good content with less effort. Where conventions live in one person's head, it produces more pages carrying the same inconsistencies, only faster. An organisation that can't explain to a new freelancer how it sounds won't be able to explain it to an agent either.\n\nA team with strong [writers](https://contentlexicon.com/writer) but no reliable way to maintain its assistant's sources has a maturity gap. Seen that way, the choice between source ownership, retrieval testing or a different service design becomes an organisational decision rather than a tool purchase. There's encouragement here too: the standards, structure and ownership that make content better for people are the same ones AI depends on. Progress means being better able to sustain useful content, not merely producing more of it.\n\n## Questions people ask\n\n**Is content maturity the same as content quality?**\n\nNo. Quality describes content; maturity describes the organisation's capability to produce and sustain it. Strong pages from an immature organisation usually mean someone is holding things together by effort alone.\n\n**Does higher maturity mean more process?**\n\nNot necessarily. Mature processes are proportionate, clear and useful, not elaborate for their own sake.\n\n**Can a small organisation be mature?**\n\nYes. Scale changes the form of systems, not the need for clarity and responsibility. A three-person team with shared standards and a habit of reviewing its pages can be more mature than a large department running on goodwill.\n\n**Where should improvement begin?**\n\nChoose a weakness with a clear consequence. If updates keep stalling because nobody can approve a change, agreeing decision rights may help more than buying another writing tool. Check whether that intervention improves the work before moving on.\n\n## Learn more\n\n- [What is content maturity?](https://contentmaturity.com/learn/what-is-content-maturity) – a clear definition of content capability, its five constituent areas and how an organisation can assess it.\n- [Seven ways to use content maturity models](https://uxcontent.com/how-to-measure-content-operations-maturity) – a useful comparison of several models and, more importantly, what teams can actually do with one.\n- [What Is the State of Content Operations in 2021?](https://review.content-science.com/what-is-the-state-of-content-operations-now) – Content Science’s 2021 survey of 109 professionals through its own maturity model; historical self-reported findings, with the full report behind a subscription." }, { "term": "Content model", "alsoCalled": [], "slug": "content-model", "definition": "A structured definition of the content types an organisation has and the fields, parts and rules that make each one work.", "category": "Structure and systems", "related": [ "backend", "cms", "content-type", "domain-model", "paddle-swim-dive", "schema", "separation-of-concerns", "structured-content", "taxonomy", "technical-debt", "ubiquitous-language" ], "updated": "2026-10-06T15:16:00.000Z", "url": "https://contentlexicon.com/content-model", "markdown": "https://contentlexicon.com/content-model.md", "credit": "Content model, from the Content Lexicon by Contentious (https://contentlexicon.com/content-model), CC BY 4.0", "body": "A good content model describes the structure behind the content, not the design wrapped around it.\n\nA simple way to think about it is as the blank form behind a finished page. If a case study always needs a client, a problem, an approach, an outcome and a [call to action](https://contentlexicon.com/call-to-action), the content model makes those parts explicit. That helps [writers](https://contentlexicon.com/writer), designers, developers and systems all work from the same shape.\n\n## Why it matters for content\n\nA content model makes content more reusable, more consistent and easier to govern. It lets the same piece of content appear in more than one place without being copied and pasted. It helps teams spot gaps and overlaps between [content types](https://contentlexicon.com/content-type). It also makes publishing systems easier to build and easier to change, because the structure is clear before anyone starts designing templates or writing code.\n\nContent modelling earns its keep when an organisation has several content types, several channels or several people creating content to different standards. It's especially useful when the same information appears across a [website](https://contentlexicon.com/website), app, chatbot, [PDF](https://contentlexicon.com/pdf) or internal [knowledge base](https://contentlexicon.com/knowledge-base), or when a team wants to move from page-by-page publishing to a more structured approach.\n\n## How AI changes the picture\n\nLanguage models handle explicit structure far better than structure implied in prose. A model tells an AI assistant what it's meant to draft, which fields matter, which parts are optional and what \"done\" looks like for a given content type. It also makes generated content easier to check, because the expected shape is explicit rather than guessed: a case study that comes back without an outcome is visibly incomplete, not just vaguely unsatisfying.\n\nRelationships matter as much as fields. A model could connect events to venues, access information and booking conditions. An assistant could then help a visitor find a workshop with step-free access and understand whether a companion needs a separate ticket, drawing on related records rather than guessing from one event description. The content model makes those connections explicit. It doesn't guarantee a correct answer, but it gives the system a clearer basis for finding the information and showing where it came from.\n\n## Questions people ask\n\n**How is a content model different from a sitemap?**\n\nA sitemap shows pages and their relationship to one another. A content model describes what each kind of content is made of.\n\n**Is a content model the same as a CMS setup?**\n\nNot quite. A [CMS](https://contentlexicon.com/cms) implementation may express a content model, but the model comes first. It's the editorial and structural thinking that the CMS should follow.\n\n**Do small organisations need one?**\n\nOften yes, though it may be light-touch. If you publish the same kind of thing more than once, you already have an implied model. Writing it down makes it usable.\n\n**Does this only matter for structured content systems?**\n\nNo. That's where it shows most, but even a team working in shared documents benefits from agreeing what a guidance page, case study or event listing should contain.\n\n## Learn more\n\n- [Karen McGrane on Content Management](https://theinformed.life/2024/10/20/episode-151-karen-mcgrane/) – Karen McGrane considers what happens when content is separated from its containers and remixed by AI, and why structure and editorial intent matter.\n- [Content modeling basics](https://www.contentful.com/help/content-models/content-modelling-basics/) – Contentful's documentation explains content types, fields, validation and relationships between records." }, { "term": "Content operating model", "alsoCalled": [], "slug": "content-operating-model", "definition": "The whole system of people, processes, tools, rules and foundations that makes content work.", "category": "Foundations", "related": [ "backend", "business-case", "content-governance", "content-maturity", "content-operations", "content-strategy", "devops", "hub-and-spoke", "kpis", "systems-thinking", "ubiquitous-language", "workflow" ], "updated": "2026-10-09T09:37:00.000Z", "url": "https://contentlexicon.com/content-operating-model", "markdown": "https://contentlexicon.com/content-operating-model.md", "credit": "Content operating model, from the Content Lexicon by Contentious (https://contentlexicon.com/content-operating-model), CC BY 4.0", "body": "A content operating model is the whole arrangement behind an organisation's content: the decisions it has made about content, and how those decisions are put into practice every day.\n\nThe decisions are its [content strategy](https://contentlexicon.com/content-strategy): why content exists, who it's for, what it covers and where it goes, settled once so they don't have to be reargued on every project. Putting them into practice means the everyday choices made within them – which story leads the homepage today, who signs off this draft – and the people, processes and tools that make those choices possible. The model covers both, across every team that creates content. It isn't a single document or [workflow](https://contentlexicon.com/workflow). It's the working arrangement behind the content estate.\n\n## Why it matters for content\n\nA content operating model matters because content quality is rarely the result of individual talent alone. It depends on structures, decisions, methods and patterns that either support good work or make it difficult. The operating model helps an organisation move from individual effort and accidental consistency to a repeatable way of producing and maintaining good content.\n\nWhen a product change affects help articles, sales material and onboarding, the model should explain how those updates connect and who can decide what. The same arrangement needs to work after the launch, not only during a clean-up project.\n\n## How AI changes the picture\n\nAn operating model needs to make room for human and automated work together. A specialist might approve a new explanation, a system might identify the affected formats and an editor might review proposed adaptations. Decide who sets the brief, which work can happen automatically and where a question returns to a person.\n\nThat arrangement can free people to do work previously squeezed out, such as testing whether readers understand an unfamiliar service. It also needs support and maintenance: shared sources, suitable access, training and a way to investigate failures. Governance is one part of the model, not the whole account of collaboration.\n\n## Questions people ask\n\n**Is a content operating model just another name for content strategy?**\n\nThe strategy is part of it: the decisions, settled once, that give the model its direction and priorities. The operating model also covers the everyday choices people and systems make within those decisions. A clear strategy can still fail if nobody has the time or authority to apply it.\n\n**Is this only for large organisations?**\n\nSmaller organisations may have lighter operating models, but they still need one.\n\n**How is this different from workflow?**\n\nWorkflow is one part of an operating model. The operating model is broader: it also includes ownership, governance, tools, structures and the organisation's purpose and knowledge.\n\n**How is this different from content operations?**\n\n[Content operations](https://contentlexicon.com/content-operations) is the day-to-day running of content work: requests, briefs, production, approval, publishing and maintenance. The operating model includes that, the strategy it works within and the wider arrangement around both – who decides, the tools, the organisation's purpose and knowledge, and the ways different teams work together on content.\n\n**Does AI mean we need a completely new operating model?**\n\nUsually not completely new, but often much more explicit. AI exposes weaknesses and adds new [components](https://contentlexicon.com/component) that need to be designed into the system.\n\n## Learn more\n\n- [The new content operating model](https://www.optimizely.com/field-notes/guides/the-new-content-operating-model) – Optimizely’s vendor report on AI-era architecture, feedback and content operations, with product and GEO claims.\n- [The content operating model in the AI age](https://contentious.ltd/resources/the-content-operating-model-in-the-ai-age) – the Contentious model connecting organisational knowledge, content machinery and human–AI collaboration." }, { "term": "Content operations", "alsoCalled": [ "ContentOps" ], "slug": "content-operations", "definition": "The day-to-day running of content work: how it is planned, produced, approved, published and maintained.", "category": "Operations", "related": [ "agile", "author-experience", "content-governance", "content-lifecycle", "content-maintenance", "content-operating-model", "content-operations-manager", "devops", "workflow" ], "updated": "2026-10-11T06:24:00.000Z", "url": "https://contentlexicon.com/content-operations", "markdown": "https://contentlexicon.com/content-operations.md", "credit": "Content operations, from the Content Lexicon by Contentious (https://contentlexicon.com/content-operations), CC BY 4.0", "body": "Content operations covers the arrangements that turn intentions into published, maintained content: how requests arrive and become briefs, how work is scheduled, drafted, reviewed and approved, which tools people use and whose time it takes, along with the routines that keep live material accurate. It's often shortened to ContentOps, by analogy with [DevOps](https://contentlexicon.com/devops). The term earns its keep wherever content is made by many people.\n\nIt sits between strategy and the content itself. Strategy decides what should exist and why; operations is how that becomes ordinary practice. Both belong to a [content operating model](https://contentlexicon.com/content-operating-model): strategy sets the rules once, and content operations makes the calls against them, week after week. The model also takes in the wider arrangement – who decides, where teams sit, what's shared – and the ways teams across the organisation collaborate. [Workflow](https://contentlexicon.com/workflow), governance and the [content lifecycle](https://contentlexicon.com/content-lifecycle) are parts of content operations rather than alternatives to it.\n\n## Why it matters for content\n\nMany content problems that look editorial turn out to be operational. A page goes out with last year's eligibility rules because the policy change never reached the [writer](https://contentlexicon.com/writer), or reached them as an attachment in someone's inbox. Requests arrive by [email](https://contentlexicon.com/email), chat and corridor conversation, so nobody can see the whole queue. A launch depends on one person who knows which forms need updating. When the Government Digital Service examined how [GOV.UK](https://www.gov.uk) content was produced in 2016, it found more than 300,000 items, 73% of them viewed by fewer than ten people a month, and traced much of the problem to how the work was organised: departments publishing in [silos](https://contentlexicon.com/silo), [content designers](https://contentlexicon.com/content-designer) stuck as intermediaries and content treated separately from the services it described.\n\nOperations also decides what an organisation can sustain. Every new format, channel or [microsite](https://contentlexicon.com/microsite) adds work that continues after launch: reviews, corrections, translations, migrations. When that work is visible, a team can plan for it, retire something or decline the next request with evidence. When it isn't, the content estate grows faster than the people looking after it, and quality slides without anyone deciding it should. The same goes for the experience of the people doing the work: a [CMS](https://contentlexicon.com/cms) that fights its users or an approval route with six stops becomes a quality problem in its own right.\n\n## How AI changes the picture\n\nCheaper drafting moves the constraint. When a first version takes minutes, the scarce resources become good briefs, reliable sources, review capacity and the time to keep everything current. Operations is where an organisation decides how to spend the hours AI frees: more research and testing, more maintenance, or more output for the same number of reviewers, which tends to end badly.\n\nAI also gives content operations new things to run. Approved source collections, instructions and [prompts](https://contentlexicon.com/prompt), evaluation examples, tool permissions and model versions all shape what gets published, and each needs an owner, a review routine and a record of changes, like any other content. A charity adapting one approved explainer into a newsletter item, social posts and a briefing for local groups has multiplied the places a correction must reach; operations decides which version counts as the source and what happens to the derivatives when it changes.\n\nThe opportunity is in the coordination work itself, which has always eaten a surprising share of [content teams](https://contentlexicon.com/content-team)' weeks. Assistants can sort incoming requests, find the pages a policy change affects, draft change summaries for owners and flag missing [metadata](https://contentlexicon.com/metadata) before publication. Those are configured uses rather than defaults, and each depends on the team having agreed what should happen next: who gets the summary, who decides and by when.\n\n## Questions people ask\n\n**When does an organisation need to think about content operations?**\n\nUsually before it notices. The signs are familiar: more than a handful of people publishing, content that outlives the people who made it, requests arriving through several channels at once and launches that depend on late nights. None of these calls for a programme of work, but each calls for someone to look at how the work runs.\n\n**Who looks after content operations?**\n\nIn a large organisation, often a [content operations manager](https://contentlexicon.com/content-operations-manager) or a small team that owns the tools, workflows and standards everyone else works within. In a smaller one it's usually part of a [head of content](https://contentlexicon.com/head-of-content)'s job. Either way, the people doing the work need a say, because they know where it sticks.\n\n**Is ContentOps DevOps for content?**\n\nIt uses some of the same ideas: repeatable processes, automation where work is predictable, versioned and structured material, and steady improvement rather than occasional rescue projects. The analogy is closest in documentation teams that manage content like code. Editorial judgement and approval don't map neatly onto a build pipeline, though, so the comparison works better as inspiration than as a blueprint.\n\n**How can you tell whether content operations are working?**\n\nOutput alone says little. More useful signs are the time from request to publication, the share of live content with a named owner and a [review date](https://contentlexicon.com/review-date), how often published content needs correcting and how much of the team's week goes on chasing and coordinating rather than on the content.\n\n## Learn more\n\n- [Content Operations from Start to Scale](https://doi.org/10.21061/content_operations_evia) – an open-access collection edited by Carlos Evia (Virginia Tech Publishing), with practitioners including Sarah O'Keefe and Rahel Anne Bailie explaining ContentOps and how it grows across an organisation.\n- [GOV.UK's content operating model: what's next after discovery](https://insidegovuk.blog.gov.uk/2016/12/05/gov-uks-content-operating-model-whats-next-after-discovery) – Trisha Doyle's 2016 account of what GDS found when it examined how government departments produced their content for the site, and why it proposed redesigning the operating model around publishers' needs.\n- [Content Maturity framework](https://contentmaturity.com/framework) – the Contentious model in which operations is one of five areas, assessed through collaboration, workflow, maintenance, skills and training." }, { "term": "Content operations manager", "alsoCalled": [], "slug": "content-operations-manager", "definition": "A role responsible for running and improving how an organisation's content work happens: its workflows, tools, standards and routines.", "category": "Roles", "related": [ "content-governance", "content-manager", "content-operating-model", "content-operations", "content-strategist", "content-team", "head-of-content", "workflow" ], "updated": "2026-10-08T20:23:00.000Z", "url": "https://contentlexicon.com/content-operations-manager", "markdown": "https://contentlexicon.com/content-operations-manager.md", "credit": "Content operations manager, from the Content Lexicon by Contentious (https://contentlexicon.com/content-operations-manager), CC BY 4.0", "body": "A [content operations](https://contentlexicon.com/content-operations) manager looks after an organisation's content operations: the [workflows](https://contentlexicon.com/workflow), tools, standards and routines that everyone else's content work depends on. They concentrate on the conditions around content rather than producing every piece, and much of what they achieve is invisible when it's going well.\n\nThe title is newer and less settled than most content roles. Large organisations sometimes have a dedicated post or a small team. In smaller ones the same responsibilities sometimes sit with a [head of content](https://contentlexicon.com/head-of-content) or a [content strategist](https://contentlexicon.com/content-strategist), whether or not anyone calls them operations.\n\n## Why an organisation needs a content operations manager\n\nOperational work is easy to leave to everyone, which in practice means nobody. Each team fixes its own corner: one keeps a spreadsheet of [review dates](https://contentlexicon.com/review-date), another builds a workaround in the [CMS](https://contentlexicon.com/cms), a third stops using the approval route because it takes too long. A content operations manager gives that work an owner who can see across teams, find where work stalls, duplicates or depends on one person's memory, and change the arrangement rather than patching it again.\n\nThe role works only with matching authority. A manager who can recommend a simpler approval route but not change it becomes a well-informed complainer. The remit commonly includes the CMS and other content tools, workflow and approval design, the [content calendar](https://contentlexicon.com/content-calendar), training for publishers and the rules for using AI in content work. It can overlap with IT, digital, editorial and governance roles, and gaps between them are where problems settle. Success shows in fewer errors, clearer responsibility and faster safe changes rather than in more output.\n\n## How AI changes the role\n\nThe job gains a technical layer. Someone has to decide which content tasks suit AI, which sources assistants may draw on, who can change their instructions and how results are checked. The content operations manager is often best placed to coordinate those decisions with technology and risk colleagues, without inheriting responsibilities that belong to them. [Prompt](https://contentlexicon.com/prompt) libraries, automated checks and monitoring may join the CMS on the list of things the role maintains. Where an organisation also has a [content manager](https://contentlexicon.com/content-manager), the operations manager builds and runs those arrangements while the content manager decides what assistants should produce and to what standard.\n\nIt also needs some resistance to automation for its own sake. Automating a reminder won't clear an approval queue if no reviewer has the time or authority to approve; the arrangement needs fixing first. Part of the work becomes making sure people can see, question and correct what automated steps do, so a wrong change can be traced and reversed.\n\n## Questions people ask\n\n**Is this a project manager?**\n\nNot exactly. A project such as a CMS migration may fall to the role, but its main work carries on after any project ends: keeping the arrangements running and improving them.\n\n**Is it a senior role?**\n\nIt needs enough influence to change processes across teams, which usually means backing from whoever leads content or digital. Scope varies by organisation: some posts lean towards platforms and tooling, others towards people and process.\n\n**Does the role write content?**\n\nNot usually. Its distinctive responsibility is improving how content work happens, and any writing time tends to get squeezed by it.\n\n**What skills does the role need?**\n\nA combination that's rarely taught together: enough editorial judgement to know what good content needs, enough technical fluency to configure tools and work with developers, and the patience to change how people work without issuing decrees. Editorial, [content design](https://contentlexicon.com/content-design) and digital project backgrounds all lead into it.\n\n## Learn more\n\n- [How to measure content operations maturity](https://uxcontent.com/how-to-measure-content-operations-maturity) – Rob Mills compares three content operations maturity models and argues for using them as conversation starters rather than prescriptions, which is useful for anyone setting an operations role's priorities.\n- [Who does what?](https://contentious.ltd/resources/who-does-what#content-operations-manager) – the Contentious guide to content roles, including where operations responsibilities sit when a team is too small for a dedicated post." }, { "term": "Content owner", "alsoCalled": [], "slug": "content-owner", "definition": "The named person responsible for a piece of content being accurate, current and still worth having.", "category": "Governance", "related": [ "content-calendar", "content-debt", "content-governance", "content-inventory", "content-maintenance", "content-plan", "content-team", "decentralised-governance", "merge-conflict", "raci", "review-date", "sign-off", "stakeholder-mapping", "technical-debt" ], "updated": "2026-10-06T15:28:00.000Z", "url": "https://contentlexicon.com/content-owner", "markdown": "https://contentlexicon.com/content-owner.md", "credit": "Content owner, from the Content Lexicon by Contentious (https://contentlexicon.com/content-owner), CC BY 4.0", "body": "A content owner may not have written the content, and they may not edit it day to day or know every detail. What makes them the owner is that they answer for it: they decide when it changes, respond when it is challenged and know when it should be reviewed or retired.\n\nOwnership is often the missing ingredient in otherwise sensible content systems. Dates, [workflows](https://contentlexicon.com/workflow) and standards are much less useful if nobody's clearly accountable for acting on them.\n\nPutting a name in a Notion field is only the start. The owner needs enough authority to make a decision, access to the people who know the subject and time to act when a review is due. Otherwise the organisation has assigned responsibility without making ownership possible.\n\nOwnership also needs a handover. When an owner changes role or leaves, their content should move to another named person rather than reverting to “the team”. Backup owners and review reminders help, but the important part is a deliberate transfer: what the content is for, which decisions are pending and where the evidence behind it lives.\n\n## Why it matters for content\n\nContent ownership makes governance real. It reduces ambiguity about who should review something, approve an update, respond to inconsistency or decide whether material is obsolete. It also helps prevent the slow drift into orphaned content: pages that still exist but no longer belong to anyone in practice.\n\nOwnership counts for most with high-stakes, long-lived or frequently updated content, and anywhere several people might assume someone else is looking after it. Which is to say: a lot of content.\n\n## How AI changes the picture\n\nWhen an assistant rewrites a page, someone still has to answer for what it now says. Content is drafted, summarised, transformed and reused in more ways than before, often by tools nobody thinks of as authors. If a chatbot quotes last year's fees from a page nobody owns, there's nobody to notice and nobody to fix it.\n\nThe list of things that need owners grows too: [prompts](https://contentlexicon.com/prompt), retrieval sources, style guidance and other machine inputs. If those assets shape outputs, someone should be accountable for their quality and maintenance.\n\n## Questions people ask\n\n**Is the content owner always the writer?**\n\nOften not. The [writer](https://contentlexicon.com/writer) creates or edits the content; the owner is the person accountable for its continuing fitness.\n\n**Can a team be the owner?**\n\nA team can share the work, but a named person is clearer. “The digital team” can't pick up the phone when a page is challenged.\n\n**Does every page need an owner?**\n\nNot with equal formality. A cookie notice and a fees page both need an owner; only one needs a quarterly review.\n\n**Is ownership the same as approval?**\n\nThey overlap. Owners may approve changes, but ownership is broader: it includes stewardship over time.\n\n## Learn more\n\n- [Owning and Maintaining Content](https://www.ncl.ac.uk/brand-hub/content-creators-guide/reviewing/content-ownership) – Newcastle University's guidance for its own content owners: named owners with access, backup owners, review intervals and retirement checks, including checking for other places that link to retired content.\n- [Bus factor](https://en.wikipedia.org/wiki/Bus_factor) – the software term for the risk that knowledge and responsibility live in one person. Content ownership meets the same exposure, which is what the handover paragraph protects against." }, { "term": "Content pattern", "alsoCalled": [], "slug": "content-pattern", "definition": "A proven, recognisable approach to something content needs to do again and again, adapted each time it is used.", "category": "Structure and systems", "related": [ "component", "content-design", "content-type", "design-system" ], "updated": "2026-10-09T10:36:00.000Z", "url": "https://contentlexicon.com/content-pattern", "markdown": "https://contentlexicon.com/content-pattern.md", "credit": "Content pattern, from the Content Lexicon by Contentious (https://contentlexicon.com/content-pattern), CC BY 4.0", "body": "A content pattern describes a situation that keeps coming up, the approach that has worked for it and the reasoning behind that approach. It might cover how to tell someone a payment has failed, ask for a date of birth, introduce a case study or explain who's eligible for a grant. Rather than supplying fixed words, it sets out the shape of a good response and what can change, so each use can fit its audience and setting.\n\nThe idea comes from architecture. In _A Pattern Language_ (1977), Christopher Alexander and his colleagues described more than 250 recurring arrangements, from a seat by a window to the shape of a neighbourhood, each written so it could be used in countless buildings and look different in every one. Software developers borrowed the idea, most famously in the 1994 book _Design Patterns_, and [design systems](https://contentlexicon.com/design-system) later brought it into digital services, where a single pattern often draws several [components](https://contentlexicon.com/component) together.\n\n## Why it matters for content\n\nPatterns keep good decisions in circulation. Research into how people respond to error messages, a legal review of consent wording or a hard-won agreement about how to describe eligibility would otherwise live in one team's memory. Written up as a pattern, with the evidence alongside the approach, it gives the next [writer](https://contentlexicon.com/writer) a reliable starting point and the means to tell when their situation needs something else. In a booking service, an error pattern could specify that the message names the missing information and explains how to correct it, while letting the words fit the task. Patterns should evolve when testing reveals a better approach. Blind reuse isn't consistency; it's habit with a template.\n\nReaders benefit too, usually without noticing. Someone who has met the same check-your-answers page on several services knows what it's for and what to do next. A familiar structure for opening times or a product recall lets people find the detail they came for without first learning a new layout, so their attention goes on the content rather than on working out how the page behaves.\n\nAlexander insisted that patterns carry a view of what makes a place good for the people in it, and content patterns do the same whether or not anyone says so. The repertoire can also be turned against readers. Deceptive patterns such as trick wording, fake urgency and confirmshaming (the opt-out link that reads something like \"No thanks, I enjoy paying full price\") are recurring, recognisable and effective, which is why they spread. Many of them are content decisions. A pattern library therefore records, among other things, how an organisation intends to treat people.\n\n## How AI changes the picture\n\nLanguage models learn patterns of a different kind. They reproduce the shapes that appear most often in their training data, which isn't the same as reproducing the shapes that work. Ask for a cancellation page with no guidance and you'll get something close to the average cancellation page, nagging and all. A maintained pattern library gives a model better examples to follow than the median of the web.\n\nAn assistant with access to that library can suggest a relevant pattern during drafting and apply its structure, though whether it has chosen the right one still needs checking. Much depends on how the patterns are written: unless guidance distinguishes fixed elements from adaptable ones, a model may copy wording that doesn't fit the audience or task. Alexander's later thinking anticipates this. In a 1996 keynote to software developers, he described his newer pattern languages as sequences of instructions, closer to code, which respond to their context and produce a different result each time. A pattern that states its purpose, its rules and its permitted variations is well on the way to being an instruction an AI system can follow.\n\nAI can also analyse existing content and support enquiries to propose candidate patterns. Those proposals need testing against real evidence, because frequency doesn't prove that an approach is good. The most common pattern in a set of sign-up journeys may well be a deceptive one.\n\n## Questions people ask\n\n**How is a pattern different from a component?**\n\nA component is a part you reuse as it stands: the same alert, the same disclaimer, maintained in one place. A pattern is an approach you reuse and adapt. Telling someone their application has been turned down might combine a heading, a short explanation, a list of reasons and a link to appeal, with guidance on [tone](https://contentlexicon.com/tone), order and what to say when the reasons can't be shared. The components stay the same; the pattern contributes the judgement about how to use them.\n\n**Is a pattern the same as a template?**\n\nA template gives you a structure to fill in. A pattern explains why that structure works and when to depart from it. One pattern can sit behind several templates, and a template cut off from its pattern tends to be filled in without much thought.\n\n**Must everyone follow a pattern to the letter?**\n\nState what's required and what can change. Exceptions should be deliberate and informative, and one that keeps recurring is often a sign that the pattern needs revising.\n\n**What's an anti-pattern?**\n\nA common response to a recurring situation that looks sensible but reliably makes things worse, such as a vague \"Something went wrong\" message or an FAQ page that collects every question nobody could fit elsewhere. Naming anti-patterns is as useful as documenting good ones, because it explains what to stop doing and why.\n\n## Learn more\n\n- [The origins of pattern theory](https://www.patternlanguage.com/archive/ieee.html) – Christopher Alexander's 1996 keynote to software developers, reflecting on how they had adopted his ideas and urging them to take on the moral purpose behind patterns: making things that are good for the people who live with them.\n- [Integration Patterns Book](https://digital.nhs.uk/developer/architecture/integration-patterns-book) – NHS England's guide for architects and engineers on moving health and social care data between systems. It isn't about content, but it shows a pattern library at its most useful: explaining when to choose each pattern and naming the anti-patterns to avoid, from bulk transfers to screen scraping.\n- [Types of deceptive pattern](https://www.deceptive.design/types) – the catalogue Harry Brignull started in 2010, from trick wording to confirmshaming, now mapped to the laws and enforcement actions each has attracted." }, { "term": "Content plan", "alsoCalled": [], "slug": "content-plan", "definition": "A working plan that shows what needs to happen to each piece of content, who owns it and what comes next.", "category": "Operations", "related": [ "content-audit", "content-brief", "content-calendar", "content-governance", "content-inventory", "content-owner", "content-strategy", "workflow" ], "updated": "2026-10-07T13:53:00.000Z", "url": "https://contentlexicon.com/content-plan", "markdown": "https://contentlexicon.com/content-plan.md", "credit": "Content plan, from the Content Lexicon by Contentious (https://contentlexicon.com/content-plan), CC BY 4.0", "body": "Where a [content audit](https://contentlexicon.com/content-audit) judges what you have, a content plan tracks each piece through the creation or improvement work that follows.\n\nA good content plan usually sits between strategy and production. It might grow out of a content audit, a [website](https://contentlexicon.com/website) restructure, a product launch, a migration or a campaign. It can be a document, but it can also be a workspace where the content work gets done: a shared place for pages, messages, assets, owners, decisions, deadlines, dependencies and project management. Its job is to help people understand the shape of the work, make sensible decisions and keep content moving without losing quality on the way.\n\n## Why it matters for content\n\nContent work often becomes vague when it's described only as “copy”, “assets” or “pages”. A content plan gives that work edges. It shows what's missing, what needs attention and who needs to do what before anything can be published.\n\nFor [content teams](https://contentlexicon.com/content-team), digital teams and clients, the plan is where strategy becomes accountable. If the strategy says a website should be clearer, more useful or better structured, the content plan shows which pages need rewriting, which messages need proof, which decisions are still unresolved and which parts of the content estate should be sent off into retirement. It also helps protect quality, because content that stays useful, readable and up to date rarely happens unless someone plans for it.\n\nA content plan becomes especially important when several people are involved. [Writers](https://contentlexicon.com/writer), subject experts, designers, developers, approvers and project leads all need a shared view of what's happening. Without that shared view, teams tend to rely on memory, meetings and a frantic search through old [email](https://contentlexicon.com/email) threads the week before launch. A content plan saves everyone from pretending that “nearly there” is a project status.\n\n## How AI changes the picture\n\nA shared content plan gives an assistant something concrete to compare when circumstances change. If a product launch moves, it can flag the related help pages, emails and approval tasks, show which deadlines depend on the launch date and suggest an updated sequence for their owners to review.\n\nIf AI is being used in content production, the content plan should make the human decisions visible. Which pages can be drafted from existing source material, and which need [subject-matter expert](https://contentlexicon.com/subject-matter-expert) input first? Which have legal, policy, [accessibility](https://contentlexicon.com/accessibility) or brand implications? And which content is canonical enough for AI to use, and which is still a pile of unverified notes? The plan helps distinguish between work that AI can support and work that still needs judgement, evidence and organisational agreement.\n\nKeeping the plan current is another job AI can help with. It can summarise audit findings, suggest draft statuses, cluster related pages, identify missing [metadata](https://contentlexicon.com/metadata) or turn messy notes into first-pass tasks. It can propose priorities against agreed criteria, but people approve the commitments and messages. It shouldn't invent substance where the source material is thin. A content plan is one of the places where [content governance](https://contentlexicon.com/content-governance) becomes practical: it shows what machines may help with, where people need to review and what standards the work must meet.\n\n## Questions people ask\n\n**How does a content plan differ from an editorial calendar?**\n\nAn [editorial calendar](https://contentlexicon.com/content-calendar) focuses on what will be published when, usually for channels such as blogs, newsletters or [social media](https://contentlexicon.com/social-media). A content plan is broader. It can include existing pages to improve, content to merge or remove, source material to find, approvals to secure, migration notes, ownership and decisions that need to happen before publishing is even possible.\n\n**Do we need a content plan for a small project?**\n\nUsually, yes, but keep it proportionate. A small project may only need a simple table with the content items, owners, status and next action. That's enough to stop important decisions living in someone’s head or inbox.\n\n**What should a content plan include?**\n\nStart with the content item, its purpose, owner, status, next action and deadline. Add source material, approvals, dependencies or migration notes when the project needs them. Drop fields that nobody uses.\n\n**Can AI write the content from the plan?**\n\nSometimes it can help, especially if the plan points to good source material, clear [user needs](https://contentlexicon.com/user-needs), approved messages and useful examples. But the plan itself should make clear where human judgement is needed. AI can draft from evidence; it can't responsibly fill organisational gaps, resolve political uncertainty or approve claims just because everyone would quite like the page to be finished by Thursday.\n\n## Learn more\n\n- [4-Part Guide To Crafting a Winning Content Plan](https://contentmarketinginstitute.com/content-marketing-strategy/4-part-guide-to-crafting-a-winning-content-plan) – governance, processes, team resources and creation/delivery as the practical foundations of a content plan.\n- [The Discipline of Content Strategy](https://alistapart.com/article/thedisciplineofcontentstrategy/) – Kristina Halvorson's 2008 argument for treating creation, publication and governance as one discipline." }, { "term": "Content policy", "alsoCalled": [], "slug": "content-policy", "definition": "The organisation's rules for what may be published, where, by whom and under what conditions.", "category": "Governance", "related": [ "ai-use-policy", "content-governance", "content-strategy", "style-guide" ], "updated": "2026-10-14T08:42:00.000Z", "url": "https://contentlexicon.com/content-policy", "markdown": "https://contentlexicon.com/content-policy.md", "credit": "Content policy, from the Content Lexicon by Contentious (https://contentlexicon.com/content-policy), CC BY 4.0", "body": "A content policy is the set of rules an organisation applies to what it publishes: what may go out, where, who can publish or approve it and what happens when something goes wrong. It deals in obligations and permissions, such as when legal review is required, how personal data is handled, how corrections are shown and when content must be removed.\n\nIt's easily confused with two neighbouring concepts. [Content strategy](https://contentlexicon.com/content-strategy) sets direction: what content should achieve, for whom and where effort should go. A [style guide](https://contentlexicon.com/style-guide) covers how to write and present content: spelling, [voice](https://contentlexicon.com/voice) and [tone](https://contentlexicon.com/tone), formatting and terminology. A content policy sits above the style guide and alongside the strategy. It says what's allowed and who decides. A useful test is consequence. Breaking a style rule produces an inconsistent page. Breaking a policy can expose personal information, make a promise the organisation can't keep or breach the law.\n\nThe term has a second meaning on platforms that host other people's material, where a content policy sets out what users may post. This is essentially the same rulebook, applied to someone else's content.\n\n## Why it matters for content\n\nPublishing creates obligations. A page may make a promise, expose personal information, misstate a service or stay online long after its facts have expired. A policy helps teams handle these risks consistently instead of rediscovering the organisation's position against every awkward deadline.\n\nUseful policies turn principles into decisions. They say who may publish a video, when legal review is required, how a correction is shown and what happens to content whose owner has left. A team might allow finance to update an approved rate but require legal approval before an eligibility rule changes. That gives people a decision they can apply, rather than asking them to consult \"appropriate colleagues\".\n\nA typical policy covers scope (which sites, channels and [content types](https://contentlexicon.com/content-type) it applies to), roles and publishing permissions, approval routes by risk, legal and [accessibility](https://contentlexicon.com/accessibility) obligations, personal data and consent, corrections and complaints, review and removal and how to appeal a decision. Each rule needs an owner. A policy nobody enforces turns into a style guide with a sterner title.\n\nKeeping policy short and stable, with detailed procedures alongside it, makes the rules easier to understand and lets procedures change without rewriting the policy. Wikipedia makes a similar split: policies set the standards everyone should follow, guidelines give advice for particular situations and essays are individual opinion. Readers can then tell a requirement from a recommendation.\n\n## How AI changes the picture\n\nAI makes it easier to produce and alter content outside established publishing teams. A policy now needs to cover machine-made drafts, synthetic media, automated personalisation and content assembled dynamically from several sources.\n\nIt should also separate creation from publication. A tool may be allowed to draft a summary while the policy still forbids it from publishing without review. The policy needs matching tool permissions and approval gates, because a sentence forbidding publication isn't an enforced boundary.\n\nPolicies written as clear, structured rules can also be given to AI tools as instructions, so an AI agent flags a draft that needs legal review before anyone asks. The rules still need a human owner.\n\n## Questions people ask\n\n**How is a content policy different from a style guide?**\n\nA style guide covers language, presentation and conventions. A policy covers permission, restriction and responsibility. One says how to write a case study. The other says whether consent is needed before publishing it.\n\n**Is this the same as content strategy?**\n\nNo. Strategy explains what content should achieve and where effort should go. Policy defines the rules within which that work happens.\n\n**How long should it be?**\n\nLong enough to settle recurring questions, short enough to use. Put detailed procedures beside the policy rather than burying the rule in forty pages of [workflow](https://contentlexicon.com/workflow).\n\n**Who owns it?**\n\nUsually a senior content or digital owner, with input from legal, accessibility, information governance and security. Name an accountable role and a route for resolving disagreements, even if several functions approve parts of the policy.\n\n## Learn more\n\n- [Content governance policy](https://thethread.renfrewshire.gov.uk/content-guidelines/content-governance-policy/) – a working council policy that connects publishing activities to named roles and permissions.\n- [Web Content Governance Policy](https://www.gre.ac.uk/__data/assets/pdf_file/0020/134183/webcontentgovernancepolicy.pdf) – the University of Greenwich's 2016 policy, more formal than the council example, with approver training, override powers and a route for appealing decisions.\n- [Wikipedia: Policies and guidelines](https://en.wikipedia.org/wiki/Wikipedia:Policies_and_guidelines) – how a volunteer community separates binding policies from guidelines and essays, and how it guards against instruction creep." }, { "term": "Content principles", "alsoCalled": [], "slug": "content-principles", "definition": "A short set of shared commitments that guide content decisions.", "category": "Foundations", "related": [ "content-standards", "content-strategy", "core-content-strategy", "style-guide", "voice" ], "updated": "2026-10-11T06:24:00.000Z", "url": "https://contentlexicon.com/content-principles", "markdown": "https://contentlexicon.com/content-principles.md", "credit": "Content principles, from the Content Lexicon by Contentious (https://contentlexicon.com/content-principles), CC BY 4.0", "body": "Content principles are a short set of statements describing what an organisation believes good content should do, used to guide decisions when several reasonable choices compete. They sit between strategy and day-to-day craft: more specific than a mission, less detailed than a [style guide](https://contentlexicon.com/style-guide). “Start with what the reader needs to do” is a principle; “use [sentence case](https://contentlexicon.com/sentence-case) for headings” is a style rule.\n\nThey borrow from a longer tradition of design principles. Dieter Rams's ten principles for good design, set out while he led design at Braun, showed how a few memorable statements could guide a whole body of work. The UK Government Digital Service's design principles, published in 2012 and opening with “Start with [user needs](https://contentlexicon.com/user-needs)”, made the form familiar in digital teams, and content principles apply the same idea to words, structure and publishing.\n\n## Why it matters for content\n\nPrinciples connect strategy with daily work. They can guide briefs, critiques, governance and prioritisation without prescribing one answer for every situation. When teams disagree, a principle lets them test the decision against an agreed commitment rather than preference or seniority, which matters most when the person asking for another homepage banner outranks the person arguing against it.\n\nGood principles are memorable, specific and usable. “Include the right people for the right reasons” changes how a review process works. “Create great content” changes nothing, because nobody set out to create dreadful content. One test is whether a sensible organisation could choose the opposite. “Maintain what exists before publishing more” is a real choice that some teams would reject, so it can settle arguments. Each principle needs a short explanation and an example of what it changes in practice.\n\nUseful principles reflect real tensions: user needs over organisational structure, sustainable maintenance over constant publication, evidence over assumption. Too many become wallpaper. A short set that leaders and practitioners cite is worth more than an inspiring page nobody can recall, and principles earn their keep when someone uses them to say no.\n\n## How AI changes the picture\n\nPrinciples can guide AI-assisted work by shaping [prompts](https://contentlexicon.com/prompt), review criteria and system behaviour. They express priorities that style rules don't capture: when to refer someone to a person, how much certainty to claim, what to leave out.\n\nAbstract statements need concrete examples before a model can apply them. AI may read “be human” as chatty filler or “be bold” as permission for unsupported certainty. Pairing each principle with examples, limits and source requirements makes it operational, and the exercise often improves the principles for human readers too.\n\nAI can check drafts against principles and flag likely conflicts, but people still decide how commitments apply when they pull in different directions or when the consequences are serious. Principles also bear on the decision to use AI in the first place. An organisation committed to evidence over assumption needs a view on generated content that nobody has checked.\n\n## Questions people ask\n\n**How are principles different from a style guide?**\n\nPrinciples guide decisions and priorities. A style guide gives detailed conventions for language and presentation, and its rules should be traceable to the principles. If a principle says to respect people's time, the style guide might set out how to front-load the answer.\n\n**How many should there be?**\n\nThree to five is a sensible starting point rather than a rule. Test them against real decisions: if two principles always give the same answer, combine them; if a recurring tension has nothing to settle it, add or revise one.\n\n**Who creates them?**\n\nA representative group, informed by audience evidence, organisational purpose and the people who make and maintain content. Principles written by leadership alone tend to describe aspirations, while those written only by practitioners may lack the authority to settle arguments. Testing drafts against recent disputes exposes vague wording quickly.\n\n**Can principles contradict each other?**\n\nThey often pull in different directions, which is part of their value. Rank them, or explain how to balance them, rather than pretending every value wins equally.\n\n## Learn more\n\n- [MnDOT content principles](https://www.dot.state.mn.us/designsystem/contentprinciples.html) – a concrete set covering refinement, deletion, findability and evidence rather than vague aspirations.\n- [Five principles of good content design](https://uxdesign.cc/principles-of-good-content-design-4c55622c9919) – a public-sector and non-profit perspective on findable, clear, connected, human and helpful content; one usable framework, not a universal set.\n- [Design Principles](https://principles.design/) – a free library of more than 1,600 real-world design principles from around 200 organisations and individuals, useful for comparing how others phrase and explain theirs.\n- [The content manifesto](https://contentious.ltd/the-content-manifesto) – the principles behind Contentious’ broad view of content as relationship, infrastructure and a force for change." }, { "term": "Content standards", "alsoCalled": [], "slug": "content-standards", "definition": "The requirements used to judge content quality before publication and throughout its life.", "category": "Governance", "related": [ "accessibility", "content-health-check", "content-principles", "style-guide" ], "updated": "2026-10-11T06:26:00.000Z", "url": "https://contentlexicon.com/content-standards", "markdown": "https://contentlexicon.com/content-standards.md", "credit": "Content standards, from the Content Lexicon by Contentious (https://contentlexicon.com/content-standards), CC BY 4.0", "body": "Content standards turn a general wish for \"good content\" into written criteria a reviewer can check, such as whether a page says who it's for, links every statistic to its source or gives every image a text alternative. They apply at the point of publication and for as long as the content stays live.\n\nStandards sit between two related things. [Content principles](https://contentlexicon.com/content-principles) express what an organisation believes about content and guide judgement where no rule applies. A [style guide](https://contentlexicon.com/style-guide) covers language and presentation: spelling, capitals, [voice](https://contentlexicon.com/voice) and [tone](https://contentlexicon.com/tone). Standards set the quality bar and the evidence used to assess.\n\nA usable standard has three parts: the criterion, the evidence that shows it's met and who checks it. \"Be accurate\" is an aspiration. \"Every statistic links to its source, checked by the editor before publication\" is a standard.\n\n## Why it matters for content\n\nWithout written standards, an assessment of quality depends on who happens to review the work. An experienced editor may catch an unexplained acronym, a missing source and an inaccessible heading structure. A busy subject expert may check only the fact they supplied. Standards give both people the same questions.\n\nThey're most useful when observable. \"Be audience-centred\" is a fine principle but a poor test. \"State the [user need](https://contentlexicon.com/user-needs), use descriptive headings and explain specialist terms\" gives a reviewer something to inspect. [Accessibility](https://contentlexicon.com/accessibility) shows how far this can go: [WCAG](https://contentlexicon.com/wcag) writes its requirements as testable success criteria, which is what makes an accessibility audit possible. Standards also make maintenance possible, because they apply after publication as well as at the gate.\n\nA common core works for most organisations, with extra requirements for particular risks and formats. A transaction page and a thought piece both need accuracy, but only one needs every step numbered.\n\n## How AI changes the picture\n\nAI-generated drafts need an explicit basis for evaluation. \"It reads well\" isn't enough when the copy may contain a confident factual error, omit a qualification or mimic the right voice while missing the reader's task. Written standards give editors, evaluation systems and automated checks a shared reference.\n\nSome criteria can be checked mechanically, including broken links, heading order and whether images have a text alternative or are marked as decorative. Those checks can't reliably judge whether a description is appropriate. A model can flag that a page makes a claim without a source; it can't always decide whether the claim is fair, proportionate and useful.\n\nStandards written as clear criteria also work as instructions. The same list an editor uses can go into a model's [prompt](https://contentlexicon.com/prompt) or an automated review, so generated drafts are held to the same bar as human ones from the start.\n\n## Questions people ask\n\n**What belongs in a content standard?**\n\nThe criterion, why it matters, what passing looks like and any exceptions. Examples help, especially where the judgement isn't binary, such as whether a summary is fair to the full report.\n\n**Should every content type use the same standards?**\n\nUse a common core, then add requirements for particular risks and formats. A form needs standards for labels and error messages that a blog post doesn't.\n\n**Can a score replace editorial review?**\n\nNo. Scores expose patterns and support decisions, but they don't make difficult editorial trade-offs disappear. A page can score well and still miss what readers came for.\n\n**How often should standards change?**\n\nReview them when audience needs, law, technology or publishing practice change. A standard that never changes may be enduring; it may also have been abandoned politely.\n\n## Learn more\n\n- [How to create and update content for GOV.UK](https://www.gov.uk/government/collections/content-and-publishing-guidance-for-government) – GOV.UK's collection of writing standards, content-type guidance and publishing instructions.\n- [Content Standards in Design Systems](https://www.nngroup.com/articles/content-design-systems/) – Nielsen Norman Group's article taking a broader view, with standards covering content structure, editorial process and style guidance, illustrated with examples from Mailchimp and Salesforce.\n- [Content Health Check](https://contenthealthcheck.com/) – the Contentious tool that assesses live content against a 15-criterion quality framework." }, { "term": "Content strategist", "alsoCalled": [], "slug": "content-strategist", "definition": "A role responsible for deciding what an organisation's content is for and designing how it is planned, structured, made and looked after.", "category": "Roles", "related": [ "content-designer", "content-manager", "content-operations-manager", "content-strategy", "content-team", "head-of-content" ], "updated": "2026-10-08T16:39:00.000Z", "url": "https://contentlexicon.com/content-strategist", "markdown": "https://contentlexicon.com/content-strategist.md", "credit": "Content strategist, from the Content Lexicon by Contentious (https://contentlexicon.com/content-strategist), CC BY 4.0", "body": "A content strategist sets the direction for an organisation's content and designs the arrangements that put it into practice: who the content serves, what it should and shouldn't cover, how it's structured, who makes it and how it's governed. The job usually involves research with audiences and colleagues, auditing what already exists and recording decisions in a form people will use. Strategists mostly make things happen through others, often with influence across many teams and line management of very few people.\n\nThe role takes many forms. Some strategists work across a whole organisation, others specialise in product content, [taxonomy](https://contentlexicon.com/taxonomy) and structure or marketing, and many work as consultants brought in for a redesign or a reset. Most are expert in part of the discipline while knowing enough about the rest to see how it fits together.\n\n## Why an organisation needs a content strategist\n\nContent decisions get made whether or not anyone is responsible for them. Without a strategist, they're made one request at a time by whichever team is publishing, and the result is an estate that reflects the organisation chart rather than the people it serves. A strategist makes those decisions deliberately and early: what to stop, what to merge, which audience comes first, what the [CMS](https://contentlexicon.com/cms) needs to support. They're often the person who can explain a content problem to a chief executive in terms of cost and risk, and to an editor in terms of what changes on Monday.\n\nThe need usually arrives with a moment of change, such as a redesign, a merger, a new service or a CMS migration. It also appears more slowly, when an organisation notices that its content has grown faster than its ability to look after it. The role needs access to senior decision-makers and permission to make choices some colleagues won't like. A strategist can't deliver the strategy alone either: a [content manager](https://contentlexicon.com/content-manager), editors or an operations lead have to turn it into everyday practice, and the strategist's influence fades if nobody owns that work once the document is signed off.\n\n## How AI changes the role\n\nMuch of the AI work that lands on [content teams](https://contentlexicon.com/content-team) is strategic before it's technical: deciding which tasks to hand to assistants, what the tools may draw on and what good output looks like. Strategists are increasingly the people asked to make those calls, often in organisations where AI arrived through a side door before anyone agreed what it was for. That adds new deliverables to the job. Alongside audits and frameworks, a strategist may now write guidelines for using AI in content work, specify the sources and standards an assistant works from and design evaluation criteria with technical colleagues.\n\nThe role also moves towards structure. Systems that retrieve and recombine content depend on it being well modelled, consistently tagged and clearly owned, so taxonomy and content modelling, once specialisms within the discipline, move closer to the centre of a strategist's job. A job description might now ask for experience of content modelling, [AI governance](https://contentlexicon.com/ai-governance) and working with engineers, alongside the familiar research, auditing and stakeholder skills. The need for someone with the standing to decide what an organisation's content is for doesn't go away; cheaper production makes that decision harder to avoid.\n\n## Questions people ask\n\n**How is a content strategist different from a content manager?**\n\nA strategist sets direction and designs the system; a manager runs it week by week. The strategist asks what the content should achieve and how it should be organised, while the manager makes sure it's planned, produced and kept in good order. Smaller organisations often combine the two, which works as long as daily demands don't crowd out the longer view.\n\n**Do we need a permanent strategist or a consultant?**\n\nEither can work. A consultant suits a defined moment, such as a redesign or a reset, and brings an outside view. A permanent strategist suits large or complex estates where decisions keep arising and someone needs to hold the direction over years. Many organisations use a consultant to set the strategy, then appoint or develop someone internally to keep it alive.\n\n**Where should a content strategist sit?**\n\nClose to the people who make organisation-wide decisions, whether in digital, communications, product or service leadership. In UK government, for example, the role is defined as an expert practitioner working alongside lead [content designers](https://contentlexicon.com/content-designer).\n\n**What makes a good content strategist?**\n\nThe ability to research, analyse and decide, combined with the patience to bring people along. Strong strategists are as comfortable with a spreadsheet as with a senior meeting, can explain structure to editors and editorial judgement to developers and know enough about each part of the discipline to see where a problem starts.\n\n## Learn more\n\n- [What Does a Content Strategist Do?](https://www.braintraffic.com/blog/what-does-a-content-strategist-do) – Kristina Halvorson's map of the activities that fall to content strategists, and her observation that every strategist specialises in a subset while needing to understand the rest.\n- [Content strategist](https://ddat-capability-framework.service.gov.uk/role/content-strategist) – the UK government's role description, which sets the job at expert-practitioner level with a focus on governance, workflows and taxonomy; one employer's definition, but a useful benchmark for a job description.\n- [The content operating model in the AI age](https://contentious.ltd/resources/the-content-operating-model-in-the-ai-age) – Contentious on the seven questions a strategy answers and the operating model that puts those answers to work." }, { "term": "Content strategy", "alsoCalled": [], "slug": "content-strategy", "definition": "Deciding why content exists, whom it serves and how it’s made, organised and looked after.", "category": "Foundations", "related": [ "business-case", "channel-strategy", "competitor-analysis", "content", "content-audit", "content-governance", "content-model", "content-plan", "content-policy", "content-strategist", "editorial-board", "head-of-content", "information-architecture", "kpis", "microsite", "workflow" ], "updated": "2026-10-11T06:24:00.000Z", "url": "https://contentlexicon.com/content-strategy", "markdown": "https://contentlexicon.com/content-strategy.md", "credit": "Content strategy, from the Content Lexicon by Contentious (https://contentlexicon.com/content-strategy), CC BY 4.0", "body": "The term names both a discipline and what it produces. A content strategy is the set of decisions an organisation has settled about its content; content strategy as a practice is the work of reaching, recording and revisiting them. Together they shape what gets made, changed, maintained or stopped.\n\nThe term means different things to different people. In marketing circles it often means [content marketing](https://contentlexicon.com/content-marketing) strategy: campaigns, channels and [editorial calendars](https://contentlexicon.com/content-calendar) designed to attract and convert customers. That’s a worthwhile discipline, but it covers only part of what the name describes. Most of an organisation’s content isn’t marketing, and the guidance, reports, training, policies and help pages need direction as much as the marketing campaigns do.\n\nOthers use content strategy to mean the core strategy alone: a short statement of what content is for and how it supports the organisation’s aims. That statement sits at the centre of the discipline, but on its own it can’t say how a [website](https://contentlexicon.com/website) should be organised or who decides when a page is retired.\n\nThe fuller view owes much to Kristina Halvorson and Brain Traffic, whose work from the late 2000s helped establish content strategy as a discipline in its own right. Brain Traffic’s original content strategy quad surrounds the core with substance, structure, [workflow](https://contentlexicon.com/workflow) and governance: what the content says, how it’s organised and the processes, tools and people that make and look after it.\n\nContentious frames content strategy as the answers to seven questions: why, for whom, what, where, by whom, when and how. Why sits at the centre, because the other answers depend on it. Each answer works at two levels, and a [content operating model](https://contentlexicon.com/content-operating-model) holds both: the strategy sets the rules, once, and day-to-day operations apply them, continually.\n\n## Why it matters for content\n\nWithout a strategy, content decisions default to whoever asked most recently or most loudly. The result is familiar: a site that grows by accretion, duplicates itself, speaks in several [voices](https://contentlexicon.com/voice) and serves internal structures rather than external needs. A strategy replaces that drift with decisions people can point to, and the test of a good one is whether it helps them say no. A document that approves of everything and rules out nothing is a wish list. A strategy narrows the field, which is uncomfortable, and then makes the remaining decisions much easier.\n\nBecause it reaches beyond the words, a strategy also shapes decisions that don’t look editorial at first glance: the [content model](https://contentlexicon.com/content-model) and [information architecture](https://contentlexicon.com/information-architecture), the roles and workflow, the tools and who has the authority to decide. A page can be beautifully written and still be a strategic mistake, and a well-chosen page can still fail if nobody has the time or responsibility to keep it current.\n\nChanges such as a redesign, merger or new service can make strategic choices urgent. The same direction also helps in ordinary weeks, when a new request might otherwise reopen decisions the team has already made.\n\n## How AI changes the picture\n\nWhen drafting was slow and expensive, the effort itself rationed what got made. AI removes that constraint, and without direction to replace it, organisations produce more content that nobody asked for, written to no clear purpose in a voice that belongs to no one. That’s the recipe for [AI slop](https://contentlexicon.com/ai-slop): fluent, plausible material that adds to the pile without helping anyone, and that someone will eventually have to check, correct or remove. A model can write competently, but it can’t decide what deserves to be written. A [prompt](https://contentlexicon.com/prompt) is only as good as the thinking behind it, and much of that thinking is strategy: knowing whom the content serves, what it needs to achieve and what good looks like.\n\nAI also adds a new layer to the questions strategy has always asked. Assistants, search systems and agents now read, summarise and act on content, so AI has become part of the audience, even if it’s rarely the final one. Strategy needs to consider what people should gain when they encounter content through an assistant. A learning publisher might support questions that help a reader understand an argument, while keeping the fuller explanation and evidence available. A service might need an approved answer with a clear route to further help. These are choices about what content does for people, not only how a machine retrieves it.\n\nAI is a creator and increasingly a maintainer too, drafting, tagging, translating and flagging pages that have drifted out of date. That changes the answers to who makes content and how. Strategy has to say where AI helps, where people lead, who reviews its output and who is accountable for what’s published. Voice guidelines, source material and editorial standards become working instructions as well as reference documents, because they shape what the tools produce.\n\nSubstance and structure matter more as a result. What makes content worth reading – original evidence, expertise, experience and a point of view – is what a model can’t supply on its own. And because machines draw on content they can parse and trust, well-modelled, well-maintained content with clear ownership becomes part of the strategy rather than a technical afterthought. The more content AI makes possible, the more it matters that someone has decided what it’s for.\n\n## Questions people ask\n\n**How long should a content strategy be?**\n\nThe core statement should be short enough to remember, often a single sentence: what the organisation’s content combines, what that achieves and for whom. A software company might say it combines clear documentation, worked examples and plain release notes so that developers can get its [API](https://contentlexicon.com/api) working on their own and trust it when it changes. The decisions about substance, structure, workflow and governance that support it can run longer, as long as people go back to them.\n\n**Is content strategy the same as a content plan?**\n\nA strategy establishes direction and priorities. A plan turns them into work: what to create, improve, merge or retire, who is responsible and what happens next. A publication schedule can form part of a plan, but it isn’t the whole thing.\n\n**How is strategy different from a content operating model?**\n\nThe strategy is part of the operating model. Each of the seven questions takes two answers: a rule, set once, and a call, made every time. A strategy settles where content goes in general: the role of each channel and how the site is organised. An editor deciding where one piece should appear is making a call within those rules. The operating model holds both, along with the people, processes, tools and governance that make each everyday call easy, consistent and checkable.\n\n**Who should own a content strategy?**\n\nSomeone with the authority to make its decisions stick, usually a senior leader responsible for communications, digital or services, and sometimes the chief executive. A [content strategist](https://contentlexicon.com/content-strategist) or [content team](https://contentlexicon.com/content-team) can draft and maintain it, but if the owner can’t settle a dispute between two departments, the strategy will be renegotiated in every meeting.\n\n**Can AI write our content strategy?**\n\nIt can help analyse evidence, compare coverage and draft options against agreed criteria. The commitments still need people to make and own them, because they depend on authority, budget and accountability for what follows. And the cheaper AI makes content, the more those commitments matter.\n\n## Learn more\n\n- [The Discipline of Content Strategy](https://alistapart.com/article/thedisciplineofcontentstrategy/) – Kristina Halvorson’s 2008 call to stop treating content as somebody else’s problem, with the definition much of the field still starts from and the specialisms it brings together.\n- [Brain Traffic’s content strategy quad](https://www.braintraffic.com/blog/brain-traffic-lands-the-quad) – the original model surrounding a core strategy with substance, structure, workflow and governance; the archived post links to Brain Traffic’s later revision.\n- [The Elements of Content Strategy](https://elements-of-content-strategy.abookapart.com/) – Erin Kissane’s short book tracing the discipline to editing, curation, marketing and information science, free to read online.\n- [The content operating model in the AI age](https://contentious.ltd/resources/the-content-operating-model-in-the-ai-age) – Contentious on the seven questions and the operating model that holds a strategy’s answers and puts them to work." }, { "term": "Content team", "alsoCalled": [], "slug": "content-team", "definition": "The group of people and roles responsible for planning, creating, governing and maintaining content.", "category": "Roles", "related": [ "agile", "centralised-governance", "content-designer", "content-editor", "content-manager", "content-operations-manager", "content-owner", "content-strategist", "decentralised-governance", "head-of-content", "hub-and-spoke", "senior-content-editor", "writer" ], "updated": "2026-10-10T17:06:00.000Z", "url": "https://contentlexicon.com/content-team", "markdown": "https://contentlexicon.com/content-team.md", "credit": "Content team, from the Content Lexicon by Contentious (https://contentlexicon.com/content-team), CC BY 4.0", "body": "A content team look after an organisation's content: everything from planning and research through writing, [editing](https://contentlexicon.com/editing) and design to governance and maintenance. It may be a single department with its own head, or a network of specialists spread across communications, product, policy and service teams who share standards and work together.\n\nIn everyday use, the team means the people whose job is content: [content strategists](https://contentlexicon.com/content-strategist), [content designers](https://contentlexicon.com/content-designer), [writers](https://contentlexicon.com/writer), editors, [content managers](https://contentlexicon.com/content-manager) and [content operations](https://contentlexicon.com/content-operations) specialists, sometimes joined by [SEO](https://contentlexicon.com/seo), [localisation](https://contentlexicon.com/localisation) or video specialists. [Subject matter experts](https://contentlexicon.com/subject-matter-expert), legal reviewers and developers shape a great deal of what's published, but they have jobs of their own and sit outside the team as close collaborators. Much of a content team's effectiveness depends on how well those relationships work.\n\n## Why an organisation needs a content team\n\nContent made by individuals scattered across an organisation tends to drift: different [voices](https://contentlexicon.com/voice), duplicated pages and no shared view of what exists or who maintains it. A content team gives that work a shape. It lets evidence reach writers, decisions happen at the right level and published material stay accurate. Strong content teams combine editorial craft with [systems thinking](https://contentlexicon.com/systems-thinking): they care about the sentence and the [workflow](https://contentlexicon.com/workflow) that will produce the next thousand sentences.\n\nCapability matters more than job titles alone. A small organisation may need several skills in one role, while a large one needs specialists and communities of practice. Check whether the work is covered, people have enough time and content expertise can influence decisions before the final [proofreading](https://contentlexicon.com/proofreading) round.\n\n## How content teams are organised\n\nMost content teams take one of three broad shapes, and many organisations move between them as they grow. Each strikes a different balance between consistency and closeness to the work, and each has its own entry.\n\nIn a centralised team, everyone who works on content sits in one unit with its own head, often within communications, marketing or digital, and the rest of the organisation brings it requests. Specialists have peers to learn from and a career path to follow, and the organisation gets a consistent voice and clear accountability. The cost is distance: the team can turn into a queue and may know less about a subject than the people who work on it. This is the team shape that goes with [centralised governance](https://contentlexicon.com/centralised-governance).\n\nA [dispersed model](https://contentlexicon.com/decentralised-governance) places content people in the teams they serve, such as a content designer in each product squad or a writer in each department. Work moves faster and stays close to subject knowledge, but embedded specialists can find themselves the only content person among engineers or policy officers, with nobody to review their work or argue its case. In its loosest form there's no content team at all, and staff publish as part of other jobs. This is how decentralised governance usually looks from the inside.\n\n[Hub and spoke](https://contentlexicon.com/hub-and-spoke) combines the two. A small central team sets standards and provides tools and support, while specialists or trained publishers in other teams do the day-to-day work. Because those publishers usually report to their own managers, the central team has to work through influence, training and a community of practice rather than instruction.\n\nNo shape suits every organisation. Size, risk, the number of publishers and the way the organisation already works all bear on the choice. A familiar path runs from dispersed by default, through a central team formed to restore order, to hub and spoke once standards are established. Agencies and freelancers can supplement any of these, adding production capacity while strategy and governance stay in-house.\n\n## How AI changes the team\n\nContent teams increasingly include AI agents as well as people: assistants that draft, check work against voice and style guidance, run rolling audits or monitor live content. That changes how a team is designed. Every agent needs instructions, sources and a named person who answers for what it produces, so someone has to own those arrangements just as someone owns the [style guide](https://contentlexicon.com/style-guide) or the [CMS](https://contentlexicon.com/cms). Teams also need new combinations of skills, such as turning editorial standards into instructions and evaluation criteria, and working with developers when an automated step goes wrong.\n\nThe shape of teams is shifting too. In US [content marketing](https://contentlexicon.com/content-marketing) job ads, demand for mid-level generalist roles has fallen sharply since 2023, while hands-on production roles and senior ownership roles have grown. Whether or not that pattern spreads to other sectors, it raises a practical question for any team: if the middle rungs disappear, where do people develop the judgement senior roles require? Training and progression deserve as much thought as tooling.\n\n## Questions people ask\n\n**How big should a content team be?**\n\nIt depends on the size, risk and complexity of the content estate. A small specialist newsletter might be sustained by a writer and editor with subject input; a multilingual service may need dedicated maintenance and localisation capacity too. These are illustrations, not staffing standards. Map workload and risk before deciding the number.\n\n**Should content sit in marketing, design or product?**\n\nAny of these can work if content has appropriate authority and close relationships across the organisation. The reporting line matters less than whether expertise reaches real decisions.\n\n**Do subject experts belong to the content team?**\n\nUsually not. They're accountable for what's factually right in their area, while the content team is accountable for making it clear and usable. The partnership works best when each side knows where its responsibility ends and experts are involved early, rather than handed a finished draft to approve.\n\n**Where should a small organisation start?**\n\nUsually with one person who can both write and manage: someone able to plan, edit and keep the estate in order, with access to subject experts and some budget for freelance or specialist help. Specialist roles such as [content design](https://contentlexicon.com/content-design) or content operations follow when volume, risk or the number of publishers justifies them.\n\n## Learn more\n\n- [Content team roles](https://shorthand.com/the-craft/content-team-roles/index.html) – a broad look at eighteen roles and how leaders shape teams.\n- [We Analyzed 8,000 Content Marketing Job Listings](https://www.semrush.com/blog/content-marketing-job-market-study) – Semrush's 2026 analysis of US job ads, showing growth at the production and senior ends and a sharp fall in mid-level generalist roles; one sector's market, not a template for team design.\n- [Who does what?](https://contentious.ltd/resources/who-does-what) – a field guide to the human roles and AI agents that now make up a capable content team." }, { "term": "Content type", "alsoCalled": [], "slug": "content-type", "definition": "A category of content with a defined purpose and structure, such as a case study or an event.", "category": "Structure and systems", "related": [ "content-model", "content-pattern", "schema", "structured-content" ], "updated": "2026-10-11T06:24:00.000Z", "url": "https://contentlexicon.com/content-type", "markdown": "https://contentlexicon.com/content-type.md", "credit": "Content type, from the Content Lexicon by Contentious (https://contentlexicon.com/content-type), CC BY 4.0", "body": "A content type is a defined kind of content with a consistent purpose and structure: an event, a case study, a job vacancy, a product page or a news story. Defining one means deciding which parts every item of that kind contains, which are optional, what format each takes and how it relates to other content. An event might always have a title, date, summary and booking route, with venue details for in-person events and joining instructions for online ones. Every event then follows the same pattern, however and wherever it's displayed.\n\nThe idea reached content work from technical communication. IBM's DITA, published in 2001 and approved as an OASIS standard in 2005, split documentation into typed topics (concept, task and reference), each with its own structure. Web content management adopted the same thinking, and Rachel Lovinger's 2012 article in A List Apart described content types as one of the three building blocks of a [content model](https://contentlexicon.com/content-model).\n\n## Why it matters for content\n\nContent types are how a pile of pages becomes a system. Once a type is defined, an organisation can make reliable promises about it: what it always contains, how it appears in a listing, how it behaves in search, who owns it, how often it's reviewed and when it should be retired. Authors get clearer guidance about what to write, developers get something to build against and readers get consistency they may never consciously notice. Every job vacancy shows its closing date in the same place, and every case study names the client and the outcome.\n\nA type also separates what content is from how it looks. The same event record can appear as a card in a listing, a full page and a line in a newsletter, each with its own template. That separation lets content move between channels without being rewritten, and it makes migrations less painful, because comparing fields shows which records convert cleanly and which exceptions need a decision.\n\nThe hard part is drawing the boundaries. With too few types, everything ends up in a generic page with one large body field, which is flexible but tells systems nothing. With too many, authors can't tell which to choose and developers maintain near-duplicates. Lovinger's tests help: something deserves its own type when it has distinct elements, distinct functional needs or distinct organisational requirements, such as press releases that must feed a newsroom automatically.\n\n## How AI changes the picture\n\nDefined types make content easier for machines to use. When dates, eligibility, prices and locations sit in their own fields, a retrieval system or assistant can find them without inferring them from prose, and a generated summary can be checked against the expected shape. A free-text page leaves the model to guess which of several dates is the deadline.\n\nAI can help design and populate types too. A model can propose candidate types and fields from existing content, cluster pages that share a structure and test a draft model against awkward cases, revealing for example that an event type has no way to record a rescheduled date. During a migration it can classify legacy pages by type and draft content into the right fields. People still need to decide which distinctions the organisation should maintain, and a valid record can still contain a wrong fact.\n\n## Questions people ask\n\n**What is the difference between a content type and a template?**\n\nA content type defines what the content is and what it contains. A template defines how one presentation of it looks. One type can have several templates: a listing card, a full page and an [email](https://contentlexicon.com/email) version, for instance.\n\n**Is a content type the same as a content model?**\n\nA content type is one part of a content model. The model is the wider picture: all the types, their fields and the relationships between them.\n\n**How many content types does an organisation need?**\n\nAs many as there are meaningful differences in purpose, structure or behaviour, and no more. A small programme might need separate event and speaker types, but not a new event type for every department. Test the model against real examples before settling the number, and merge types that turn out to be identical apart from their names.\n\n**Can a page belong to two types?**\n\nA record normally has one type, but a page can assemble several: an event, a speaker profile and related guidance, for example. Be clear whether you're classifying the record or the presentation.\n\n## Learn more\n\n- [Content Modelling: A Master Skill](https://alistapart.com/article/content-modelling-a-master-skill/) – Rachel Lovinger's A List Apart article on building a content model from an assembly model, content types and attributes, with tests for when something deserves its own type.\n- [History of DITA](https://dita-lang.org/history) – how IBM's concept, task and reference topics grew from an internal workgroup into an OASIS standard, with a timeline of later releases.\n- [An introduction to structured content](https://digital.gov/resources/an-introduction-to-structured-content) – Digital.gov on publishing content as modular, machine-readable pieces, using an event that can feed a calendar, a news feed and search results as its example." }, { "term": "Context engineering", "alsoCalled": [], "slug": "context-engineering", "definition": "Designing the instructions, examples, source material and surrounding information an AI system needs to do useful work.", "category": "AI and tooling", "related": [ "content-brief", "context-window", "conversational-design", "llm", "markdown", "prompt", "rag", "skill", "system-prompt", "ubiquitous-language" ], "updated": "2026-10-06T16:06:00.000Z", "url": "https://contentlexicon.com/context-engineering", "markdown": "https://contentlexicon.com/context-engineering.md", "credit": "Context engineering, from the Content Lexicon by Contentious (https://contentlexicon.com/context-engineering), CC BY 4.0", "body": "Context engineering covers instructions, examples, source documents, retrieved material, [schemas](https://contentlexicon.com/schema), memory, tools, constraints and other information that shapes how an AI model performs. The focus is less on wording one clever [prompt](https://contentlexicon.com/prompt) and more on assembling the right working context around the task.\n\nIt's a helpful shift, because many AI failures aren't failures of raw intelligence so much as failures of setup. The model was given the wrong material, too much irrelevant material, no clear standard to work to, or no reliable way to tell what counted as success.\n\n## Why it matters for content\n\nContext engineering is where a lot of content work meets AI work directly. [Style guides](https://contentlexicon.com/style-guide), [voice](https://contentlexicon.com/voice) and [tone](https://contentlexicon.com/tone) rules, [taxonomies](https://contentlexicon.com/taxonomy), examples, definitions of done, [knowledge bases](https://contentlexicon.com/knowledge-base) and structured source material all become part of the context an assistant uses to generate, transform or evaluate content. If that material is vague, contradictory or inaccessible, the output is more likely to be vague, contradictory or unreliable too.\n\nThe payoff is biggest when AI is used for repeated tasks rather than one-off experiments: drafting pages, summarising research, rewriting copy, classifying material, checking content against standards or answering questions from a document set.\n\n## How context engineering changes the picture\n\nFor [content teams](https://contentlexicon.com/content-team), familiar editorial assets stop being passive reference material and start acting as inputs. The style guide stops being a [PDF](https://contentlexicon.com/pdf) people skim and becomes the instruction the assistant follows on every draft. The taxonomy that once organised a navigation menu now decides which documents get retrieved.\n\nThat raises the stakes on clarity, structure and maintenance. A style guide that says \"be concise\" in one section and \"explain fully\" in another used to produce two kinds of draft from two [writers](https://contentlexicon.com/writer), who could ask which rule applied. An assistant can ask too, if clarification is part of its [workflow](https://contentlexicon.com/workflow). Without a clear priority or a way to resolve the conflict, its drafts may follow different rules from one attempt to the next. An out-of-date policy page left in the source set gets quoted with the same confidence as the current one.\n\n## Questions people ask\n\n**Is context engineering the same as prompt engineering?**\n\nIt includes it. Prompt engineering focuses on the wording of the instruction; context engineering also covers the examples, sources, structure, tools and constraints around it.\n\n**Do I need to be technical to do context engineering?**\n\nNot always. Technical implementation may be needed for more advanced systems, but content people often do a large part of the useful work by improving source material, examples, standards and structure.\n\n**Is this only relevant for AI agents?**\n\nNo. It applies to simple assistants and repeatable prompting as well. The more often a task is repeated, the more useful it is to engineer the context rather than improvise every time.\n\n**Does better context remove the need for review?**\n\nBetter context improves the chances of good output, but review is still needed, especially where the stakes are high or the judgement is editorial rather than mechanical.\n\n## Learn more\n\n- [Effective context engineering for AI agents](https://www.anthropic.com/engineering/effective-context-engineering-for-ai-agents) – Anthropic's September 2025 account of selecting context, designing tools and maintaining longer-running agents through compaction and notes.\n- [Context engineering vs. prompt engineering](https://www.elastic.co/search-labs/blog/context-engineering-vs-prompt-engineering) – a January 2026 example of focused retrieval from a 103,063-book dataset, built with Elastic's own tools.\n- [Context Engineering Guide](https://www.promptingguide.ai/guides/context-engineering-guide) – a broad, example-led introduction to prompts, memory, retrieval and tools." }, { "term": "Context window", "alsoCalled": [], "slug": "context-window", "definition": "The amount of information an AI system can consider at once when generating a response.", "category": "AI and tooling", "related": [ "context-engineering", "llm", "prompt", "rag", "system-prompt" ], "updated": "2026-10-06T15:28:00.000Z", "url": "https://contentlexicon.com/context-window", "markdown": "https://contentlexicon.com/context-window.md", "credit": "Context window, from the Content Lexicon by Contentious (https://contentlexicon.com/context-window), CC BY 4.0", "body": "A context window covers everything an AI model has in play for a response: the instructions it has been given, the documents supplied to it, the conversation so far and the answer it is producing. Beyond that limit, things fall away.\n\nIt helps to think of it as a desk rather than a memory. Whatever is on the desk can be used. Whatever has been cleared off is gone, however important it seemed twenty minutes ago. Larger windows mean bigger desks, but not necessarily better judgement about what deserves to be on them.\n\n## Why it matters for content\n\nThe context window is the reason AI output quality depends so heavily on what you put in front of the model. A [style guide](https://contentlexicon.com/style-guide), a set of examples, a factual source and a clear brief will produce far better content than a clever [prompt](https://contentlexicon.com/prompt) alone – but only if they fit, and only if they're legible enough to be used well.\n\nIt also explains some familiar frustrations. An assistant that followed your style rules beautifully at the start of a long session and has stopped doing so has probably let those rules slide under a pile of newer papers. Some tools trim long uploads, so a summary of a 200-page report may rest on a fraction of it. Neither is mysterious once you know the limit exists.\n\nContent work that leans on substantial source material feels this hardest: brand guidelines, technical documentation, research transcripts, large [content inventories](https://contentlexicon.com/content-inventory). There, deciding what goes into the window is an editorial act.\n\n## How context windows change the picture\n\nWindows have grown enormously, and the temptation is to treat capacity as a solution to everything. It isn't. Filling a large window with everything available tends to dilute attention, bury the important instructions and produce blander, vaguer output. Choosing less, and choosing well, works better, as [content teams](https://contentlexicon.com/content-team) learned about [websites](https://contentlexicon.com/website) long before anyone had a context window.\n\nThis is why retrieval matters. Rather than pasting an entire [knowledge base](https://contentlexicon.com/knowledge-base) in, well-designed systems fetch the few relevant pieces and put those on the desk. That only works if content and context are structured, chunked sensibly and written so that a section makes sense away from its neighbours, which long undifferentiated pages rarely do.\n\nIt also reframes what content teams are producing. Style guides, [glossaries](https://contentlexicon.com/glossary), examples, canonical policy pages and clean source documents aren't just reference material for colleagues any more; they're inputs competing for finite space in a machine's attention. Content written to be extracted, quoted and reused gets used. Content written as an atmospheric brochure doesn't.\n\n## Questions people ask\n\n**Does a bigger context window mean better answers?**\n\nNot reliably. It means more can be considered, not that the right things will be.\n\n**Why does the assistant forget my instructions in a long conversation?**\n\nBecause long conversations dilute attention to early instructions, and many tools trim or summarise earlier turns once the window fills. Repeating key instructions, or putting them in a [system prompt](https://contentlexicon.com/system-prompt), keeps them in play.\n\n**Can I just paste our whole website in?**\n\nUsually you can, and usually you shouldn't. You'll get better results from the handful of pages that answer the question.\n\n**What should I put in the window for content work?**\n\nThe brief, the relevant source of truth, the [voice](https://contentlexicon.com/voice) and style rules, and one or two examples of the output you want. That combination does most of the work.\n\n## Learn more\n\n- [What is a context window?](https://www.ibm.com/think/topics/context-window) – an accessible explanation of tokens, limits and why size isn't the same as understanding.\n- [Context Rot: How Increasing Input Tokens Impacts LLM Performance](https://www.trychroma.com/research/context-rot) – Chroma's 2025 tests of 18 models, showing output getting less reliable as input grows, even on simple tasks. The introduction is readable; the rest is technical.\n- [Lost in the Middle: How Language Models Use Long Contexts](https://arxiv.org/abs/2307.03172) – the 2023 paper that found models use information at the start and end of a long input far better than material in the middle." }, { "term": "Continuous integration", "alsoCalled": [], "slug": "continuous-integration", "definition": "A practice where changes are regularly checked and merged so problems surface early.", "category": "AI and tooling", "related": [ "automated-testing", "content-health-check", "content-maintenance", "devops", "linting" ], "updated": "2026-10-11T13:23:00.000Z", "url": "https://contentlexicon.com/continuous-integration", "markdown": "https://contentlexicon.com/continuous-integration.md", "credit": "Continuous integration, from the Content Lexicon by Contentious (https://contentlexicon.com/continuous-integration), CC BY 4.0", "body": "Continuous integration is the practice of merging small changes into a shared main version frequently, often several times a day, with automated checks run on every merge. Problems show up while each change is still small enough to understand and fix.\n\nIt comes from Extreme Programming, the software method Kent Beck developed in the late 1990s, and Martin Fowler's article Continuous Integration, first published in 2000, spread it more widely. The working rules are short. Everyone integrates their work with the main version at least daily. Each change triggers an automated build and a set of tests. If the build breaks, fixing it comes before any new work, so the main version is always in a usable state.\n\n## Why it matters for content\n\nContent often builds up in documents until a migration or launch, when hundreds of changes meet at once. Frequent integration reveals missing fields, broken links, terminology differences and structural problems earlier, while they're cheap to fix.\n\nA content version runs checks every time someone proposes a change. A prose [linter](https://contentlexicon.com/linting) such as Vale tests the text against the [style guide](https://contentlexicon.com/style-guide), flagging banned terms, inconsistent product names or headings that break the capitalisation rule. Link checkers catch broken internal and external links. Validators confirm that required [metadata](https://contentlexicon.com/metadata) is present and structured data is well formed. Automated [accessibility](https://contentlexicon.com/accessibility) tests catch specified failures, such as missing [alt text](https://contentlexicon.com/alt-text) or skipped heading levels, though they can't establish that the whole experience is accessible. Docs-as-code teams usually run these checks on each [pull request](https://contentlexicon.com/pull-request), so the [writer](https://contentlexicon.com/writer) sees the results before a reviewer does.\n\nThe same rule about broken builds applies. If the checks fail on the main version, someone fixes that before publishing anything else.\n\nContinuous integration depends on a shared source and version history. Publishing more often isn't the same thing; changes have to be combined and checked in a reliable process. Review also needs to cover meaning, evidence and how people use the result.\n\n## How AI changes the picture\n\n[Prompts](https://contentlexicon.com/prompt), knowledge sources and evaluation cases can enter the same process. A team could test whether a changed eligibility rule still produces the right explanation for a set of sample questions. These tests check selected behaviour. They can't predict every future answer.\n\nAI can help write tests and explain failures, but generated checks need review. A system can pass every formal rule while producing an unhelpful answer.\n\nSmall, visible changes make AI behaviour easier to trace than occasional wholesale prompt rewrites.\n\n## Questions people ask\n\n**Is continuous integration the same as continuous delivery?**\n\nNo. Integration combines and checks changes. Continuous delivery keeps successful changes ready for release, and continuous deployment goes further by releasing them automatically after the required checks.\n\n**Must content be stored in Git?**\n\nNo, though the system needs version history, a shared source and a way to run checks on each change.\n\n**What content checks can be automated?**\n\nStyle rules, terminology, links, required fields, formats, some accessibility patterns and some consistency checks are good candidates. Start with a few rules and add more as the team trusts them.\n\n**Does automation remove editorial review?**\n\nIt can reduce routine checking. Someone still needs to decide whether the checks are appropriate and review what they don't cover.\n\n## Learn more\n\n- [Continuous Integration](https://www.martinfowler.com/articles/continuousIntegration.html) – Martin Fowler's detailed account of frequent integration, automated builds and repairing failures; a software practice to adapt thoughtfully to content.\n- [Continuous documentation deployment](https://docs.readthedocs.com/platform/stable/continuous-deployment.html) – a concrete example of documentation entering the same build, review and release rhythm as software.\n- [How to introduce prose linters at your workplace](https://passo.uno/prose-linters-implement-workplace-howto/) – Fabrizio Ferri-Benedetti's practical advice on bringing a style linter such as Vale into a writing team, including adding rules to the pipeline gradually." }, { "term": "Conversational design", "alsoCalled": [ "Conversation design" ], "slug": "conversational-design", "definition": "Designing how an interaction unfolds in language, especially in chat, voice and other back-and-forth interfaces.", "category": "Editorial craft", "related": [ "agentic-ai", "content-design", "context-engineering", "email", "frontend", "microcopy", "tone" ], "updated": "2026-10-07T12:54:00.000Z", "url": "https://contentlexicon.com/conversational-design", "markdown": "https://contentlexicon.com/conversational-design.md", "credit": "Conversational design, from the Content Lexicon by Contentious (https://contentlexicon.com/conversational-design), CC BY 4.0", "body": "Conversational design covers what a system says, what a person might say next, how the exchange moves forward and how it recovers when something goes wrong. It is used in chatbots, voice interfaces, assistants, onboarding flows and other places where content behaves more like a conversation than a static page.\n\nIt is not simply writing in a chatty [tone](https://contentlexicon.com/tone). Good conversational design considers intent, turn-taking, context, timing, repair and the actions available on both sides. A system that sprinkles “Great question!” over a baffling journey has acquired manners, not conversational design.\n\n## Why it matters for content\n\nIn a conversational interface, words do much of the structural work usually shared by navigation, layout and controls. The content must tell people what the system can do, invite an appropriate response and make the next step clear. Small ambiguities can derail the whole exchange because there may be no page around the words to provide context.\n\nConversational design brings together [content design](https://contentlexicon.com/content-design), interaction design, service design and user research. It needs an understanding of the language people naturally use, the task they are trying to complete and the limits of the service behind the interface. It also needs careful work on failure: misunderstood requests, unavailable information, sensitive topics, silence, interruptions and requests the system should not fulfil.\n\nTrust depends on honest framing. People should know whether they are dealing with automation, what will happen to their information and when a person can help. Pretending a system is more capable or more human than it is may produce a charming first minute and a very sour second one.\n\n## How AI changes the picture\n\nTraditional conversational interfaces often followed designed paths with recognised intents and prepared responses. Generative AI makes the exchange more flexible: people can phrase requests freely and the system can produce answers that weren't written in advance. That opens up conversation as a way to learn and explore, as well as complete a task.\n\nAn exhibition visitor could ask questions of its source material, request an explanation at a different level or follow a connection between two objects. The material wouldn't need to have been written as a script for every possible question. This is an illustrative design possibility, not a claim about a particular museum. The exchange should help someone develop their own understanding, with routes back to the objects and evidence.\n\nThe designer's work moves beyond scripting individual lines to include system instructions, source content, boundaries, tool behaviour, memory and escalation. The conversation needs to reveal where information came from, allow correction and make uncertainty clear. [Voice](https://contentlexicon.com/voice) and tone guidance should help with difficult moments and sensitive language, keeping warmth distinct from false intimacy. People need to know what they're talking to and what it can support.\n\n## Questions people ask\n\n**Is conversational design just chatbot copywriting?**\n\nNo. The words matter, but so do the possible paths, system capabilities, errors, timing and recovery. It’s the design of the interaction, not merely the polish on each message.\n\n**Should a conversational interface sound human?**\n\nIt should sound clear and natural, but it should not deceive people about what it is. A useful voice is better than an elaborate personality that hides limitations.\n\n**Do we still need designed flows with generative AI?**\n\nYes. People need predictable ways to start, correct, confirm and escalate. Generative freedom works best inside a deliberately designed service.\n\n**How do we test a conversation?**\n\nUse realistic tasks with varied participants, observe where expectations diverge and test difficult states as seriously as successful ones. A transcript can look plausible while the experience remains frustrating.\n\n## Learn more\n\n- [Conversational Design for You](https://www.muledesign.com/blog/conversational-design-for-you) – Erika Hall introduces her book and offers an authorised download; the post describes its availability as limited-time.\n- [Conversational Design](https://abookapart.com/products/conversational-design.html) – A Book Apart’s description of Erika Hall’s book on language-based interactions, with readers now directed to the author rather than a shop.\n- [AI turns news into a conversation](https://www.niemanlab.org/2024/12/ai-turns-news-into-a-conversation) – Nikita Roy’s December 2024 forecast imagines journalism that readers can question and explore through dialogue, offering a perspective beyond service chatbots." }, { "term": "Copy editing", "alsoCalled": [ "Copyediting" ], "slug": "copy-editing", "definition": "Editing a draft for clarity, consistency and correctness within an agreed scope.", "category": "Editorial craft", "related": [ "automated-testing", "content-editor", "editing", "grammar", "linting", "proofreading", "pull-request", "senior-content-editor", "style-guide", "word-list" ], "updated": "2026-10-11T09:06:00.000Z", "url": "https://contentlexicon.com/copy-editing", "markdown": "https://contentlexicon.com/copy-editing.md", "credit": "Copy editing, from the Content Lexicon by Contentious (https://contentlexicon.com/copy-editing), CC BY 4.0", "body": "Copy [editing](https://contentlexicon.com/editing) works sentence by sentence on a draft whose purpose and structure are already settled. A copy editor corrects [grammar](https://contentlexicon.com/grammar), spelling and [punctuation](https://contentlexicon.com/punctuation), applies house style, makes names, numbers and terms consistent throughout, checks references and links, and flags claims that look unsupported or risky. Its depth varies: a light edit may change little beyond style points, while a heavy one rewords sentences and queries whole passages. Agree which you're commissioning before the work begins, including whether fact checking and rewriting are in scope.\n\nThe copy editor might be a freelance specialist, a member of a publications team or, in many organisations, a [content editor](https://contentlexicon.com/content-editor) or colleague doing it alongside other work. Much of the craft lies in the query: a note asking the [writer](https://contentlexicon.com/writer) to confirm a figure, resolve an ambiguity or choose between two readings, in place of an editor's silent guess. Copy editors often keep a style sheet as they go, recording decisions specific to the piece, such as how a name is spelled or whether a term takes a capital, so that the choice made on page four is still the choice on page forty.\n\n## Why it matters for content\n\nReaders judge reliability from details. A figure that differs between the summary and the table, a programme called by three different names or a link to a retired page tells them something about the care behind the whole piece, even when the argument is sound. Copy editing is where those details get caught, and where small ambiguities that would trip a reader up (does “they” mean the trustees or the beneficiaries?) are settled with the writer instead of being left for the reader to guess.\n\nConsistency also matters beyond the single piece. When a service, condition or product is named and described the same way across a [website](https://contentlexicon.com/website), people searching for it find it, cross-references hold and AI agents drawing on the content are less likely to treat one thing as two. A [style guide](https://contentlexicon.com/style-guide) and [word list](https://contentlexicon.com/word-list) record the decisions; copy editing applies them and notices when a new case needs a decision nobody has made yet. It's often the practical moment, too, when link text, heading order, [alt text](https://contentlexicon.com/alt-text) and abbreviations get checked for [accessibility](https://contentlexicon.com/accessibility).\n\n## How AI changes the picture\n\nCopy editing is the stage of editing that automation reaches most easily, because so much of it can be written down as rules. Prose [linters](https://contentlexicon.com/linting) such as Vale turn a style guide into checks that run as people write: a preferred term, a banned phrase, a capitalisation pattern. Language models extend this to judgements that are hard to express as rules, suggesting a clearer sentence or spotting that a paragraph contradicts an earlier one. Together they can take a large share of consistency checking off a person's desk and apply it across thousands of pages rather than one document.\n\nThat changes where a copy editor's attention goes. Someone has to write and maintain the rules, decide which suggestions a tool may apply automatically and which it may only flag, and keep the style sheet in a form that people and tools can both use. Reviewing automated changes becomes a skill in its own right. A model asked to tidy a sentence may turn “may” into “will”, correct a deliberately unusual spelling, smooth a direct quotation or “fix” someone's name, and each change is easy to miss if it isn't tracked. Insisting that tools make visible, reviewable suggestions in place of silent edits keeps the habit of querying that makes copy editing trustworthy.\n\nAI-assisted drafts need a further check. Language models are known to produce references that don't exist and statistics that sound right without actually being true, so a copy edit of such a draft should confirm that cited sources exist and say what the text claims they say. That's slower work than tidying commas.\n\n## Questions people ask\n\n**What should we give a copy editor before they start?**\n\nThe style guide and word list, the intended audience, a note on the depth of edit you want and the name of whoever will answer queries. Say whether fact checking, link checking and accessibility checks are included. Most delays in copy editing come from queries waiting for answers.\n\n**Where does copy editing end and proofreading begin?**\n\nCopy editing improves the text; [proofreading](https://contentlexicon.com/proofreading) checks the final version, in its finished layout, for errors that remain or crept in along the way. On a website the two often collapse into one pass, which is workable for small updates. For a report or anything designed, keep a separate proofread of the laid-out version, ideally by someone who didn't do the copy edit.\n\n**Will a copy editor check our facts?**\n\nUsually only lightly: names, dates, links, sums and figures that can be compared with a source supplied alongside the draft. Full fact checking is a different job and needs commissioning as such. In most organisations, confirming that a claim is right remains the responsibility of the writer or a subject expert, with the copy editor's queries pointing to where it's needed.\n\n**Can a style checker replace a copy editor?**\n\nFor rule-based consistency on routine content it can do much of the work. It can't tell whether a figure is right, whether “that” or “which” matches what the writer meant, or whether a grammatically correct sentence will mislead its reader. A sensible arrangement gives tools what can be specified and keeps people for meaning, accuracy and queries.\n\n## Learn more\n\n- [Glossary of editorial and publishing terms](https://www.ciep.uk/resource/editorial-glossary.html) – CIEP’s definitions of copyediting, line editing and proofreading side by side, including how far a copy edit may extend depending on the state of the original.\n- [Confessions of a Comma Queen](https://www.newyorker.com/magazine/2015/02/23/holy-writ) – Mary Norris's New Yorker essay on decades at the magazine's copy desk, a lively account of house style and of how many copy-editing decisions are interpretive rather than mechanical." }, { "term": "Copyright and licensing", "alsoCalled": [], "slug": "copyright-and-licensing", "definition": "The rights that decide who owns content, what other people may do with it and the conditions attached to reuse.", "category": "Governance", "related": [ "ai-use-policy", "machine-readable" ], "updated": "2026-10-14T11:54:00.000Z", "url": "https://contentlexicon.com/copyright-and-licensing", "markdown": "https://contentlexicon.com/copyright-and-licensing.md", "credit": "Copyright and licensing, from the Content Lexicon by Contentious (https://contentlexicon.com/copyright-and-licensing), CC BY 4.0", "body": "Copyright is the legal right that gives creators, or whoever holds the rights, control over copying, adapting and sharing original work. A licence is permission from the rights holder to use that work under stated conditions. Together they decide what an organisation can publish, quote, adapt, translate and reuse, and what others may do with the organisation's own content.\n\n[Content teams](https://contentlexicon.com/content-team) meet them everywhere: stock photographs, commissioned copy, charts, quotations, interview transcripts, open data and a paragraph copied into a slide deck because it was \"already on the internet\". Being available isn't the same as being permitted. Details vary by country, but the basic pattern is broadly shared: copyright arises automatically, lasts for decades and covers the expression of an idea rather than the idea itself.\n\n## Why it matters for content\n\nOwnership should be settled before publication, especially for commissioned work and material made with partners. A contract may pay for an asset without transferring copyright. It should also cover practical uses, such as whether a commissioned illustration may be translated, adapted or used in another publication. Stock licences often limit where and how long an image can appear.\n\nCreative Commons licences give a standard way to grant permission in advance. All of them require attribution. Some add conditions: NonCommercial restricts commercial use, NoDerivatives forbids sharing adapted versions and ShareAlike requires adaptations to carry the same licence. These are different conditions and shouldn't be treated as interchangeable.\n\nLicensing can also be a deliberate choice about your own content. An organisation may want people to quote its guidance, adapt training materials or translate a public resource. The Content Lexicon, for example, can be quoted freely as long as the quote links back. A clear statement like that removes unnecessary permission requests and protects the condition that matters.\n\n## How AI changes the picture\n\nGenerative AI has made the boundary between reading, copying and training hotly contested. Different jurisdictions are reaching different answers. In the US, dozens of lawsuits turn on whether training counts as fair use. The UK government's 2026 report describes a developing licensing market, unresolved evidence and several possible policy approaches.\n\nThat leaves [content owners](https://contentlexicon.com/content-owner) with a newer question: whether to let AI companies train on their content, and how to say so. The tools work at different levels. Robots.txt rules can ask named crawlers to stay away, but only crawlers that choose to respect them. Licensing deals set terms with specific AI companies. [Machine-readable](https://contentlexicon.com/machine-readable) licensing standards, such as RSL, let a publisher state terms for AI use alongside its content, from free with attribution to paid. Proposals such as llms.txt help AI tools find content rather than set permissions. None of these is a universal switch, so decide the position first and choose the signals that express it.\n\nCreative Commons warns that choosing a more restrictive CC licence may block useful human reuse without reliably stopping AI training. For material with real risk, take legal advice rather than relying on folklore passed around a group chat.\n\n## Questions people ask\n\n**Can we use something if we credit the creator?**\n\nNot necessarily. Attribution doesn't create permission. Check the licence, check whether a copyright exception applies or ask for consent.\n\n**Who owns commissioned content?**\n\nThe contract should say. Paying for work doesn't automatically answer every ownership and reuse question.\n\n**Does robots.txt stop AI training?**\n\nIt can signal or control access for crawlers that choose to respect it. It isn't a copyright licence, a guarantee or a substitute for clear terms.\n\n**Can AI output itself be copyrighted?**\n\nThe answer depends on jurisdiction and the human contribution. Treat important output as a legal question rather than assuming the [prompt](https://contentlexicon.com/prompt) created ordinary authorship.\n\n## Learn more\n\n- [Understanding CC Licenses and AI Training: A Legal Primer](https://creativecommons.org/2025/05/15/understanding-cc-licenses-and-ai-training-a-legal-primer/) – Creative Commons's 2025 introduction to why licence restrictions may not prevent AI training, with links to its fuller legal analysis.\n- [Report on Copyright and Artificial Intelligence](https://www.gov.uk/government/publications/report-and-impact-assessment-on-copyright-and-artificial-intelligence/report-on-copyright-and-artificial-intelligence) – the UK government's 2026 report on AI training, transparency, technical controls and licensing; one jurisdiction's policy account, not universal legal permission.\n- [RSL: Really Simple Licensing](https://rslstandard.org/) – the open standard for publishing machine-readable licensing terms for AI use, with examples ranging from Creative Commons attribution to paid licences." }, { "term": "Copywriting", "alsoCalled": [], "slug": "copywriting", "definition": "Writing persuasive messages that help people understand an offer and decide whether to act.", "category": "Editorial craft", "related": [ "call-to-action", "content-marketing", "copy-editing", "storytelling", "tone", "writer" ], "updated": "2026-10-11T13:23:00.000Z", "url": "https://contentlexicon.com/copywriting", "markdown": "https://contentlexicon.com/copywriting.md", "credit": "Copywriting, from the Content Lexicon by Contentious (https://contentlexicon.com/copywriting), CC BY 4.0", "body": "Copywriting is writing whose job is to move someone towards an action: buying, signing up, donating, booking or reading on. It covers headlines, straplines, product descriptions, landing pages, adverts, fundraising appeals, [email](https://contentlexicon.com/email) subject lines and the buttons at the end of them. What sets it apart from other writing is that it's judged by what readers do next, as well as by whether they understood.\n\nGood copywriting starts before the words. It needs a clear audience, an offer worth having and evidence for the claims, then language that makes the case quickly and credibly. Adjectives can't rescue a weak proposition: a charity appeal that names one specific thing a gift pays for gives readers more to act on than three paragraphs about passion and commitment.\n\nThe trade grew up in print advertising, and some of its habits have lasted. Claude Hopkins argued in Scientific Advertising (1923) that adverts should be tested and judged by results rather than taste. David Ogilvy's account of his own process in the 1950s – research first, agree the problem with the client, write many headlines, edit repeatedly – reads much like a good brief today.\n\n## Why it matters for content\n\nCopy often sits at the point where someone decides. A headline determines whether a page gets read, a [call to action](https://contentlexicon.com/call-to-action) whether a form gets started, a subject line whether an email gets opened. Small changes at those points can shift outcomes, so copy repays testing and revisiting rather than being written once and left.\n\nPersuasion carries responsibility. Copy that hides conditions, invents urgency, shames people for declining (\"No thanks, I don't like saving money\") or buries the real cost can lift a number for a week and erode trust for much longer. These tactics are well documented as deceptive patterns, and in the UK advertising rules require objective claims to be backed by evidence and material information to be shown clearly. [Content designers](https://contentlexicon.com/content-designer) working alongside marketing teams often end up being the ones who ask whether a promise is one the service can keep.\n\n## How AI changes the picture\n\nCopy is one of the first places organisations put generative AI to work, because the outputs are short, plentiful and easy to compare. A model can produce fifty headline variants in seconds, which echoes the old discipline of writing many and choosing one. The bottleneck moves to choosing: knowing which option fits the audience, the brand [voice](https://contentlexicon.com/voice) and the facts, and which only sounds like advertising.\n\nModels trained on vast amounts of marketing drift towards its most common moves – superlatives, urgency, \"unlock\" and \"elevate\" – and will add a claim nobody approved if it makes the line flow. Give the model the approved facts, offers and limits as inputs, and check every claim in the output against them. A model's prediction of which variant will win doesn't replace running the test with real people in the real context.\n\n## Questions people ask\n\n**How is copywriting different from content writing?**\n\nCopywriting usually aims at a specific action in a short space. Content writing more often explains or informs over a longer stretch, building trust that may lead to action later. The line blurs on product pages and in email, where both jobs happen at once.\n\n**Does every piece of copy need a call to action?**\n\nOnly when there's a next step the reader wants. A confirmation email after a purchase can reassure without asking for anything else, and adding \"Shop more\" at that moment can feel grasping.\n\n**Is clever copy effective?**\n\nIt can be, when the reader gets the point straight away. Wit that delays understanding is charging the audience admission.\n\n**Should copy be long or short?**\n\nAs long as the decision needs. A free newsletter sign-up might need a single line; a course costing thousands needs answers to every question a careful buyer will ask. Long copy works when each paragraph earns its place and fails when it repeats itself.\n\n## Learn more\n\n- [David Ogilvy’s Copywriting Process](https://chasereeves.co/2013/david-ogilvys-copywriting-process) – a 1955 letter, reproduced on Chase Reeves's blog, in which Ogilvy describes studying earlier campaigns, agreeing the problem with the client, writing a long list of headlines and editing each draft several times.\n- [Misleading advertising](https://www.asa.org.uk/advice-online/misleading-advertising.html) – advice from the UK's advertising rule-makers on where persuasion crosses the line, with rulings on unsupported claims, missing information and exaggeration, and when obvious puffery is allowed.\n- [Deceptive Patterns](https://www.deceptive.design/) – a long-running catalogue of design and copy tricks in apps, websites and AI systems that steer people into choices against their interests." }, { "term": "Core content strategy", "alsoCalled": [], "slug": "core-content-strategy", "definition": "A short statement of what content must do for the organisation and its users.", "category": "Foundations", "related": [ "content-principles", "content-strategy", "message-architecture", "persona" ], "updated": "2026-10-11T13:22:00.000Z", "url": "https://contentlexicon.com/core-content-strategy", "markdown": "https://contentlexicon.com/core-content-strategy.md", "credit": "Core content strategy, from the Content Lexicon by Contentious (https://contentlexicon.com/core-content-strategy), CC BY 4.0", "body": "A core [content strategy](https://contentlexicon.com/content-strategy) puts an organisation's content choices into one statement, usually a single sentence, that says what content it will provide, who it's for and what it should achieve. The idea comes from Brain Traffic, the content strategy consultancy founded by Kristina Halvorson, which calls it a core strategy statement and aims to keep it under 30 words.\n\nA university's statement might read: \"Give prospective students clear, comparable course information so they can choose the degree that suits them, and apply with confidence.\" It names the content (course information), the audience (prospective students) and the outcome (a confident, well-matched choice). It also leaves things out, which is where much of its value lies.\n\n## Why it matters for content\n\nContent strategies can become long documents whose recommendations nobody remembers. A core statement gives a compact reference for everyday choices. Teams can ask whether a proposed page, [content model](https://contentlexicon.com/content-model), [workflow](https://contentlexicon.com/workflow) or campaign supports it.\n\nThe statement should connect audience needs with organisational purpose. A promise of \"engaging content\" for \"everyone\" makes no choices. Useful strategy excludes things and gives people a basis for saying no.\n\nThat's how the statement earns its place. Suppose the marketing team asks the university's web team for a weekly series on campus life, and a department wants its staff news on the course pages. Neither helps a prospective student compare courses, so both fall outside the statement. The web team can decline, or point the requests to the channel that owns them, by referring to an agreed decision rather than a personal opinion. The same test applies to existing work. A hundred old news stories that serve no one in the statement become candidates for removal.\n\nSaying no works best when the statement has visible support. Leaders who approved it need to back it when a senior request falls outside it, or it becomes a slogan people quote and ignore.\n\nThe surrounding strategy still matters. An annotated core statement can show the research and reasoning behind each phrase, while governance, structure, operations and measurement explain how it'll be realised.\n\n## How AI changes the picture\n\nAI makes production easier, which increases the need for strategic restraint. A core strategy gives people and assistants a basis for comparing proposals. An assistant could flag a proposed article that doesn't serve the stated student decision, then ask the team whether the strategy or the proposal needs revisiting.\n\nThe statement can guide [prompts](https://contentlexicon.com/prompt), briefs and evaluation criteria. It isn't a magic sentence: systems also need source material, specific instructions and human judgement.\n\nAI may open up new channels and formats, but the core purpose can stay stable. A strong strategy directs adaptation without tying the organisation to one platform or tool.\n\n## Questions people ask\n\n**How long should it be?**\n\nUsually one sentence, supported by annotations where necessary. Brevity should sharpen choices, not remove substance.\n\n**Is it the whole content strategy?**\n\nNo. It's the compact centre. The wider strategy covers evidence, content models, governance, operations, channels and measures.\n\n**Who is it for?**\n\nEveryone making or approving content, including leaders, subject experts, designers, [writers](https://contentlexicon.com/writer) and technical teams.\n\n**How often should it change?**\n\nReview it when audience needs, organisational purpose or the content environment change significantly, not whenever a new tool appears.\n\n## Learn more\n\n- [Keep your content on strategy with a single statement](https://contentmarketinginstitute.com/content-marketing-strategy/keep-your-content-on-strategy-with-this-single-statement-templates) – Meghan Casey's practical template and examples for making real choices in one sentence.\n- [Core Strategy vs. Strategic Priorities: Which Is Right for You?](https://www.braintraffic.com/blog/core-strategy-vs-strategic-priorities-which-is-right-for-you) – Kristina Halvorson on what a core strategy statement covers, with examples, and when a set of strategic priorities works better.\n- [Content transformation, not digital transformation](https://contentious.ltd/thoughts/content-transformation-not-digital-transformation) – an argument for making content, not technology alone, the organising force behind meaningful change." }, { "term": "Create once, publish everywhere", "alsoCalled": [ "COPE" ], "slug": "create-once-publish-everywhere", "definition": "The idea that content should be created once in a structured form and reused across channels instead of rewritten each time.", "category": "Structure and systems", "related": [ "atomic-content", "channel-strategy", "dry", "headless-cms", "separation-of-concerns", "single-source-of-truth", "structured-content" ], "updated": "2026-10-10T17:07:00.000Z", "url": "https://contentlexicon.com/create-once-publish-everywhere", "markdown": "https://contentlexicon.com/create-once-publish-everywhere.md", "credit": "Create once, publish everywhere, from the Content Lexicon by Contentious (https://contentlexicon.com/create-once-publish-everywhere), CC BY 4.0", "body": "Create once, publish everywhere is often shortened to COPE. The useful version is not “write once, paste everywhere”. Different contexts may need different lengths, formats, introductions or [calls to action](https://contentlexicon.com/call-to-action).\n\nCOPE separates the durable substance from its presentation, so a reliable source can support appropriate versions without spawning unrelated copies.\n\n## Why it matters for content\n\nRecreating the same information across web pages, apps, [emails](https://contentlexicon.com/email) and documents causes duplication. Each copy costs time to produce and becomes another place where facts can drift. A changed phone number or eligibility rule must then be found and repaired everywhere, often by people who don't know where “everywhere” is.\n\nStructured reuse can improve consistency and maintenance. A product description, event, office address or piece of guidance can be stored once with clear fields, then presented in ways that suit each channel. The content remains connected to its source even when its appearance changes.\n\nCOPE also affects [author experience](https://contentlexicon.com/author-experience). [Content models](https://contentlexicon.com/content-model) and publishing tools need to reflect meaningful units rather than forcing everything into a page-shaped rich-text field. Too much structure makes writing laborious; too little makes reuse unreliable. Structure information where consistency, filtering or reuse creates real value; there's no need to make every sentence a separate [component](https://contentlexicon.com/component).\n\n## How AI changes the picture\n\nAI makes it easier to adapt source content for different channels. A model can shorten, reframe, translate or change format while preserving a central set of facts. This extends COPE beyond identical reuse towards controlled transformation, provided the canonical source is clear and the transformation has rules.\n\nWithout that discipline, AI can multiply divergence. Ten generated versions may begin from different documents, introduce unsupported claims or change meaning to fit a [tone](https://contentlexicon.com/tone). Teams need to separate protected facts from adaptable expression, record where versions came from and decide which changes require review.\n\n[Structured content](https://contentlexicon.com/structured-content) is also easier for AI systems to retrieve and use. Fields, [metadata](https://contentlexicon.com/metadata) and relationships help distinguish a title from a date or an instruction from an example. COPE therefore supports both publishing and AI-era knowledge use: the same well-managed source can feed pages, services, assistants and future channels that have not yet acquired a name or a launch keynote.\n\n## Questions people ask\n\n**Does COPE mean identical content on every channel?**\n\nCOPE means reusing a reliable source. Presentation and emphasis can change to fit the context, provided the underlying meaning and facts remain governed.\n\n**Is COPE only for large publishers?**\n\nSmall organisations also repeat addresses, biographies, service details and event information. Starting with one frequently duplicated [content type](https://contentlexicon.com/content-type) can be worthwhile.\n\n**Do we need a headless CMS?**\n\nNot necessarily. A suitable content model and clear source can support reuse in many systems. Technology helps, but the first questions concern content, governance and channels.\n\n**What content should not be reused automatically?**\n\nMaterial whose meaning depends heavily on context, audience or moment may need fresh judgement. Sensitive messages and persuasive copy are not interchangeable fields merely because software can move them.\n\n## Learn more\n\n- [Clean Content = Portable Content](https://www.npr.org/sections/inside/2009/02/clean_content_portable_content.html) – Daniel Jacobson's February 2009 explanation of NPR's COPE principles: separate content from display and avoid platform-specific markup.\n- [Understanding Create Once Publish Everywhere (COPE)](https://www.lullabot.com/articles/understanding-create-once-publish-everywhere-cope) – a practical account of COPE and structured content.\n- [Adapting Ourselves to Adaptive Content](https://karenmcgrane.com/talks/adapting-ourselves-to-adaptive-content/) – Karen McGrane's talk transcript on structured content for many devices, using NPR's COPE approach and API as the central example." }, { "term": "Database", "alsoCalled": [], "slug": "database", "definition": "A structured place to store information so it can be queried, related and reused.", "category": "Structure and systems", "related": [ "backend", "content-inventory", "metadata", "schema", "structured-content" ], "updated": "2026-10-11T11:34:00.000Z", "url": "https://contentlexicon.com/database", "markdown": "https://contentlexicon.com/database.md", "credit": "Database, from the Content Lexicon by Contentious (https://contentlexicon.com/database), CC BY 4.0", "body": "A database stores information as records, one for each thing it keeps track of, such as an article, a course or a person. Each record holds the same details in named fields. Because every record follows the same pattern, you can ask questions of the whole collection at once, such as which descriptions haven't been reviewed this year or which courses a particular provider runs.\n\nRecords can also point to one another. Each course can link to its provider's record instead of repeating the provider's details, so a correction made once reaches every page that draws on it.\n\nThat's the difference from a folder of documents, where the details sit inside each file and someone has to open them one by one to find out who owns what. A spreadsheet comes closer: each row can be a record and each column a field. But it will accept anything in any cell, and connecting a row to one on another sheet takes formulas that somebody has to keep working.\n\nA database can insist that a status comes from an agreed list, that a date is a date and that every course has a provider. Notion and Airtable work this way, and so does the storage behind most [content management systems](https://contentlexicon.com/cms) (CMSs). In each case the structure matters more than the product: which details get their own field, how records relate and what people need to find or change.\n\n## Why it matters for content\n\nStoring content as records changes the work of writing it. Instead of composing a page, an author fills in fields: a title of a set length, a summary, a body, a start date, a link to the right provider. The same course description can then appear on the course's own page, in search results, in a printed prospectus and in a partner's listings, and one update to the record changes all of them. Audits get easier too. A question like \"what hasn't been reviewed in two years?\" becomes a filter rather than a week of opening pages.\n\nThe cost is that [writers](https://contentlexicon.com/writer) can lose sight of the finished thing. A form full of fields shows none of the pages it will feed, so a summary that reads well on its own may sit badly beside an image or get cut off on a phone. Good systems give authors previews and guidance for each field.\n\nThe structure also hardens: once templates, feeds and reports depend on a field, renaming or splitting it means changing all of them. That's why decisions about fields and relationships belong in content modelling, with content people in the room, rather than being left to whoever builds the system. And a field nobody fills in, or fills in inconsistently, is worse than having no field, because people and systems will trust it anyway.\n\n## How AI changes the picture\n\nStructured records let an assistant give exact answers to some questions. Asked which courses start in March, a system that can query a start-date field returns a list; one that has to read course pages is guessing from whichever passages it finds. Fields for status, owner and [review date](https://contentlexicon.com/review-date) can also keep withdrawn or expired material out of an answer, provided the retrieval is set up to use them. None of this makes a stored claim true. An out-of-date description in a well-labelled field is still out of date.\n\nFor questions that don't map onto fields, systems increasingly search by meaning. A vector database, or a vector index added to an ordinary database, stores each passage as a long list of numbers, called an embedding, that places it near passages with similar meaning.\n\nGraph databases, which store connections between people, topics and organisations, can help a system follow relationships that the text only implies.\n\nAI can also help with the slowest part of building a content database: getting existing material into it. A model can propose summaries, tags and [content types](https://contentlexicon.com/content-type) across thousands of legacy pages far faster than a person could, flag records where the fields contradict the body and spot a field that's doing two jobs. People still need to check its proposals, and reviewing a sample shows how far to trust each kind. But a structure that once needed months of manual migration can now be filled well enough to start using, then improved as it's used.\n\n## Questions people ask\n\n**Where do we start?**\n\nWith content you already maintain over and over, such as events, courses, products or policies. Pick one type and list the details people look up, display or change. Those are your candidate fields. Add a relationship only where a real change needs to travel, such as a provider's new address reaching every course. Then enter twenty real items before committing: gaps and awkward fields show up fast.\n\n**How many fields should a content type have?**\n\nAs many as someone uses to find, display, check or reuse the content, and no more. Each field needs a person who fills it in and a reason someone or something reads it. Splitting a body into many small fields makes reuse easier but writing feel like form-filling; one large body field is pleasant to write in and hard to reuse. Writing three real items is the quickest way to find the balance.\n\n**Do content people need to understand technical databases?**\n\nThey don't need to become database engineers. They need to understand enough structure to ask better questions: what are the content types, what fields matter, how do items relate, what needs to be reusable and which source is authoritative?\n\n**How does a database relate to a CMS?**\n\nA CMS often uses databases behind the scenes to store content and [metadata](https://contentlexicon.com/metadata). [Content teams](https://contentlexicon.com/content-team) may never touch that database directly, but their [content model](https://contentlexicon.com/content-model), fields and [workflows](https://contentlexicon.com/workflow) determine whether the system is useful or awkward.\n\n## Learn more\n\n- [Content Modelling: A Master Skill](https://alistapart.com/article/content-modelling-a-master-skill/) – Rachel Lovinger on documenting content types, their attributes and relationships, and keeping that model as a living reference shared by designers, developers and authors.\n- [Embeddings: What they are and why they matter](https://simonwillison.net/2023/Oct/23/embeddings/) – Simon Willison's explainer on finding content by meaning, built around a related-posts feature for his own site; he argues an existing database with a vector index often does the job without a specialist vector database.\n- [Database as a Symbolic Form](https://manovich.net/index.php/projects/database-as-a-symbolic-form) – Lev Manovich’s 1998 essay on how collections, interfaces and narrative shape different ways of encountering material." }, { "term": "Decentralised governance", "alsoCalled": [ "Decentralised model", "Distributed publishing", "Dispersed model" ], "slug": "decentralised-governance", "definition": "A governance model in which separate parts of an organisation plan, create and publish their own content, with little central control.", "category": "Governance", "related": [ "author-experience", "centralised-governance", "content-audit", "content-governance", "content-owner", "content-team", "hub-and-spoke", "rot-content", "silo", "single-source-of-truth", "style-guide" ], "updated": "2026-10-14T12:03:00.000Z", "url": "https://contentlexicon.com/decentralised-governance", "markdown": "https://contentlexicon.com/decentralised-governance.md", "credit": "Decentralised governance, from the Content Lexicon by Contentious (https://contentlexicon.com/decentralised-governance), CC BY 4.0", "body": "Decentralised governance spreads the authority to publish across an organisation. Departments, product teams, regional offices or local [branches](https://contentlexicon.com/branch) decide what content they need, produce it and manage it themselves. Some have content specialists of their own, such as a [content designer](https://contentlexicon.com/content-designer) in each product team or a [writer](https://contentlexicon.com/writer) in each department; in others the work falls to staff who publish alongside another job. Shared standards may exist on paper, but nobody has the authority or the time to apply them across the whole.\n\nIt sits at the opposite end of the spectrum from [centralised governance](https://contentlexicon.com/centralised-governance), with [hub and spoke](https://contentlexicon.com/hub-and-spoke) in between. Few organisations choose it outright. More often it's where they end up as [websites](https://contentlexicon.com/website), intranets and channels multiply faster than anyone can coordinate them, and it's familiar in universities, federations of charities and companies that have grown through acquisitions.\n\n## Why it matters for content\n\nDecentralisation keeps content close to the knowledge it depends on. The people who run a service or study a subject can publish when something changes, without waiting in another team's queue, and they often know their own audience better than a central team could. For organisations with very different products, markets or local communities, that autonomy can be the right choice.\n\nThe costs build up across the estate rather than on any one page. [Voice](https://contentlexicon.com/voice) and quality drift apart, pages on the same topic contradict one another, navigation follows the organisation chart instead of what readers are looking for and nobody can say how much has been published or what's out of date. Mel Conway observed in 1968 that organisations produce systems that copy their own communication structures, and a decentralised website is a vivid example: a reader who wants to make a complaint first has to work out which department handles complaints. Embedded specialists face their own version of the problem. Without peers, review or a manager who understands their craft, they can end up defending standards alone, with no clear route to progress.\n\nDecentralisation works best when it's chosen deliberately and given some shared infrastructure: a common platform, a short set of standards, named owners and a community where publishers compare notes. At that point it starts to resemble hub and spoke, which is where many organisations move once the costs become visible.\n\n## How AI changes the picture\n\nGood AI infrastructure can narrow the gap between well-resourced teams and stretched ones. A local team can draft, check [readability](https://contentlexicon.com/readability) and translate without calling on specialists it doesn't have, and shared guidance such as a [style guide](https://contentlexicon.com/style-guide) or [content principles](https://contentlexicon.com/content-principles) can be built into the tools each team uses, so standards travel with the work rather than sitting in a document few people open.\n\nAI can also give a decentralised organisation something it has usually lacked, which is a view of the whole. Comparing thousands of pages to find duplicates, contradictions and abandoned sections once needed a central team with months to spare. Language models can now produce a first pass much faster, for people to check. That doesn't settle who should fix what, but it makes the cost of sprawl harder to ignore. The same tools speed up the sprawl itself. When more teams can publish more material faster, duplication grows with the output, and chatbots or AI search drawing on that material may serve readers whichever version they find first.\n\n## Questions people ask\n\n**Does decentralised mean ungoverned?**\n\nNot necessarily. Some decentralised organisations have clear owners, shared standards and regular reviews, and leave decisions about what to publish to the teams themselves. The trouble starts when decentralisation is a default rather than a decision, and nobody is responsible for the whole.\n\n**Who should embedded content specialists report to?**\n\nReporting to a team they serve keeps them close to the work but can leave them without anyone who understands their craft. Some organisations keep the reporting line within a content or design discipline while placing people in product teams for long stretches. Either way, a community of practice helps prevent isolation.\n\n**How do you know decentralisation has gone too far?**\n\nSeveral pages answer the same question differently, readers and staff aren't sure which site is official, sections belong to people who left years ago and nobody can say how many pages exist. A [content audit](https://contentlexicon.com/content-audit) usually confirms the suspicion quickly.\n\n**Can a decentralised organisation have one voice?**\n\nIt can share one, but it takes work: a clear style guide, examples people can borrow, training for occasional publishers and someone who reads a sample of what's published and talks to teams about it. Some variation in [tone](https://contentlexicon.com/tone) between, say, a research unit and a fundraising team is appropriate, provided the voice underneath stays recognisable.\n\n## Learn more\n\n- [Organising structures for content teams: a guide](https://lapope.com/resource/organising-structures-for-content-teams-a-guide) – Lauren Pope's comparison of five structures, rated for control, scalability and agility, which separates decentralised teams of content professionals from federated models that devolve publishing to non-specialists.\n- [How Do Committees Invent?](https://www.melconway.com/research/committees.html) – Melvin Conway's 1968 paper, the origin of Conway's law; written about the design of systems such as software, its argument applies readily to websites built department by department.\n- [Managed-UX Integration: A Team Model for UX Autonomy and Product Alignment](https://www.nngroup.com/articles/managed-ux/) – Anna Kaley's proposal for embedding designers in product teams while they report to design leadership, aimed at the isolation and missing community common in decentralised models; written for UX, it transfers directly to content design." }, { "term": "Decision models", "alsoCalled": [], "slug": "system-one-models", "definition": "An AI model built to decide rather than write: it picks from answers you've defined and says how likely each one is.", "category": "AI and tooling", "related": [ "agentic-ai", "automated-testing", "guardrails", "hallucination", "human-in-the-loop", "llm", "structured-content", "taxonomy" ], "updated": "2026-10-06T21:15:00.000Z", "url": "https://contentlexicon.com/system-one-models", "markdown": "https://contentlexicon.com/system-one-models.md", "credit": "Decision models, from the Content Lexicon by Contentious (https://contentlexicon.com/system-one-models), CC BY 4.0", "body": "A decision model receives some content, such as a support ticket or photo caption, and questions with defined answer options. It returns bounded results: a category with probabilities, a score or a yes-or-no likelihood.\n\nTypeSafe AI calls these 'System One models', a nod to fast, instinctive human thinking. The comparison is a metaphor, not evidence that a model thinks like a person.\n\n'Decision model' has an older and broader life in operations research and decision theory, where it means any formal representation of a decision problem. The modern version is an AI model that returns bounded, typed decisions – a category, a score or a yes-or-no probability – rather than generated prose.\n\n## Why it matters for content\n\nMany [content operations](https://contentlexicon.com/content-operations) depend on repeated judgements with known answers: which queue should receive a support message, whether a passage breaches a policy, how relevant a document is to a query or which [taxonomy](https://contentlexicon.com/taxonomy) term fits an item. These jobs don't always need a paragraph from an [LLM](https://contentlexicon.com/llm). They need a value that software can use without first persuading a free-form answer to behave like data.\n\nThat makes the content around the decision unusually important. Someone has to define the available options, write questions that draw meaningful distinctions and choose which part of the source material the model sees. A muddled taxonomy or vague quality scale will still produce tidy numbers.\n\nThe probabilities also create a practical route for review. A system might route high-confidence decisions automatically, send uncertain ones to a person and record disagreements for evaluation. However, teams still need to test whether the confidence scores match real outcomes; a number with two decimal places can look more settled than it deserves.\n\n## How decision models change the picture\n\nDecision models separate judgement from generation. An LLM can still draft a reply or summary, while a decision model handles bounded tasks such as classification, scoring, routing or checking. In an agentic system, that could mean using one model to choose the next permitted action, code to enforce the rules and an LLM to write whatever a person will read.\n\nFor [content teams](https://contentlexicon.com/content-team), this gives [structured content](https://contentlexicon.com/structured-content) another job. Take a charity case study stored as separate fields: the story, a pull quote, the photo credit and a consent record. A check on whether the quote is cleared for use needs the quote and the consent record, and nothing else. Whether you can send just those two depends on decisions made long before anyone bought an AI tool: how the [content model](https://contentlexicon.com/content-model) splits things up, and whether consent was recorded as data or left in someone's inbox.\n\nThe fixed output removes one familiar failure: the model cannot invent a category outside the options or return malformed prose where code expects a number. It can still misread the content, choose a perfectly well-formed wrong answer and attach a high probability to it. Format guarantees make systems easier to build; they don't guarantee to supply truth.\n\n## Questions people ask\n\n**Is a decision model an LLM?**\n\nIt can be built from one. Cloudflare describes its Clef models as using Qwen language models to interpret the input, then scoring the allowed answers in parallel rather than generating text token by token. The useful distinction is what the model is built to return: bounded decisions rather than open-ended language.\n\n**Is this just a classification model with a grander name?**\n\nThere's considerable overlap. Classification is one of its main jobs, but the proposed category also covers scoring and yes-or-no probabilities, with answer options supplied at the time of the request. These [APIs](https://contentlexicon.com/api) can ask for choices, scores and yes-or-no probabilities in the same request. Cloudflare describes its API as compatible with Jev. Similar interfaces don't yet establish a settled category. Whether that deserves a lasting new category will become clearer as more models and independent evidence appear.\n\n**Can a decision model hallucinate?**\n\nWith no text to write, it can't make up a quote, a source or a category that wasn't on the list. It can still get the answer wrong: a support message sent to the wrong queue with 94% confidence is still in the wrong queue.\n\n**When should a content team use one?**\n\nUse one for a decision you make often, where you can write down every acceptable answer in advance and software needs to act on the result quickly. If the job needs words – an explanation, a draft, a reason someone can read – it's a job for an LLM or a person.\n\n## Learn more\n\n- [Distilling System 2 into System 1](https://arxiv.org/html/2407.06023v3) – research using “System 1” in the broader sense of producing an answer directly, which helps distinguish the existing academic usage from the newer category.\n- [Introducing System One Models & Jev](https://typesafe.ai/blog/introducing-system-one-models-and-jev) – TypeSafe's 2026 announcement of the category and its first model, with the company's technical claims, examples and caveats.\n- [Cloudflare Clef decision models](https://blog.cloudflare.com/clef-decision-models/) – Cloudflare's 2026 announcement of open-source Clef and Clef-flash, describing their parallel-scoring approach and its RL fine-tuning platform.\n- [Perplexity Decisions API](https://docs.perplexity.ai/docs/decisions/quickstart) – product documentation for the three question types, with worked examples using pplx-decider-v1-27b." }, { "term": "Design system", "alsoCalled": [], "slug": "design-system", "definition": "A shared, maintained collection of components, patterns and guidance that teams use to design and build digital products and services.", "category": "Structure and systems", "related": [ "component", "content-pattern", "frontend", "separation-of-concerns", "style-guide" ], "updated": "2026-10-11T06:24:00.000Z", "url": "https://contentlexicon.com/design-system", "markdown": "https://contentlexicon.com/design-system.md", "credit": "Design system, from the Content Lexicon by Contentious (https://contentlexicon.com/design-system), CC BY 4.0", "body": "A design system brings together the parts of a digital product – visual styles, interface [components](https://contentlexicon.com/component), [content patterns](https://contentlexicon.com/content-pattern) and code – with the guidance that explains when and how to use them. Its value lies less in the parts than in the decisions they carry: why an alert looks the way it does, what it should say and when it shouldn't be used at all. Without that reasoning, a component library is a well-stocked cupboard that everyone raids differently.\n\nOrganisations have codified how they present themselves for decades. The New York City Transit Authority's 1970 graphics standards manual, produced by Massimo Vignelli and Bob Noorda at Unimark, set out how the subway's signs should look and included an approved list of station names, so content was in the system from the start. Web teams later built [style guides](https://contentlexicon.com/style-guide) and pattern libraries, and the modern design system took shape in the 2010s, when the parts became working code and the button a designer drew could be the same button a developer shipped.\n\n## Why it matters for content\n\nContent is often the missing layer. Plenty of systems specify a button's padding to the pixel and say nothing about what the button should say. Content specialists can define how components behave with real language, different lengths, errors and translations rather than squeezing copy into boxes designed around [lorem ipsum](https://contentlexicon.com/lorem-ipsum). An error component needs more than red styling; it needs rules for naming the problem and helping someone recover. [Voice](https://contentlexicon.com/voice) and [tone](https://contentlexicon.com/tone) guidance, a [word list](https://contentlexicon.com/word-list) of product names and component-level content rules mean that the many people who put words in front of users, from product managers to lawyers, start from the same decisions.\n\nThe system's documentation is also content, with readers of its own. Designers and developers consult it in a hurry, usually halfway through a task. If the guidance is vague, buried or out of date, they guess, and consistency erodes one workaround at a time. Writing, structuring and maintaining that documentation is content work, and it often decides whether a system is adopted or admired and ignored.\n\nDesign systems multiply whatever they contain. Improve the wording of an error component and the improvement reaches every product that uses it; get it wrong and the mistake ships everywhere too. Nathan Curtis argued in 2016 that a design system is a product serving other products, with users, a roadmap and someone accountable for it, rather than a project that ends at launch. Contributions, ownership, [versioning](https://contentlexicon.com/versioning) and deprecation prevent the library from becoming a museum of almost-identical buttons.\n\n## How AI changes the picture\n\nDesign systems are acquiring a second audience. Coding assistants and design tools increasingly read a system directly, through its code, its documentation and [machine-readable](https://contentlexicon.com/machine-readable) formats such as design tokens, which record decisions about colour, type and spacing in a form different tools can share. A W3C community group published the first stable specification for them in 2025. When an assistant builds an interface from approved components, the system marks the boundaries of what it can produce.\n\nThat puts unusual weight on the guidance. A person reading \"use sparingly\" can ask a colleague what it means; a model will settle on its own interpretation. Rules about when not to use a component, what its content must include and which combinations don't work need to be explicit and illustrated, because an assistant can assemble technically valid combinations that make no sense together. A team could generate trial empty-state messages against an approved pattern, then compare their meaning in the actual screen.\n\nAI can also take on some of the upkeep: drafting documentation from code and design files, spotting where it has drifted from what's shipped and presenting the same guidance differently to a junior designer and a senior developer. Generated documentation still needs review by people who know why the decisions were made. Lose the reasoning and what remains is a parts catalogue.\n\n## Questions people ask\n\n**Is a style guide a design system?**\n\nNot on its own, though it may form part of one. A style guide sets conventions for language and presentation; a design system connects those conventions to the components, patterns and code that teams use directly.\n\n**How does a design system relate to components and patterns?**\n\nComponents are the reusable parts, such as buttons, alerts and cards. Patterns describe how to combine and adapt them for recurring tasks. The design system holds both, along with the principles, standards and documentation that explain them and the process for changing them.\n\n**Does using a design system guarantee accessibility?**\n\nAccessible components help, but their combination, content and implementation still need testing. A well-built button doesn't make an unclear or inaccessible booking journey work.\n\n**Do smaller organisations need one?**\n\nNot a large one. A charity with a single [website](https://contentlexicon.com/website) might need a short, maintained record of its components, key content patterns and word list, kept close to the code. If people keep solving the same problems in different ways, even a modest system will save time and confusion.\n\n## Learn more\n\n- [A Design System isn't a Project. It's a Product, Serving Products.](https://medium.com/eightshapes-llc/a-design-system-isn-t-a-project-it-s-a-product-serving-products-74dcfffef935) – Nathan Curtis's essay arguing that launch is where the work begins, and that a system needs product management, funding and a roadmap to survive.\n- [Content in, for, and of Design Systems](https://indeed.design/article/content-in-for-and-of-design-systems) – Michael Lawrence and Michael Haggerty-Villa's piece for Indeed Design, separating three kinds of content work: the system's own documentation, the content standards it sets for everyone else and the story that persuades people to use it.\n- [AI and Design Systems](https://bradfrost.com/blog/post/ai-and-design-systems/) – Brad Frost's 2024 account of how his team uses AI for component code, testing, accessibility review and documentation, including documentation that reshapes itself for each reader." }, { "term": "Design thinking", "alsoCalled": [], "slug": "design-thinking", "definition": "A problem-solving approach that uses empathy, exploration, prototyping and testing to create better services, products or content.", "category": "Research and evidence", "related": [ "agile", "card-sorting", "user-needs", "user-testing" ], "updated": "2026-10-07T12:54:00.000Z", "url": "https://contentlexicon.com/design-thinking", "markdown": "https://contentlexicon.com/design-thinking.md", "credit": "Design thinking, from the Content Lexicon by Contentious (https://contentlexicon.com/design-thinking), CC BY 4.0", "body": "Design thinking is often described through stages such as empathise, define, ideate, prototype and test, though real projects tend to move backwards, sideways and around again.\n\nFor content work, its value lies in replacing assumption with evidence and debate with something people can try. A team can explore several ways to explain a service, put rough versions in front of users and learn before committing to a finished design or a small mountain of copy.\n\n## Why it matters for content\n\nOrganisations often arrive with a requested output: a new [website](https://contentlexicon.com/website), campaign, guide or chatbot. Design thinking encourages the team to ask what problem sits underneath the request and whose problem it is. The answer may reveal that people need a clearer process, a timely message or one trustworthy source rather than more content.\n\nPrototyping is especially useful. Content can be tested in sketches, paper journeys, draft messages and simple interactive flows. These need not look finished. Their job is to reveal whether people understand the language, find the right information and know what to do next.\n\nThe approach also gives different disciplines a shared way to learn. Researchers bring evidence, content specialists shape meaning, designers explore interactions and subject experts test accuracy. It works best when participants can change the outcome, rather than attending a workshop-shaped interval before the original decision resumes unchanged.\n\n## How AI changes the picture\n\nAI can accelerate exploration by generating variations, summarising research and turning rough ideas into prototypes. Teams can test more possibilities and create realistic edge cases earlier. This is useful when speed serves learning rather than merely helping the first idea acquire a glossy surface.\n\nGenerative tools can also reproduce assumptions in their [prompts](https://contentlexicon.com/prompt) and training data. An AI-created [persona](https://contentlexicon.com/persona) or journey isn't user research. It may help a team formulate questions, but it can't provide evidence about what real people need, experience or understand.\n\nAI products themselves benefit from design thinking because their behaviour is uncertain and shaped by context. Teams need to prototype failed retrieval, unclear requests, correction, escalation and consent alongside the successful cases. Human-centred principles remain essential when an interface can produce something nobody explicitly designed.\n\n## Questions people ask\n\n**Must the stages happen in order?**\n\nThe stages describe ways of working, not a ceremonial sequence. Teams move between understanding, making and testing as they learn.\n\n**Is it the same as user-centred design?**\n\nThey overlap. Human-centred or user-centred design expresses a commitment to people’s needs, while design thinking describes methods and mindsets for applying that commitment to problems.\n\n**Do content teams need to prototype?**\n\nYes, when structure, language or interaction is uncertain. A rough content prototype often reveals more than discussing an abstract requirement for another hour.\n\n**Can AI replace user research?**\n\nNo. It can organise evidence or help prepare research, but generated users can't tell you what actual people do, need or misunderstand.\n\n## Learn more\n\n- [Design Thinking 101](https://www.nngroup.com/articles/design-thinking/) – Nielsen Norman Group's six iterative phases, including implementation, rather than a fixed five-step sequence.\n- [Design thinking and human-centred design](https://designthinking.ideo.com/introduction) – IDEO's distinction between a people-first orientation and the methods for applying it.\n- [Principles of human-centred design](https://www.nngroup.com/videos/principles-human-centered-design-don-norman/) – Don Norman on keeping people and the wider system in view." }, { "term": "Devops", "alsoCalled": [], "slug": "devops", "definition": "Shared practices that bring software development and operations together to deliver and maintain systems reliably.", "category": "AI and tooling", "related": [ "agile", "automated-testing", "branch", "content-operating-model", "content-operations", "continuous-integration", "feature-flag", "git", "merge-conflict", "pull-request", "refactoring", "rollback", "staging", "technical-debt", "workflow" ], "updated": "2026-10-11T13:22:00.000Z", "url": "https://contentlexicon.com/devops", "markdown": "https://contentlexicon.com/devops.md", "credit": "Devops, from the Content Lexicon by Contentious (https://contentlexicon.com/devops), CC BY 4.0", "body": "Devops is a way of working in which the people who build software share responsibility for how it runs once it's live. It combines a culture of shared ownership with practices such as version control, [automated testing](https://contentlexicon.com/automated-testing), [continuous integration](https://contentlexicon.com/continuous-integration), monitoring and fast recovery.\n\nIt grew out of a familiar problem. Developers wrote code and handed it over the wall to an operations team, who had to deploy and support something they hadn't helped design. Developers were rewarded for change and operations for stability, so each side blamed the other when a release failed. The name comes from DevOpsDays, the conference Patrick Debois organised in Ghent in 2009.\n\n## Why it matters for content\n\nContent has its own wall. [Writers](https://contentlexicon.com/writer) and subject experts create material and hand it over to be published, and then responsibility fades. Nobody owns how the page performs, whether it still matches the service or who updates it when a policy changes. The devops answer is shared responsibility for what happens after publishing, so the people who create content also see how it's used, how it fails and how it's maintained. In practice, the writer of a help article might see its search terms and feedback, and join the review when the support team reports that it's causing calls.\n\nMany people and tools now create, review, translate, publish and update connected material. Yet [workflows](https://contentlexicon.com/workflow) often rely on documents, memory and favours rather than visible changes and repeatable checks. Devops offers habits that transfer: make small changes, keep versions, review differences, automate routine tests, watch what happens after release and learn from failures without blame. Applied thoughtfully, these habits improve [content operations](https://contentlexicon.com/content-operations) without asking writers to become software engineers.\n\nThe comparison has limits. Content involves judgement, meaning and cultural context that tests can't capture. Borrow the discipline where it helps and keep people responsible for editorial decisions.\n\n## How AI changes the picture\n\nFor AI-assisted content, these practices become more relevant. [Prompts](https://contentlexicon.com/prompt), knowledge sources and evaluation cases can be versioned, tested before release, monitored afterwards and rolled back when quality falls.\n\nContinuous integration can check links and required fields when content changes. Automated [accessibility](https://contentlexicon.com/accessibility) checks cover particular failures, not the whole experience. A documentation team could also test that a changed setup instruction still matches the current product, then investigate failures before releasing it.\n\nThe deeper change is cultural. Content, design, technology and operations share ownership of how the system behaves, rather than passing work downstream and hoping for the best.\n\n## Questions people ask\n\n**Is devops a job or a practice?**\n\nIt appears in job titles, but it's mainly a set of shared practices and responsibilities across teams.\n\n**What can content teams borrow first?**\n\nStart by keeping versions and reviewing visible differences before publication. For a small team, that might mean checking one changed instruction against its source and recording who approved it. Then agree who watches each important page after it goes live. Add [staging](https://contentlexicon.com/staging) or automated tests when the work warrants them.\n\n**Can content quality be automated?**\n\nParts can be checked automatically. Accuracy, usefulness, ethics and [tone](https://contentlexicon.com/tone) still need evidence and judgement.\n\n**Do we need developer tools?**\n\nNot always. Start with the practices and choose tools that fit the team's capability and risk.\n\n## Learn more\n\n- [DORA's software delivery performance metrics](https://dora.dev/guides/dora-metrics/) – the evolving measures of software delivery, with warnings against gaming targets and comparing unlike services; not a ready-made score for editorial quality.\n- [The Three Ways: The Principles Underpinning DevOps](https://itrevolution.com/articles/the-three-ways-principles-underpinning-devops/) – Gene Kim's summary of the principles behind devops: flow across the whole system, fast feedback and a culture of experimentation and learning.\n- [Continuous Integration](https://martinfowler.com/articles/continuousIntegration.html) – Martin Fowler's detailed guide to the small, frequent changes that make the wider approach possible." }, { "term": "Domain model", "alsoCalled": [], "slug": "domain-model", "definition": "A representation of the important concepts, relationships and rules in a defined area of knowledge or activity.", "category": "Structure and systems", "related": [ "content-model", "structured-content", "ubiquitous-language" ], "updated": "2026-10-20T15:43:00.000Z", "url": "https://contentlexicon.com/domain-model", "markdown": "https://contentlexicon.com/domain-model.md", "credit": "Domain model, from the Content Lexicon by Contentious (https://contentlexicon.com/domain-model), CC BY 4.0", "body": "A domain model describes the important concepts in an area of knowledge or activity, how they relate and which rules shape those relationships. It makes a team's understanding explicit enough to discuss and test. In content work, it's often a conceptual diagram with short definitions; in software design, the term can also refer to an implemented model that combines data and behaviour.\n\nThe model is selective, not an attempt to catalogue everything. Its scope follows the subject and purpose, rather than the organisation chart or the pages on an existing [website](https://contentlexicon.com/website). A model of a lending service, for example, might distinguish an item, a borrower and a loan, including the rule that one borrower can have several loans. The concepts and their names develop together through [Ubiquitous language](https://contentlexicon.com/ubiquitous-language): agreeing a label can reveal that people haven't agreed what the thing is.\n\n## Why it matters for content\n\nA domain model gives [content teams](https://contentlexicon.com/content-team) a way to understand their subject before deciding how to publish it. Separating a loan from the item being borrowed helps a team see that an item's description and a borrower's return deadline are different kinds of information. The distinction can inform public explanations, navigation and the records that supply them. Starting with existing pages can hide it, because several concepts may already be bundled into one page.\n\nA [Content model](https://contentlexicon.com/content-model) translates relevant parts of that understanding into publishable structures: [content types](https://contentlexicon.com/content-type), fields and relationships. The translation involves editorial choices. Some domain concepts need their own reusable records; others need only a field, an explanation or no published representation at all. A guidance article may explain several concepts without being a concept in the domain itself. There needn't be one page, or even one content type, for every box in the diagram.\n\nThe Content Lexicon offers a related example. Its entries define concepts across content work, while links between terms make some connections visible. In that sense, it can be read as a partial, evolving model of the field – expressed through definitions and links rather than a formal diagram. It doesn't attempt to map every concept, relationship or rule.\n\nModelling also exposes disagreements before a publishing system gives them permanent homes. Domain experts can test the model against real cases, while research with users can reveal distinctions or language the organisation has overlooked. Different services may need different models of the same subject. Their boundaries and connections matter more than forcing everything into a single organisation-wide diagram.\n\n## How AI changes the picture\n\nAn AI assistant can help turn interviews and source documents into a provisional set of concepts, suggest relationships and identify apparently conflicting definitions. That can give a team something concrete to challenge earlier in the work. Its suggestions still need testing against the domain: a plausible relationship isn't evidence that the organisation operates that way, and an unusual case may reveal a missing concept rather than an error to tidy away.\n\nOnce agreed definitions and relationships are represented in usable source material or connected records, they can help an assistant interpret questions and find relevant information. It could distinguish a borrowing limit from a return deadline, for example, instead of treating everything about loans as interchangeable. The diagram alone doesn't create that capability; the system needs access to the definitions, relevant records and a way to use their relationships. Where rules determine eligibility or other consequential decisions, implementing and validating those rules matters as well as explaining them.\n\n## Questions people ask\n\n**Does “domain” mean a website address?**\n\nHere it means a subject or area of activity, such as lending, conservation or recruitment. A domain model can serve several websites, services and internal systems.\n\n**Is a domain model the same as a taxonomy?**\n\nA [taxonomy](https://contentlexicon.com/taxonomy) classifies things. A domain model can express wider relationships, such as who borrows an item and which loan records that borrowing. Classification may be part of the model, but doesn't describe all its connections or rules.\n\n**Does it have to become a database?**\n\nIt can remain a shared conceptual model. It may inform a [database](https://contentlexicon.com/database) [schema](https://contentlexicon.com/schema) or software design later, but storage choices aren't its starting point. A database also needs implementation decisions that the conceptual model may leave open.\n\n**How much detail is enough?**\n\nEnough to support the decisions it's being used for. A small service may need a handful of concepts and relationships. More detail earns its place when it resolves an ambiguity, captures a consequential rule or helps explain a real case. Extra boxes don't make a model more accurate.\n\n## Learn more\n\n- [Designing Connected Content – publisher's sample](https://ptgmedia.pearsoncmg.com/images/9780134763385/samplepages/9780134763385_Sample.pdf) – Mike Atherton and Carrie Hane introduce their subject-first approach to publishing. The sample includes an interview with Annette Priest about using domain and content modelling across a complex product portfolio.\n- [Domain Model](https://martinfowler.com/eaaCatalog/domainModel.html) – Martin Fowler's concise definition of the software pattern, where interconnected objects represent the domain and combine behaviour with data. A useful distinction from a purely conceptual diagram.\n- [Bounded Context](https://martinfowler.com/bliki/BoundedContext.html) – Fowler explains why different parts of an organisation may need different models, with explicit relationships between them rather than a single universal account." }, { "term": "DRY", "alsoCalled": [ "Don’t repeat yourself" ], "slug": "dry", "definition": "“Don’t repeat yourself”: keep one reliable source of truth instead of many drifting copies.", "category": "Structure and systems", "related": [ "atomic-content", "create-once-publish-everywhere", "single-source-of-truth", "structured-content", "technical-debt" ], "updated": "2026-10-11T06:04:00.000Z", "url": "https://contentlexicon.com/dry", "markdown": "https://contentlexicon.com/dry.md", "credit": "DRY, from the Content Lexicon by Contentious (https://contentlexicon.com/dry), CC BY 4.0", "body": "DRY stands for \"don't repeat yourself\". In software, it means every piece of knowledge should have a single, authoritative representation, so a change is made in one place. In content, the same principle applies to definitions, policy wording, product copy, service descriptions, statistics, instructions and guidance. If the same fact or passage exists in five places, those five places eventually start disagreeing with one another.\n\nDRY doesn't mean everything should be written only once, in identical words. Sometimes repetition is useful for readers. What needs one authoritative home is the underlying fact, decision or source, even if it's expressed in different ways for different contexts.\n\n## Why it matters for content\n\nDRY reduces drift, contradiction and needless maintenance. It helps teams update information once rather than hunting for every copy, and it makes content easier to govern because there's less ambiguity about which version is the right one. It's especially valuable where the same information appears across channels, teams or formats.\n\nThe facts and assets that change over time need it most: prices, statistics, policy wording, service descriptions, eligibility criteria, boilerplate copy, brand guidance and recurring explanations. They're also the things organisations copy and paste most freely.\n\n## How AI changes the picture\n\nGenerating a new copy of something now costs almost nothing, while maintaining it costs what it always did. Ask an assistant for a fresh eligibility summary for a campaign page and you get a near-duplicate, worded slightly differently, that nobody will remember to update when the criteria change. Retrieval adds a second problem: a chatbot that finds three versions of the same policy may quote the oldest, or blend them into a fourth that appears nowhere.\n\nUsed well, AI supports DRY: it can rework one structured source into a summary, an FAQ answer or a social post, each drawn from the same original. That depends on there being a reliable original to draw from.\n\n## Questions people ask\n\n**Does DRY mean never repeating anything?**\n\nNo. Readers often need reinforcement, summaries or context-specific wording. DRY is about not maintaining several competing sources for the same underlying information.\n\n**Is DRY only for technical teams?**\n\nNot at all. [Content teams](https://contentlexicon.com/content-team) deal with repeated facts, definitions and guidance all the time. In some organisations they deal with it more than developers do.\n\n**Can DRY go too far?**\n\nYes. If everything is forced into one abstract source when readers need tailored wording, the result can become harder to use. DRY should reduce duplication without flattening context.\n\n**Is DRY the same as a single source of truth?**\n\nThey're closely related. DRY is the principle of reducing unnecessary duplication; a [single source of truth](https://contentlexicon.com/single-source-of-truth) is often the practical way of doing it.\n\n## Learn more\n\n- [The Pragmatic Programmer on DRY](https://media.pragprog.com/titles/tpp20/dry.pdf) – the authorised “The Evils of Duplication” excerpt from The Pragmatic Programmer, explaining duplication of knowledge rather than only repeated code.\n- [Avoiding Repetition](https://www.martinfowler.com/ieeeSoftware/repetition.pdf) – Martin Fowler's IEEE Software essay on why repeated knowledge creates maintenance trouble.\n- [The Wrong Abstraction](https://sandimetz.com/blog/2016/1/20/the-wrong-abstraction) – Sandi Metz's counterpoint that “duplication is far cheaper than the wrong abstraction”, and how forcing shared code onto cases that have drifted apart makes it harder to change." }, { "term": "Editing", "alsoCalled": [], "slug": "editing", "definition": "Improving a draft or published piece so that it says what it needs to, clearly and correctly, for its intended reader.", "category": "Editorial craft", "related": [ "content-brief", "content-editor", "content-maintenance", "copy-editing", "pair-writing", "proofreading", "senior-content-editor", "style-guide", "voice", "writer" ], "updated": "2026-10-14T12:10:00.000Z", "url": "https://contentlexicon.com/editing", "markdown": "https://contentlexicon.com/editing.md", "credit": "Editing, from the Content Lexicon by Contentious (https://contentlexicon.com/editing), CC BY 4.0", "body": "Editing covers everything from deciding that a report's third chapter should become its first to noticing that a phone number is missing a digit. Editors usually describe it in levels. A structural or developmental edit looks at the whole: what the piece is arguing, what it leaves out and whether its shape suits the reader. A line edit works on sentences, rhythm and [voice](https://contentlexicon.com/voice). [Copy editing](https://contentlexicon.com/copy-editing) brings consistency and correctness, and [proofreading](https://contentlexicon.com/proofreading) is the final check before publication. Organisations rarely run all four as separate stages, but knowing which one a draft needs prevents a lot of wasted effort.\n\nThe editor may be a [content editor](https://contentlexicon.com/content-editor), a copy editor, a colleague or the [writer](https://contentlexicon.com/writer) returning to their own draft after a break. What makes it editing rather than more writing is the change of position: reading as the intended reader would and asking what the piece needs to work for them. Robert Gottlieb, who edited Joseph Heller, said the editor's relationship to a book should be invisible. Readers should notice only that the argument has become clearer.\n\n## Why it matters for content\n\nMost of what an organisation publishes starts with someone who knows the subject far better than the reader does. That knowledge is the reason the content exists, and also why first drafts so often assume context, bury the main point or answer a question nobody asked. Editing is usually the first moment anyone stands in for the reader before publication. Without it, the organisation's internal logic – its structure, terminology and priorities – tends to go straight onto the page.\n\nEditing for the web also differs from editing for print in two important respects: the piece never stands alone and it’s rarely finished. A page sits among hundreds of others, so an editor has to ask whether it duplicates, contradicts or undermines something elsewhere on the site. [GOV.UK](http://GOV.UK)'s editorial review, known as 2i or \"second pair of eyes\", made this explicit: reviewers read the whole page even when a single line had changed, because one small update can cause problems elsewhere. Much organisational editing happens long after launch, when policies change, prices rise or a service moves. An editing process designed only for new content leaves the larger part of the estate without one.\n\n## How AI changes the picture\n\nEditing has always assumed a writer with something to say. The editor's job is to understand that intention and help the text serve it, querying where meaning is uncertain rather than inventing it. Generated drafts weaken that assumption. They often arrive fluent, well organised and free of errors, so the familiar signals that a draft needs work are missing, while the questions underneath, about whether it's accurate and whether anyone stands behind it, are harder to answer. The editor Mary Kole describes being hired to edit two AI-assisted novels and finding herself making hundreds of decisions that belonged to the author. Organisations meet the same problem in less literary forms: a briefing note or web page that reads smoothly but has no identifiable source for its claims and no view of what matters most.\n\nTwo practical consequences follow. Editing an AI-assisted draft often means returning to the brief and the evidence before touching any sentences, and sometimes concluding that rewriting from good notes would be quicker. Scope and disclosure matter more too, since an editor can't judge how much work a draft needs without knowing how it was made.\n\nAI is also a capable assistant to editing. It can suggest a structure, say what a draft appears to argue or play an impatient reader asking what a paragraph means. That gives writers some of the distance editing provides at the moment they need it, before a draft reaches anyone else. Its suggestions tend towards the most common way of saying things, so it's better at removing obstacles than at deciding what a piece should do. Judgement about purpose, emphasis and what to leave out stays with people, and editing is where much of that judgement gets exercised.\n\n## Questions people ask\n\n**Which kind of edit does a draft need?**\n\nStart with the biggest problem. If the argument is unclear or the piece is aimed at the wrong reader, polishing sentences is wasted work, since some of them will be cut. A sensible order is purpose and structure first, then sentences, then consistency and correctness, with a proofread last. Agreeing the level in advance also saves the awkward moment when a writer expecting a light tidy receives their draft rearranged.\n\n**How much should an editor change?**\n\nAs much as the piece needs to work for its reader, and no more. Good editors keep the writer's voice and meaning, explain significant changes and ask when they're unsure what was intended. Changing a sentence because you'd have written it differently isn't editing.\n\n**Can you edit your own work?**\n\nTo a point. Time away from a draft, reading it aloud or seeing it in a different format all help you read it as a stranger would. What self-editing can't easily supply is ignorance: you know what you meant, so you read it into the text. A second reader, or [pair writing](https://contentlexicon.com/pair-writing), catches what the writer can't see.\n\n**Does every piece need an editor?**\n\nEvery piece doesn't need the same edit. Proportion it to risk and reach: a change to opening hours needs a quick check, while a public report, health information or anything with legal consequences deserves a full edit and a separate proofread. What should stay constant is that someone other than the writer reads anything that matters before it goes out.\n\n## Learn more\n\n- [Robert Gottlieb, The Art of Editing No. 1](https://www.theparisreview.org/interviews/1760/the-art-of-editing-no-1-robert-gottlieb) – a Paris Review interview in which one of publishing's most celebrated editors argues that an editor's relationship to a book should be invisible, and describes the difficult balance between authority and possessiveness.\n- [Editorial reviews and mainstream content](https://insidegovuk.blog.gov.uk/2013/07/31/editorial-reviews-and-mainstream-content) – Jon Sanger's 2013 Inside [GOV.UK](http://GOV.UK) post on the 2i review, a clear account of editing for a whole website rather than a single page; [GOV.UK](http://GOV.UK)'s publishing guidance has since moved on, but the questions it asks of every update haven't dated.\n- [After Editing Two AI-Assisted Manuscripts, I Rewrote My Client Agreement](https://janefriedman.com/after-editing-two-ai-assisted-manuscripts-i-rewrote-my-client-agreement/) – Mary Kole's 2026 guest post for Jane Friedman on how editing generated drafts turned into authorship work, and why she changed her scope, disclosure terms and pricing in response." }, { "term": "Editorial board", "alsoCalled": [ "Content steering group, editorial committee" ], "slug": "editorial-board", "definition": "A group drawn from across an organisation that shapes content ideas into priorities, upholds standards and settles decisions that cut across teams.", "category": "Governance", "related": [ "content-governance", "content-strategy" ], "updated": "2026-10-14T15:24:00.000Z", "url": "https://contentlexicon.com/editorial-board", "markdown": "https://contentlexicon.com/editorial-board.md", "credit": "Editorial board, from the Content Lexicon by Contentious (https://contentlexicon.com/editorial-board), CC BY 4.0", "body": "An editorial board brings together people from across an organisation to decide what its content should say and how that content should be run. In many organisations it's where ideas are shared and shaped into a plan: members bring themes, stories and requests from their own corners, and the board decides which deserve attention and how they might fit together. It also settles questions too big, too contested or too risky for any one team or editor, such as priorities, standards, disputes over ownership and the handling of high-risk content. It works on the programme rather than on individual drafts, which stay with the people writing and [editing](https://contentlexicon.com/editing) them.\n\nThe name comes from newspapers, where the board argues over the issues of the day and decides the positions the paper takes in its editorials; the New York Times has had one since 1896. In-house [content teams](https://contentlexicon.com/content-team) borrowed the term, and the balance between its two jobs varies. Some boards are mostly planning meetings that feed an [editorial calendar](https://contentlexicon.com/content-calendar); others are mostly governance bodies that rule on standards and exceptions. Many do both. Whatever the mix, a useful board has defined decisions, suitable members and the authority to act. A monthly gathering that receives updates and refers every decision elsewhere neither generates ideas nor governs anything.\n\n## Why it matters for content\n\nGood ideas for content turn up all over an organisation. A fundraiser hears what supporters keep asking, a product manager knows what's about to change and a policy lead can see which debate is coming. Without somewhere to take them, those ideas arrive as one-off requests or not at all. An editorial board gives them a regular place to be pitched, argued over and combined, so a single strong theme can serve several teams instead of each commissioning its own version. The content team sees the whole programme rather than a queue of separate asks, and members who see their ideas become published work have a reason to keep coming.\n\nThe same group gives a home to conflicts that one editor can't settle: competing priorities, duplicated guidance, disputed ownership and exceptions to policy. Typical decisions include which content work gets funded this quarter, whether a new section of the site is needed, which of two teams owns a disputed topic, whether to approve a standard or an exception to it, and how to handle content with legal or reputational risk, such as a sensitive campaign or public statements during an incident. Day-to-day editorial judgement stays with the people doing the work.\n\nMembers are chosen for the decisions and the perspectives they bring, not for seniority alone. A board usually includes the [head of content](https://contentlexicon.com/head-of-content), senior owners from the main services or products, a subject or policy lead and someone who can commit budget. People who hear directly from the audience, such as support, sales or fundraising staff, are often a rich source of ideas. Legal, [accessibility](https://contentlexicon.com/accessibility), data protection or technology specialists join when the agenda needs them. Recording decisions and the reasons for them lets teams apply them later without asking again.\n\n## How AI changes the picture\n\nAI adds decisions that cross content, technology, legal and data protection: which sources an assistant may use, what requires human review and how poor answers are investigated. An editorial board can connect those decisions to existing [content policy](https://contentlexicon.com/content-policy) rather than letting each AI project invent its own miniature constitution.\n\nA board might decide how to respond when an automated answer repeatedly misstates a qualification: pause the use, investigate the source or change the review requirement. It shouldn't become the approval queue for every generated draft. It sets risk levels, standards and escalation routes, and teams work within them, bringing back exceptions and evidence of recurring failure.\n\nGenerating ideas has also become cheap. A model can suggest fifty angles on a theme in a minute, or summarise a month of search queries and support tickets before a planning meeting. Discussion can then start from evidence rather than from whoever spoke first, and the work shifts towards choosing. Members' knowledge of what the audience needs, what the organisation can credibly say and what has already been tried becomes the scarcer contribution.\n\n## Questions people ask\n\n**Who should sit on an editorial board?**\n\nThe people needed for its decisions and those with the clearest view of what the audience needs: usually content, service or product, subject expertise and a senior owner. Legal, accessibility or technology specialists can join when the agenda requires them.\n\n**How often should it meet?**\n\nOften enough to make timely decisions and act on ideas while they're still current. Monthly suits many boards, sometimes with a longer planning session each quarter, while a major migration or AI launch may need a shorter rhythm.\n\n**Does every organisation need one?**\n\nNo. A small team with clear ownership can share ideas and govern content directly. Boards help when ideas and decisions span teams and nobody has authority across the joins.\n\n**How do we stop it becoming another sign-off layer?**\n\nWrite down what the board decides and what it delegates, and keep the delegated list longer. Routine approvals belong with [content owners](https://contentlexicon.com/content-owner); the board sees exceptions, disputes and items above an agreed level of risk. A standing slot for new ideas helps too, since a board that only reviews and approves soon becomes something people work around. If the agenda fills up with individual pages, the delegation needs revisiting.\n\n## Learn more\n\n- [What Is an Editorial Board?](https://www.nytimes.com/2020/01/13/reader-center/editorial-board-explainer.html) – James Bennet's 2020 explainer on the New York Times board: what it does, how it stays separate from the newsroom and how it reaches decisions.\n- [Frequently Asked Questions](https://rosenfeldmedia.com/books/managing-chaos-frequently-asked-questions/) – short answers from Lisa Welchman's Managing Chaos: Digital Governance by Design on accountability for digital strategy, policy and standards, and why governance needn't slow teams down.\n- [Content Marketing Governance](https://contentmarketinginstitute.com/content-optimization/get-your-content-governance-in-order-with-this-1-thing) – a Content Marketing Institute article on the editorial board as a cross-functional group that gathers ideas, sets creative priorities and produces a shared calendar, including the story of one that grew out of an informal sharing meeting between two teams." }, { "term": "Email", "alsoCalled": [], "slug": "email", "definition": "A channel for correspondence, publications and service messages, delivered to a recipient's inbox.", "category": "Operations", "related": [ "channel-strategy", "content-calendar", "content-marketing", "conversational-design" ], "updated": "2026-10-11T06:24:00.000Z", "url": "https://contentlexicon.com/email", "markdown": "https://contentlexicon.com/email.md", "credit": "Email, from the Content Lexicon by Contentious (https://contentlexicon.com/email), CC BY 4.0", "body": "Unlike a web page waiting to be found, an email is delivered. It arrives in a place people often experience as personal, albeit crowded.\n\nEmail is often dismissed as yesterday’s technology, but its familiarity and ubiquity are strengths. People can communicate across organisations, devices and services without joining the same social platform. An email address provides a continuing point of contact even as other channels fall in and out of favour.\n\nThat proximity gives email room for a distinctive [voice](https://contentlexicon.com/voice) and an ongoing relationship. A newsletter can become something readers anticipate: a familiar perspective, a story unfolding over several editions or a thoughtful selection of things they might otherwise miss. Permission and relevance matter because the sender is entering a space the reader uses for much of their life.\n\n## Why it matters for content\n\nNewsletters are publications in their own right. They can carry essays, reporting, curated discoveries, personal reflections and serial narratives, with enough substance to reward reading in the inbox. Sending readers elsewhere is one possible purpose, not a requirement for every edition.\n\nThe experience develops over time. Voice, design, frequency and recurring features help readers recognise a publication and decide what place it has in their lives. A monthly letter can invite leisurely reading; a daily briefing can become part of a routine. Subject line, sender name and opening sentence set expectations, while the content needs to fulfil them. A familiar sender still has to earn attention.\n\nEmail also supports conversations. Replies can bring questions, experiences and ideas back to the people creating the content, informing future editions and making the relationship more reciprocal. Teams should decide whether they welcome replies and who will respond, rather than offer an invitation that leads to an unattended inbox.\n\nService messages have their own important role: confirming what happened, explaining a next step or reminding someone before a deadline. Their outcome or required action should be unmistakable. Across editorial and service email, [accessibility](https://contentlexicon.com/accessibility), readable structure and meaningful links matter. Important information shouldn’t be trapped in an image or attached [PDF](https://contentlexicon.com/pdf) without explanation, and connections to [websites](https://contentlexicon.com/website) or other services need to make sense.\n\n## How AI changes the picture\n\nAI offers possibilities for more responsive editorial experiences. With appropriate data and permission, personalisation can adapt the selection, depth, timing and presentation of content to a reader’s interests and circumstances. A newsletter might offer a newcomer an explanation of an unfamiliar idea and a specialist a more detailed treatment. The editorial challenge is to preserve a coherent voice and shared perspective while making individual editions more relevant.\n\nAI agents can also conduct conversations through email. A reader could reply to a newsletter with a question, ask for related reading or explore how an idea applies to their situation. An agent connected to reliable source material could continue that exchange over several messages. Agents could also help coordinate arrangements or guide someone through a service, using a channel they already know rather than requiring a new interface.\n\nThese possibilities depend on trust. People need to know when they’re corresponding with an agent, what information it can use and what it has permission to do. They should be able to stop the exchange or reach a person when needed. Personalisation should serve interests people have expressed, without making intrusive assumptions or narrowing everything they encounter to what they already know.\n\nAI can help with drafting, summarising and classifying replies too, while recipients’ assistants may summarise incoming newsletters or answer questions about them. Clear context and descriptive subjects help both people and machines interpret messages. Facts, dates, links and consent need reliable sources, and consequential messages or actions need appropriate human oversight. Generating more email is easy; giving people a reason to welcome it remains the harder work.\n\n## Questions people ask\n\n**What makes a newsletter worth subscribing to?**\n\nA clear editorial promise, fulfilled consistently. That might mean a distinctive perspective, useful curation, absorbing stories or access to knowledge readers value. Frequency and format should suit that promise, rather than simply fill a publishing schedule.\n\n**How long should an email be?**\n\nLong enough to reward the attention it asks for. A confirmation rarely needs an essay, but an essay newsletter can deserve an unhurried read. Put essential actions first in service messages; give editorial content a structure and pace that suit the reading experience.\n\n**Should a newsletter contain the whole piece or link elsewhere?**\n\nEither can work. Publishing a complete piece in the inbox lets people read without going elsewhere. An excerpt or curated link can help them discover material worth visiting. Be clear about what each edition offers, and link to a stable source when information is likely to change or requires a web-based task.\n\n**How could an AI agent change a newsletter?**\n\nIt could let readers ask questions, request further explanation or explore recommendations through replies. That creates scope for an ongoing exchange around the publication, provided the agent has reliable sources, clear limits and a way to involve a person.\n\n## Learn more\n\n- [Email Is Dead, Long Live Email](https://annhandley.com/long-live-email) – Ann Handley's essay on how distinctive voice, storytelling and relationships can help newsletters remain worth reading as AI assistants mediate the inbox.\n- [AI turns news into a conversation](https://www.niemanlab.org/2024/12/ai-turns-news-into-a-conversation) – Nikita Roy's December 2024 prediction about journalism readers can question and explore through voice-first dialogue. A future-facing argument about responsive publications, not evidence of email deployment.\n- [Email Newsletters Are a New Literary Genre](https://www.thecut.com/2021/07/email-newsletters-new-literary-style.html) – Molly Fischer's essay on newsletters as a form of voice, intimacy and belonging, with a critical eye on the blurred boundaries between personal writing and personal branding.\n- [AgentMail webhooks](https://docs.agentmail.to/webhooks-overview) – technical documentation showing how incoming messages can trigger an email agent, including thread identifiers and the limits of event payloads. A concrete capability example from AgentMail, not a neutral survey of providers." }, { "term": "Evergreen content", "alsoCalled": [], "slug": "evergreen-content", "definition": "Content whose core value remains useful over time, with review and maintenance when needed.", "category": "Operations", "related": [ "content-decay", "content-maintenance", "longform", "slow-content" ], "updated": "2026-10-11T06:25:00.000Z", "url": "https://contentlexicon.com/evergreen-content", "markdown": "https://contentlexicon.com/evergreen-content.md", "credit": "Evergreen content, from the Content Lexicon by Contentious (https://contentlexicon.com/evergreen-content), CC BY 4.0", "body": "Evergreen content addresses subjects whose interest or significance outlasts a particular news cycle, event or campaign. A guide to writing [alt text](https://contentlexicon.com/alt-text) may remain useful as a reference; an essay about belonging can still move readers or change how they see the world years after publication. Its value doesn't depend on being new. The term comes from journalism, where an evergreen is a story that can be published at different times without losing its relevance.\n\nEvergreen content sits at one end of a range. Timely content, such as news, event pages and campaign material, is meant to be useful for a short period and then archived or removed. Evergreen content is meant to last, but it still needs looking after: examples age, links break and better evidence appears. Left alone, it becomes [ROT content](https://contentlexicon.com/rot-content) (redundant, outdated or trivial) like anything else, only more slowly.\n\n## Why it matters for content\n\nEvergreen material becomes a dependable part of a content estate. It answers recurring needs, attracts links and search traffic over years and reduces the pressure to publish constantly to look active.\n\nBecause its value accumulates, it deserves more investment in structure, [accessibility](https://contentlexicon.com/accessibility) and maintenance than a one-off post. One canonical piece on a topic, kept current, usually does more than repeated introductions to the same subject. Vox put this into practice in 2015 by asking its [writers](https://contentlexicon.com/writer) to update and republish older articles every week.\n\nDates need care. Readers should know when material was written and last reviewed, but changing a timestamp without checking the content disguises old content as new.\n\n## How AI changes the picture\n\nEvergreen sources are valuable for AI retrieval because they provide stable explanations and definitions. Clear ownership and review history help teams and retrieval systems pick current, approved material over older or derivative versions.\n\nAI can suggest where evergreen pages need refreshed examples, updated links or clearer structure. It can also generate any number of derivative versions of the same explainer, which makes it more important to know which one is canonical and which facts in it are protected.\n\nDurable content can be written for reuse without becoming bland. Strong examples and a clear point of view stay useful when they're tied to lasting principles rather than passing fashion, and they give a retrieval system something more distinctive to cite than a generic summary.\n\n## Questions people ask\n\n**Is evergreen content always educational?**\n\nOften, but not always. Tools, reference material, service explanations and lasting case studies can be evergreen too.\n\n**How often should it be reviewed?**\n\nAccording to risk and rate of change. Durable content may need less frequent review than news or campaign pages, but it still needs some, and a [review date](https://contentlexicon.com/review-date) makes sure it happens.\n\n**Should old publication dates be removed?**\n\nNo. Be open about when something was published and reviewed. Age alone doesn't make accurate content worse, and readers use dates to judge whether advice still applies.\n\n**Does evergreen mean generic?**\n\nNo. Specific expertise and examples are often what give durable content its lasting value. A generic explainer is easy to replace; one built on experience isn't.\n\n## Learn more\n\n- [Refreshing the evergreen](https://www.vox.com/2015/1/15/7546877/evergreen-experiment) – Matthew Yglesias on Vox's 2015 experiment in updating and republishing older articles, which brought more than 500,000 readers to 88 refreshed stories in five days.\n- [Routinising the Expected? Defining and Characterising Evergreen Journalism Online](https://www.tandfonline.com/doi/full/10.1080/1461670X.2025.2588231) – Cecilia Arregui Olivera's open-access study of how the news industry defines evergreen journalism, from filling slow news days to getting lasting value from published work.\n- [Why you need an evergreen content strategy](https://econsultancy.com/why-you-need-an-evergreen-content-strategy) – Graham Charlton's account of reference posts earning traffic over time, illustrated with Econsultancy's own publishing results." }, { "term": "Feature flag", "alsoCalled": [ "Feature toggle" ], "slug": "feature-flag", "definition": "A software-controlled switch that determines whether a feature or content variant is available, and to whom.", "category": "AI and tooling", "related": [ "content-decay", "content-maintenance" ], "updated": "2026-10-10T20:45:00.000Z", "url": "https://contentlexicon.com/feature-flag", "markdown": "https://contentlexicon.com/feature-flag.md", "credit": "Feature flag, from the Content Lexicon by Contentious (https://contentlexicon.com/feature-flag), CC BY 4.0", "body": "A feature flag is a switch in software that decides whether a feature, or a version of it, is available and to whom, without anyone changing the code. Both versions go live and the flag decides which one people see. If something misbehaves, it can often be switched off in seconds.\n\nFlags separate putting a change live from releasing it to people. A team can deploy new work, test it with colleagues and decide later when, or whether, everyone else gets it. Flickr's engineers described the approach in 2009, and it's now routine in software development.\n\nMost flags do one of four jobs: hiding unfinished work until launch, showing different versions to different groups (as in [A/B testing](https://contentlexicon.com/a-b-testing)), letting a team switch off a failing feature or giving certain users access that others don't have. Launch flags should be short-lived; access flags can last for years.\n\n## Why it matters for content\n\nEvery feature behind a flag has content behind it too: labels, error messages, onboarding, help articles, release notes and support scripts. That content needs to be ready before the flag is switched on, which often means writing for something most people can't see yet. Help pages may need to describe both states or sit behind the same flag, so people without the feature aren't sent looking for a button they don't have.\n\n[Writers](https://contentlexicon.com/writer) and reviewers also need a way to preview each variant in context. Content in an experiment needs the same review as anything else, because one group of real users will read the version that loses. Turning a feature off needs content too, such as an explanation or an alternative route, rather than a button that silently vanishes.\n\n[Content teams](https://contentlexicon.com/content-team) use flags directly as well, for seasonal guidance, regional information or trial wording, without creating unmanaged duplicate pages.\n\nFlag debt is the cost of flags nobody removes. Each one adds states that need testing, and old ones can hide content that nobody checks or maintains. Every flag needs an owner, a reason and a removal date.\n\n## How AI changes the picture\n\nAI features suit controlled release because their behaviour can be hard to predict. Teams can expose a new assistant, [prompt](https://contentlexicon.com/prompt) or knowledge source to a small group, watch for failures and widen access carefully. A flag also gives a fast way to switch off an assistant that starts giving bad answers.\n\nFlags can compare different instructions, models or retrieval sources. The experiment still needs ethical limits, clear measures and protection against giving one group a materially poorer or riskier service, especially where the answers affect money, health or legal rights.\n\nAI can help analyse feedback from flagged variants, but people must decide what counts as success and when a feature is safe to release to everyone.\n\n## Questions people ask\n\n**Do feature flags always involve an experiment?**\n\nNo. Experiments are one use. Many flags exist for release control, permissions or emergency shut-off, with no comparison involved.\n\n**Can flags replace separate pages?**\n\nSometimes. They work well where content shares a source and changes by a defined condition, such as region or date. Distinct [user needs](https://contentlexicon.com/user-needs) may still call for distinct content.\n\n**Who should control the flag?**\n\nA named owner with enough product, content and operational context to understand the consequences of switching it. For flags that change what people read, that should include someone responsible for the content.\n\n**When should a flag be removed?**\n\nWhen the feature is fully launched, abandoned or no longer needs conditional behaviour. Temporary flags should have [review dates](https://contentlexicon.com/review-date), and removing the flag should include removing the content for the state that's gone.\n\n## Learn more\n\n- [Feature Toggles](https://martinfowler.com/articles/feature-toggles.html) – Pete Hodgson's taxonomy of release, experiment, operational and permissioning toggles.\n- [Flipping Out](https://code.flickr.net/2009/12/02/flipping-out) – Ross Harmes on how Flickr used feature flags and per-user \"flippers\" in 2009 to deploy from a single branch several times a day, and the clean-up each launch left behind.\n- [Knight Capital Americas LLC](https://www.sec.gov/files/litigation/admin/2013/34-70694.pdf) – the US Securities and Exchange Commission's 2013 order on the firm's 2012 trading incident, describing how a repurposed flag triggered old code on a server that had missed a deployment." }, { "term": "Findability", "alsoCalled": [], "slug": "findability", "definition": "The ease with which people and systems can locate content they know or suspect exists.", "category": "Structure and systems", "related": [ "card-sorting", "information-architecture", "metadata", "paddle-swim-dive", "seo", "taxonomy" ], "updated": "2026-10-11T06:26:00.000Z", "url": "https://contentlexicon.com/findability", "markdown": "https://contentlexicon.com/findability.md", "credit": "Findability, from the Content Lexicon by Contentious (https://contentlexicon.com/findability), CC BY 4.0", "body": "Findability describes how easily someone can get to content they're looking for, by whatever route they try. It belongs to the content and the systems around it together: an excellent page can still have poor findability if its title uses internal [jargon](https://contentlexicon.com/jargon), it sits three levels deep under a misleading label or search ranks an old copy above it. Peter Morville popularised the term in the early 2000s and explored it in his 2005 book Ambient Findability.\n\nPeople find content through navigation, internal search, external search engines, links from other pages, recommendations and, increasingly, AI assistants. Findability is about locating something a person knows or suspects exists. Discoverability, a related idea, covers coming across useful things they didn't know to look for. Both rest on the same groundwork but are tested differently.\n\n## Why it matters for content\n\nPeople use their own language and mental models, not the organisation chart. [Information architecture](https://contentlexicon.com/information-architecture), labels, [metadata](https://contentlexicon.com/metadata) and internal links should reflect real tasks and the words people use for them. A council might file rubbish collections under \"Environmental services\", but residents search for \"bins\".\n\nFindability also depends on the content itself. Clear titles, descriptive headings and explicit subjects help search systems judge relevance. Duplicates and outdated pages make the right answer harder to identify, even when everything is technically indexed.\n\nFindability can be measured, by testing whether people can follow a route rather than checking that a page exists. Tree testing asks people to find things using only the bare menu structure, which shows whether labels and categories make sense before any design work. Search logs show the words people use, and queries that return nothing, or that people repeat with different wording, point to missing or mislabelled content. Task-based [usability testing](https://contentlexicon.com/user-testing), first-click data and support contacts fill in the rest.\n\n## How AI changes the picture\n\nAI assistants create new routes to content. Someone who remembers a report's finding but not its title can describe it, narrow down the options with a follow-up question and open the source. That helps when people's words don't match the organisation's, but it doesn't fix weak sources or permission problems. Finding the document and receiving a plausible summary of it are also different outcomes, and only one of them lets the reader check the source.\n\nMetadata, canonical status and relationships help assistants pick the current, authoritative item. Without them, a fluent answer may come from the wrong page.\n\nFindability now extends beyond ranking pages. Content needs to be locatable, interpretable and citable by systems while staying clear to people.\n\n## Questions people ask\n\n**Is findability the same as SEO?**\n\nNo. [SEO](https://contentlexicon.com/seo) covers external search engines. Findability also includes navigation, internal search, links and AI retrieval, and much of it depends on decisions made inside the organisation.\n\n**Can good search fix poor navigation?**\n\nIt can provide another route, but it shouldn't excuse confusing structure and labels. Many people browse before they search, and search results inherit the same unclear titles.\n\n**Why can't people find a page that's in the menu?**\n\nUsually because the label doesn't match what they're looking for, or the page sits under a category they wouldn't think to open. Tree testing and search logs tend to show which.\n\n**Does metadata matter if AI understands language?**\n\nYes. Metadata supplies explicit context, authority and relationships that reduce guesswork, such as which of two similar pages is current and who it's for.\n\n## Learn more\n\n- [Ambient Findability](https://archive.org/details/ambientfindabili0000morv) – Peter Morville's curious, wide-ranging book on search, wayfinding and information overload.\n- [Tree Testing: Evaluate Menu Labels and Categories](https://www.nngroup.com/articles/tree-testing/) – Page Laubheimer of Nielsen Norman Group on testing whether people can find things in a navigation structure before any visual design exists.\n- [Site Search Analytics: A Conversation with Lou Rosenfeld](https://www.uxmatters.com/mt/archives/2011/07/site-search-analytics-a-conversation-with-lou-rosenfeld.php) – a UXmatters interview on using search query data to improve findability, including why searches that return no results are a quick way to spot missing, mislabelled or jargon-heavy content." }, { "term": "Front-loading", "alsoCalled": [ "Inverted pyramid" ], "slug": "front-loading", "definition": "Putting the information readers need most at the start of a sentence, section or page.", "category": "Editorial craft", "related": [ "information-architecture", "paddle-swim-dive", "plain-language", "readability" ], "updated": "2026-10-12T13:46:00.000Z", "url": "https://contentlexicon.com/front-loading", "markdown": "https://contentlexicon.com/front-loading.md", "credit": "Front-loading, from the Content Lexicon by Contentious (https://contentlexicon.com/front-loading), CC BY 4.0", "body": "Front-loading gives people the answer before the history, the action before the explanation and the meaningful words before generic labels.\n\nThe approach resembles journalism’s inverted pyramid, where the central fact comes first and supporting detail follows. It suits digital reading because people scan, arrive with a task and may read only a fraction of what is available.\n\n## Why it matters for content\n\nFront-loading helps readers decide quickly whether content is relevant. A heading such as “Apply for a grant” carries more information than “Further information”. An [email](https://contentlexicon.com/email) that begins with the outcome respects someone’s time more than two paragraphs of organisational throat-clearing.\n\nIt also supports [accessibility](https://contentlexicon.com/accessibility). Screen-reader users may navigate through headings and links out of context, while people with cognitive or visual impairments benefit from predictable structure. Important words at the beginning make lists, navigation and search results easier to distinguish.\n\nFront-loading doesn't mean removing nuance or forcing every piece into blunt instructions. It means establishing orientation before adding context. [Longform](https://contentlexicon.com/longform) arguments can still unfold carefully, but readers should not have to excavate the subject from paragraph four.\n\n## How AI changes the picture\n\nAI can reorganise drafts, propose descriptive headings and summarise long passages. It can help teams notice where introductions delay the answer or where repeated phrases push distinctive information to the end.\n\nAI systems also extract snippets and summaries from content. Clear openings, explicit headings and self-contained sentences make it easier for them to represent the source accurately. This isn't a reason to write for robots instead of people; both benefit from content that states its point.\n\nGenerated prose often begins with generic scene-setting because it's a safe statistical pattern. For task-focused content, [prompts](https://contentlexicon.com/prompt) and [editing](https://contentlexicon.com/editing) guidance can ask for the answer first. A story or exploratory argument may need a different order; orientation doesn't always mean revealing the ending. Human judgement is still needed to decide what matters most for the reader’s purpose.\n\n## Questions people ask\n\n**Does front-loading mean giving away the conclusion?**\n\nYes, when the conclusion is what the reader needs. Digital content isn't a detective novel, unless it happens to be a detective novel.\n\n**Should every sentence use the same structure?**\n\nNo. Rhythm matters. Front-load where it improves orientation, especially in headings, instructions, links and opening paragraphs.\n\n**Is front-loading the same as writing briefly?**\n\nNo. Long content can be front-loaded and short content can bury its point. The principle concerns order rather than length.\n\n**Does it help SEO?**\n\nClear topic language can help discovery, but the primary value is comprehension. Write the strongest opening for the reader rather than repeating keywords mechanically.\n\n## Learn more\n\n- [Inverted Pyramid: Writing for Comprehension](https://www.nngroup.com/articles/inverted-pyramid) – Amy Schade’s explanation of conclusion-first writing, its web-reading benefits and how to arrange supporting detail.\n- [Writing for the web](https://web.wsu.edu/web-training/resources/writing-for-the-web) – a compact collection connecting front-loading with scanning patterns and readable layouts.\n- [First things first](https://yoast.com/inverted-pyramid) – Yoast’s writing perspective on the inverted pyramid for readers and search; useful organisation advice, not proof of ranking or AI-summary effects." }, { "term": "Frontend", "alsoCalled": [ "Front end" ], "slug": "frontend", "definition": "The part of a website, app or other channel that people see, hear and interact with.", "category": "Structure and systems", "related": [ "accessibility", "backend", "content-design", "conversational-design", "design-system", "headless-cms", "microcopy", "staging", "ux" ], "updated": "2026-10-11T06:24:00.000Z", "url": "https://contentlexicon.com/frontend", "markdown": "https://contentlexicon.com/frontend.md", "credit": "Frontend, from the Content Lexicon by Contentious (https://contentlexicon.com/frontend), CC BY 4.0", "body": "In software, the frontend is the client side of a [website](https://contentlexicon.com/website) or app: the HTML, CSS and JavaScript that run in a browser or on a device to present pages, navigation, forms, buttons and media. It handles layout, interaction and much of what makes a service usable and accessible, while the [backend](https://contentlexicon.com/backend) decides which information to send in the first place.\n\nFor content, the frontend is wherever content reaches people. That might be a web page or an app screen, but also an [email](https://contentlexicon.com/email), a push notification, a printed leaflet, a voice assistant's reply or a chatbot's answer. It includes the words on buttons and in error messages as much as articles, along with the order, emphasis and timing in which they appear. Ann Rockley's frontend [content strategy](https://contentlexicon.com/content-strategy) describes this end of the work: who the audience is, what they need, which [content types](https://contentlexicon.com/content-type) serve them and how to write for them.\n\n## Why it matters for content\n\nMeaning changes with presentation. A sentence in a [CMS](https://contentlexicon.com/cms) field may become a heading on one screen, a notification on another and a spoken answer on a third, each with different constraints of length, context and attention. A warning that reads clearly in the [editing](https://contentlexicon.com/editing) interface can be easy to miss once it's placed below the action button. Headings, labels, link text and reading order carry meaning in the way [punctuation](https://contentlexicon.com/punctuation) does for a sentence, which is why content people need to see and test the frontend rather than approve isolated strings. The same structure, expressed in well-formed headings and semantic HTML, is what assistive technology, search engines and AI systems rely on to make sense of a page.\n\nThe frontend is also where the backend gets judged. Readers never see the [content model](https://contentlexicon.com/content-model), the [taxonomy](https://contentlexicon.com/taxonomy) or the approval route. They see whether the fee is right, whether the page answers their question and whether search turns up what they came for. In a [content operating model](https://contentlexicon.com/content-operating-model), the answers to what is published and for whom are tested here, so what people do at the frontend should feed back into the models, [workflows](https://contentlexicon.com/workflow) and guidance behind it. [Components](https://contentlexicon.com/component) work best when content, design and development shape them together, around real language, errors, translations and edge cases rather than placeholder text.\n\n## How AI changes the picture\n\nMore of the frontend is now composed at the moment someone asks. A chatbot or search assistant writes an answer that no editor saw in that exact form, drawing on whatever sources it can reach. Editorial work moves from approving finished screens towards setting the patterns, sources, boundaries and fallbacks for what may be composed: how to express uncertainty, when to hand over to a person and what to say when the answer isn't known. Language becomes the main control surface, so wording and recovery are design decisions in their own right.\n\nPeople also meet content on frontends the organisation doesn't control. AI summaries in search results, assistants inside other products and agents that browse on someone's behalf all present an organisation's content in their own words and layouts. You can't design those screens, but titles, summaries, clear structure and accurate [metadata](https://contentlexicon.com/metadata) influence how well content holds up when it's presented elsewhere. It helps to check what these tools say about your services, as you would test your own pages.\n\nBuilding frontends has changed too. AI can generate interface code and working prototypes quickly, which lets [content designers](https://contentlexicon.com/content-designer) try real wording in a real layout much earlier than a design handover would allow. A generated interface still knows nothing about the service behind a button, so content priorities, [accessibility](https://contentlexicon.com/accessibility) tests and real source content remain the measure of whether it works.\n\n## Questions people ask\n\n**Is the frontend mostly visual design?**\n\nNo. It includes structure, behaviour, sound and content as well as appearance. A keyboard user's route through a form and the order in which a screen reader announces a page are part of the frontend too.\n\n**Is a chatbot a frontend?**\n\nYes, a conversational one. The same backend can feed a web page and a chatbot, but the chatbot usually composes its own wording from that material, so its instructions, sources and limits need as much editorial attention as a page template.\n\n**How much do content people need to know about frontend code?**\n\nEnough to work with components and constraints. Knowing what a heading level does, how a [design system](https://contentlexicon.com/design-system)'s components handle long or translated text and why link text matters for screen readers improves decisions and conversations with developers, without anyone needing to write production code.\n\n**What can we do about frontends we don't control?**\n\nConcentrate on what travels. Clear titles and summaries, a sensible structure, accurate metadata and pages that state their key facts plainly give search engines, social previews and AI tools better material to work with. Then look at the results: search for your own services and read what comes back.\n\n## Learn more\n\n- [The Interface is Content](https://borism.medium.com/the-interface-is-content-b13f4d693d66) – Boris Müller's essay arguing that interface elements, like typography for text, carry meaning and can't be stripped away without making content harder to understand.\n- [Content and Design Are Inseparable Work Partners](https://medium.com/uie-brain-sparks/content-and-design-are-inseparable-work-partners-5e1450ac5bba) – Jared Spool's case that content is whatever a user needs at that moment, wherever it lives, and that organisations should integrate content and design teams to match.\n- [Atlassian Design System](https://atlassian.design/design-system) – one product family's foundations, components and content design standards in a single system, including guidance for AI-powered moments; a concrete example to explore, not evidence of effectiveness." }, { "term": "Gap analysis", "alsoCalled": [], "slug": "gap-analysis", "definition": "A comparison between what exists now and what should exist, so the missing parts can be found.", "category": "Research and evidence", "related": [ "content-audit", "content-inventory", "user-needs" ], "updated": "2026-10-06T15:28:00.000Z", "url": "https://contentlexicon.com/gap-analysis", "markdown": "https://contentlexicon.com/gap-analysis.md", "credit": "Gap analysis, from the Content Lexicon by Contentious (https://contentlexicon.com/gap-analysis), CC BY 4.0", "body": "In content work a gap analysis usually sits alongside an inventory or an audit: the inventory says what you have, the audit says what state it is in, and the gap analysis says what is absent.\n\nAnyone can run one. Its value comes entirely from the standard you compare against. Against [user needs](https://contentlexicon.com/user-needs) it shows unanswered questions; against a competitor set, omissions; against a [content model](https://contentlexicon.com/content-model), structural holes. Without a stated standard, it's just a list of things somebody happens to have noticed.\n\n## Why it matters for content\n\nMissing content is harder to spot than bad content, because nothing on the screen draws attention to it. Nobody complains about a page that was never written; they ring the office, abandon the form, guess the answer or go elsewhere. A gap analysis is how those silences become evidence you can act on.\n\nIt's also one of the most persuasive pieces of work in a content project. A gap expressed as an unanswered user question – asked forty times a month in site search and answered nowhere – tends to move decisions that a general plea for better content never will.\n\nThe best time to run one is before commissioning anything new, whether that's a site, a section, a campaign or a support hub.\n\n## How AI changes the picture\n\nThe mechanical part of gap analysis goes much faster with AI. It can cluster search queries and support enquiries into themes, match those themes against existing pages, compare your coverage with a set of external sources and produce a first list of candidate gaps in an afternoon rather than a fortnight.\n\nIt also creates a new kind of gap to look for. If an AI assistant is answering questions about your organisation, the interesting question is no longer only what your site lacks, but what a machine can't find, can't trust or gets wrong. Asking several assistants your users' most common questions is a crude but revealing form of gap analysis. When three of them give different opening hours for the same office, the source content is usually missing, buried or contradicting itself.\n\nThe judgement remains human. AI will cheerfully identify gaps nobody should bother filling, and it can't tell you which absences reflect a deliberate decision rather than an oversight. A gap you chose is a scope boundary, not a failure.\n\n## Questions people ask\n\n**Is a gap analysis the same as a content audit?**\n\nThey answer different questions. An audit assesses the content you have; a gap analysis identifies the content you should have and don't.\n\n**Do we need a complete inventory first?**\n\nNot necessarily. A gap analysis against a clear set of user needs can be done on a section at a time, which is usually more useful than waiting for a complete inventory.\n\n**What do we compare against if we have no research?**\n\nSite search terms, support enquiries, the questions staff answer repeatedly and your own task list. Imperfect evidence beats a guess dressed as a strategy.\n\n**Does every gap need filling?**\n\nNo. Some gaps are somebody else's job, and some are content you've rightly decided not to own. Record the decision so the gap stops resurfacing.\n\n## Learn more\n\n- [Learning about users and their needs](https://www.gov.uk/service-manual/user-research/start-by-learning-user-needs) – the UK government's Service Manual on finding and recording user needs, the standard most content gap analyses should be measured against.\n- [Content gap analysis: a step-by-step guide](https://www.semrush.com/blog/content-gap-analysis/) – a search-led method, now including checking which prompts and AI citations surface competitors and whether content is easy to extract, though it leans on Semrush's own tools." }, { "term": "GEO", "alsoCalled": [], "slug": "geo", "definition": "Improving the likelihood that content is selected, accurately represented and cited in AI-generated answers.", "category": "AI and tooling", "related": [ "findability", "machine-readable", "rag", "seo", "structured-content", "website" ], "updated": "2026-10-11T11:41:00.000Z", "url": "https://contentlexicon.com/geo", "markdown": "https://contentlexicon.com/geo.md", "credit": "GEO, from the Content Lexicon by Contentious (https://contentlexicon.com/geo), CC BY 4.0", "body": "GEO stands for generative engine optimisation. It concerns how content appears within AI-generated answers, rather than only where a page ranks in a list of search results. Selection, representation and attribution are separate questions: being mentioned doesn't mean being represented accurately.\n\nThe term covers a changing set of systems and practices. Clear sources and structure can help, but an organisation doesn't control every engine's retrieval or answer generation. Treat techniques as hypotheses to test, not a checklist that guarantees visibility.\n\n## Why it matters for content\n\nGEO matters because discovery is changing. People increasingly encounter information through generated summaries, chat interfaces and tools that synthesise multiple sources before the person ever visits a page. That affects how organisations think about [findability](https://contentlexicon.com/findability), authority, citation and the shape of their content.\n\nFor [content teams](https://contentlexicon.com/content-team), the useful response isn't to chase every rumoured trick. Make the material easier for people and systems to use: identify its subject, support its claims and show when a qualification applies. Those improvements have value even when an engine doesn't select the page.\n\n## How GEO changes the picture\n\nA useful answer may reach someone without a visit to its source. Teams therefore need to examine both visibility and fidelity: did the answer cite the right material, preserve its qualifications and give the reader a route to check it? Traffic alone won't show that.\n\nA publisher could test a defined set of questions and compare the answers before and after making a source clearer. Results will depend on the engine, question and time of testing. Keep the publication useful on its own; changing it only to win a citation can damage the experience for people who actually read it.\n\n## Questions people ask\n\n**How does GEO differ from SEO?**\n\n[SEO](https://contentlexicon.com/seo) and GEO overlap in helping sources be found, but GEO also considers how generated answers select, represent and cite them. Those systems can produce a summary without a click. Neither process establishes that the information is true.\n\n**Should we create content only for AI answers?**\n\nKeep the reader’s purpose primary. Structure and evidence may help machines use the material, but an essay, guide or experience should still be valuable when someone reads it directly.\n\n**What content qualities help GEO?**\n\nStart with reliable information and make its scope clear. Descriptive headings and relevant structured data can aid discovery and interpretation, but their effects vary. Use evidence and examples because they help the reader, not as decorative citation bait.\n\n**How do we measure it?**\n\nMeasurement is still developing. Teams can track citations in AI answers, branded mentions, referral traffic from AI tools where visible, changes in search behaviour and the quality of generated summaries. They should also watch whether content remains accurate when summarised.\n\n## Learn more\n\n- [GEO: Generative Engine Optimization](https://arxiv.org/pdf/2311.09735) – the 2024 conference paper’s benchmark and visibility experiments, with effects varying by domain and method; not a recipe for every current answer engine.\n- [Creating helpful, reliable, people-first content](https://developers.google.com/search/docs/fundamentals/creating-helpful-content) – a useful guard against turning GEO into another excuse for synthetic search filler.\n- [Generative Engine Optimization: How to Dominate AI Search](https://arxiv.org/abs/2509.08919) – a 2025 comparative study finding that AI search leans heavily on earned, third-party media rather than brand-owned and social content, unlike Google’s results." }, { "term": "Git", "alsoCalled": [], "slug": "git", "definition": "A version control system for managing change, collaboration, review and release across files over time.", "category": "AI and tooling", "related": [ "branch", "markdown", "merge-conflict", "pull-request", "versioning" ], "updated": "2026-10-11T06:05:00.000Z", "url": "https://contentlexicon.com/git", "markdown": "https://contentlexicon.com/git.md", "credit": "Git, from the Content Lexicon by Contentious (https://contentlexicon.com/git), CC BY 4.0", "body": "Git records changes to files, but more importantly it gives teams a way to work on the same material without overwriting one another, to review changes before they are accepted, to compare versions in detail, to merge work from different people and to restore an earlier state when something goes wrong. Its basic units are commits (saved changes), [branches](https://contentlexicon.com/branch) (parallel lines of work), merges (bringing changes together) and tags (named states you can return to). Hosting services such as GitHub and GitLab add [pull requests](https://contentlexicon.com/pull-request), a formal review step before changes are merged.\n\nVersion history is only the start. Git is a system for change management and collaboration: who changed what, in which branch, against which version, with whose review and how that change became the current state. That's why software teams use it for far more than code, and why [content teams](https://contentlexicon.com/content-team) increasingly use it for documentation, [content models](https://contentlexicon.com/content-model), [prompts](https://contentlexicon.com/prompt), [schemas](https://contentlexicon.com/schema), design tokens, configuration, [Markdown](https://contentlexicon.com/markdown) content and whole [websites](https://contentlexicon.com/website).\n\n## Why it matters for content\n\nGit makes content work safer, more visible and more deliberate. It lets a team draft a major change on a branch without touching the live version, compare two drafts line by line, use a configured hosting [workflow](https://contentlexicon.com/workflow) to require review before something is merged, see what changed between releases and roll back if a change breaks something. It's especially useful where content is structured, reused or shipped through systems rather than pasted into a single visual editor, the kind that shows the page as you type.\n\nA content team can use Git to manage a [style guide](https://contentlexicon.com/style-guide) as a living document, maintain prompts and AI instructions with a proper audit trail, coordinate edits to a content model with developers, keep translations and regional variants aligned, and publish documentation or websites from a trusted source rather than from whichever copy someone last downloaded.\n\nIt earns its place when several people touch the same material, or when getting the live version wrong would be expensive.\n\n## How AI changes the picture\n\nPrompts, retrieval sources, schemas and guidance change constantly, and small edits can have large effects. When an assistant starts answering differently on a Tuesday, a diff can show that someone changed one line of its instructions on the Monday. Git gives those assets provenance, review and [rollback](https://contentlexicon.com/rollback).\n\nIt also makes experimentation safer: branches for testing, pull requests for scrutiny, tags for stable versions and diffs that show whether a change altered substance or only wording.\n\n## Questions people ask\n\n**Do I have to use the command line?**\n\nNo. Many people use Git through interfaces such as GitHub Desktop, VS Code, GitKraken or the built-in Git features in other tools. The command line is still the most direct and flexible way to use it, and it makes the underlying model easier to understand, but it isn't a requirement.\n\n**Do I have to understand pull requests?**\n\nIf you're working alone, not necessarily. If you're working with other people, probably yes. On GitHub (GitLab calls it a merge request), a pull request is a formal way to propose a change, review it and discuss it before it becomes the new main version. For content teams, that's the discipline that stops important material being changed casually or invisibly.\n\n**Is Git the same as GitHub?**\n\nGit is the version control system; GitHub is one popular service built around it, a place to use Git with other people. You can use Git without GitHub, and there are other Git hosting services such as GitLab and Bitbucket.\n\n**Is Git overkill for content?**\n\nSometimes. If one person edits a handful of pages in one [CMS](https://contentlexicon.com/cms), Git may add more ceremony than value. But where content is shared, reused, translated, reviewed, published across channels or maintained alongside code and structured systems, Git often becomes the cleanest way to manage change.\n\n## Learn more\n\n- [What is Git?](https://git-scm.com/book/en/v2/Getting-Started-What-is-Git%3F) – the Pro Git chapter explaining snapshots, local history and the modified, staged and committed states.\n- [Learn Git Branching](https://learngitbranching.js.org/) – an interactive way to practise commits, branches and merges through visual exercises.\n- [Confusing git terminology](https://jvns.ca/blog/2023/11/01/confusing-git-terminology/) – Julia Evans’s plain-language guide to the git terms that trip people up, from detached HEAD to “ours” and “theirs” and fast-forward." }, { "term": "Glossary", "alsoCalled": [], "slug": "glossary", "definition": "A set of definitions for terms used in a particular subject, organisation or body of content.", "category": "Editorial craft", "related": [ "content-model", "jargon", "taxonomy", "word-list" ], "updated": "2026-10-14T15:51:00.000Z", "url": "https://contentlexicon.com/glossary", "markdown": "https://contentlexicon.com/glossary.md", "credit": "Glossary, from the Content Lexicon by Contentious (https://contentlexicon.com/glossary), CC BY 4.0", "body": "A glossary can explain unfamiliar subject terms for readers or settle language a team uses in its work. These purposes overlap, but a customer guide shouldn't become a catalogue of internal project names.\n\nWriting definitions for the people who'll use them means readers can understand a term without needing another glossary to explain the glossary. A term may need examples or links to related concepts, particularly where everyday language and specialist usage differ. An agreed definition should also have a route for revision when the subject changes.\n\n## Why it matters for content\n\nTerminology shapes thought and work. When a team uses one word consistently, requirements become clearer, interfaces become more coherent and readers are less likely to meet three labels for one action. A glossary also helps new colleagues and partners participate without having to absorb every private phrase by osmosis.\n\nA useful glossary does more than collect definitions. It identifies preferred terms, synonyms, discouraged language, context and ownership. Entries should be searchable, linked to related concepts and reviewed when products or policy change.\n\nA glossary differs from a [word list](https://contentlexicon.com/word-list). The glossary explains concepts; the word list settles spelling, capitalisation and usage. The two can connect, and smaller organisations may combine them, but their questions are distinct.\n\n## How AI changes the picture\n\nAI can extract candidate terms from documents, suggest duplicates and help keep definitions consistent. It can also use a glossary to draft with approved vocabulary, expand unfamiliar terms and improve search across synonyms.\n\nThis makes glossary quality operational. Vague or conflicting definitions may influence generated content at scale. Important entries need human owners and reliable sources rather than a model averaging the organisation’s disagreements into something smooth.\n\nA glossary can also support a conversational explanation: an assistant could recognise an alias and offer the maintained definition, then let a reader ask for an example. That depends on how the glossary is connected to the system. Structured glossaries can support retrieval, [metadata](https://contentlexicon.com/metadata) and [knowledge graphs](https://contentlexicon.com/knowledge-graph). Terms, aliases and relationships help systems connect the language users choose with the concepts the organisation maintains.\n\n## Questions people ask\n\n**Which terms belong in a glossary?**\n\nInclude words that are important, ambiguous, specialised or used inconsistently. Do not reproduce an ordinary dictionary unless the organisation gives a word a specific meaning.\n\n**Who owns the definitions?**\n\nOwnership may be shared between subject experts and content specialists. Someone should have authority to resolve disagreements and maintain each important entry.\n\n**How long should a definition be?**\n\nLong enough to distinguish the concept clearly. Start with a concise definition, then add context, examples and related terms where they help.\n\n**Can AI create the glossary for us?**\n\nIt can propose candidates and draft definitions from sources. People must verify meaning, choose preferred terminology and resolve contradictions.\n\n**Is a lexicon just a grand name for a glossary?**\n\nOften, yes. Strictly, a lexicon is the whole vocabulary of a language or field, while a glossary defines selected terms for particular readers. In practice, a published lexicon tends to be a glossary with ambitions: the Content Lexicon, for instance, is a glossary that couldn't stop at one line per term.\n\n## Learn more\n\n- [Building a living term base](https://lokalise.com/blog/term-base) – practical guidance on definitions, context, ownership and keeping terminology alive rather than embalmed in a spreadsheet.\n- [Content design glossary template](https://www.atlassian.com/software/confluence/templates/content-design-glossary) – A Confluence template with a transferable sequence: identify terms, seek feedback and add usage examples; the surrounding page also promotes the product.\n- [Glossary of editorial and publishing terms](https://www.ciep.uk/resource/editorial-glossary.html) – a real professional glossary worth browsing for how concise definitions distinguish related concepts." }, { "term": "Grammar", "alsoCalled": [], "slug": "grammar", "definition": "The patterns and conventions that govern how words form meaningful sentences.", "category": "Editorial craft", "related": [ "copy-editing", "plain-language", "proofreading", "punctuation", "sentence-case", "style-guide" ], "updated": "2026-10-11T11:46:00.000Z", "url": "https://contentlexicon.com/grammar", "markdown": "https://contentlexicon.com/grammar.md", "credit": "Grammar, from the Content Lexicon by Contentious (https://contentlexicon.com/grammar), CC BY 4.0", "body": "Grammar covers how sentences are formed, how verbs relate to subjects, how clauses connect and what pronouns refer to. People who speak a language already use many of its patterns without having to name them.\n\nEditors learn to describe grammar so they can explain why a sentence is unclear or misleading rather than simply feeling that something has gone awry. This is different from enforcing personal preferences as universal rules.\n\n## Why it matters for content\n\nGrammar matters when it affects meaning, trust or ease of reading. A misplaced modifier can attach an action to the wrong person. An unclear pronoun can leave two possible subjects. Changes in tense or agreement can interrupt understanding. Good [editing](https://contentlexicon.com/editing) removes these obstacles so readers can concentrate on the subject rather than reverse-engineering the sentence.\n\nIt helps to separate grammatical problems that affect meaning from schoolroom folklore. English allows sentences to begin with “And”, infinitives to be split and prepositions to appear at the end. Sometimes those choices make prose clearer and more natural. A [style guide](https://contentlexicon.com/style-guide) should give [writers](https://contentlexicon.com/writer) confidence to use the language deliberately, not hand them a sack of inherited prohibitions.\n\nOrganisations still need decisions where accepted usage varies: singular “they”, collective nouns, technical conventions or differences between British and American English. These are style choices as much as grammar. Decide what suits the audience, record it and avoid turning every edit into a small constitutional crisis.\n\n## How AI changes the picture\n\nAI can identify agreement errors, awkward syntax and inconsistent tense. It can explain alternatives and help people edit in a language they know less well. This is useful assistance, particularly for routine checking and early drafts.\n\nIt can also “correct” intentional [voice](https://contentlexicon.com/voice), regional usage or [inclusive language](https://contentlexicon.com/inclusive-language) towards whatever pattern appears most conventional. Models may confidently enforce myths, flatten rhythm or rewrite a precise sentence into smoother ambiguity. Grammar suggestions therefore need context and editorial judgement.\n\nClear guidance improves AI output. A well-maintained style guide can distinguish firm grammatical requirements from house preferences and give examples of both. Teams should test whether automated checks preserve meaning, names and terminology rather than measuring success by the number of green ticks produced.\n\n## Questions people ask\n\n**Is grammar fixed?**\n\nLanguages change through use, though conventions provide enough stability for people to understand one another. Guidance should reflect the language and setting, rather than treating every inherited rule as permanent.\n\n**What is the difference between grammar and style?**\n\nGrammar describes how language forms meaning. Style concerns choices among valid alternatives, such as [tone](https://contentlexicon.com/tone), spelling conventions and preferred constructions. The boundary can be fuzzy, which is why editors need reasons rather than reflexes.\n\n**Can AI proofread grammar reliably?**\n\nIt can catch many errors, but it may miss ambiguity or change intended meaning. A person should review consequential copy and any suggestion that substantially rewrites the sentence.\n\n**What should we do when references disagree?**\n\nCheck whether they cover different varieties of English, registers or publication dates. For valid alternatives, follow the agreed house preference and record the choice; for a question of meaning, explain the evidence rather than appealing to authority alone.\n\n## Learn more\n\n- [Merriam-Webster grammar and usage](https://www.merriam-webster.com/grammar) – a lively reference for real usage questions rather than a list of institutional preferences.\n- [The Guardian style guide, 2004](https://image.guardian.co.uk/sys-files/Guardian/documents/2004/07/15/styleguidepdfjuly2004.pdf) – an official historical edition explaining the difference between house consistency and imposing one writing style.\n- [On style guides and dictionaries](https://victoryediting.com/on-style-guides-and-dictionaries/) – a book editor on choosing a primary reference and recording justified exceptions; its Chicago/Merriam-Webster defaults are an American publishing perspective." }, { "term": "Graph engineering", "alsoCalled": [], "slug": "graph-engineering", "definition": "Designing an AI system as an explicit map of steps and routes, so work moves between models, tools, agents and people in controlled ways.", "category": "AI and tooling", "related": [ "agentic-ai", "context-engineering", "knowledge-graph", "separation-of-concerns", "workflow" ], "updated": "2026-10-14T22:06:00.000Z", "url": "https://contentlexicon.com/graph-engineering", "markdown": "https://contentlexicon.com/graph-engineering.md", "credit": "Graph engineering, from the Content Lexicon by Contentious (https://contentlexicon.com/graph-engineering), CC BY 4.0", "body": "In graph engineering, each step of an AI system is a node – a model call, a tool, a check or a whole agent – and each edge is a route the work is allowed to take next. A shared record of the work's state travels along those routes, so each step knows what's been done and what's still missing. The design decides where a model can use its judgement and where the system follows fixed rules, such as an approval that every draft must pass before publication. Routes can loop back, so a draft that fails a check returns for revision rather than carrying on regardless.\n\nThe mathematical idea of a graph is much older. In 1736, Leonhard Euler represented Königsberg’s landmasses as points and its seven bridges as connections, then showed that no route could cross every bridge exactly once. His solution is often treated as the beginning of graph theory. Today’s AI graphs use the same basic abstraction, though graph engineering itself draws on [workflow](https://contentlexicon.com/workflow) design and state machines.\n\nThe term surfaced in mid-2026 as the latest in a sequence of AI \"engineerings\". [Prompt](https://contentlexicon.com/prompt) engineering shapes a single exchange, [context engineering](https://contentlexicon.com/context-engineering) decides what information a model sees, and harness and loop engineering organise the tools and repeated attempts around one agent. Graph engineering works at the level of the whole system, coordinating several steps, agents and people. The idea is older than the name, drawing on workflow design and the state machines of conventional software.\n\n## Why it matters for content\n\nMost content work already follows a graph, though it usually lives in people's heads, a process document or a long thread of messages: brief, draft, fact-check, legal [sign-off](https://contentlexicon.com/sign-off), publish, review. Graph engineering turns that route into something a system runs. The benefit is that the route becomes explicit and open to inspection, showing which checks always happen, where a draft goes when it fails and who is asked when the evidence is unclear. The cost is that nothing can be left implied. A step that isn't drawn doesn't happen, and information that isn't carried in the state never reaches the next node.\n\nContent pipelines are a favourite example in graph engineering tutorials, and they show what goes missing when a graph is drawn without a content person in the room. A 2026 Analytics Vidhya guide builds a planner, researcher, [writer](https://contentlexicon.com/writer) and evaluator, with a revision loop and human review. Its state holds a topic, a plan, evidence, a draft and feedback. Nothing in it records who the piece is for or what it should help them do, so the evaluator can check accuracy, grounding, completeness and clarity but not whether the piece will be useful to anyone. The decision to make the piece at all is taken before the graph begins, and the route ends at a node called finalize, with no owner, [review date](https://contentlexicon.com/review-date) or path to retirement. Those are the parts of a content workflow a [content strategist](https://contentlexicon.com/content-strategist) would draw first.\n\nA graph also depends on its words meaning the same thing at every step. If \"approved\" means passed by the evaluator at one node and signed off by legal at another, the routing will follow the wrong path without complaint. The 2026 survey that set out graph engineering as an academic paradigm turns to this problem in its future directions: the same concept can be defined differently across parts of a system, so agent graphs need shared definitions of their entities and relationships. Definitions of done, [glossaries](https://contentlexicon.com/glossary) and approval rules are where [content teams](https://contentlexicon.com/content-team) already do that work.\n\n## How graph engineering changes the picture\n\nGraph engineering moves part of editorial governance from documents into the structure of the system. A [style guide](https://contentlexicon.com/style-guide) can ask writers to get legal approval for health claims; a graph can send every draft flagged as containing one to a legal review node, with no edge that bypasses it. Where a model may use its judgement and where a fixed rule applies becomes a design decision made in advance, rather than something left to the model's discretion or a reviewer's memory. Instructions describe behaviour, but routes and permissions enforce it.\n\nBecause each piece of work can leave a record of the route it took, a team can see where drafts are repeatedly sent back, which check catches most problems and where work waits for a person. That makes the workflow something to improve with evidence. The graph needs maintaining like any other content asset, though. When a review policy changes, the route has to change with it, or the system will go on enforcing last year's out-of-date process with impeccable consistency.\n\nNot every kind of content work suits a fixed map. Open-ended research, where the next step depends on what turns up, can be better served by an agent planning its own route within limits; LangChain moved its own deep research agent from a predefined graph to a more agentic loop for that reason. Graphs suit work with a predictable shape and checkpoints that must always happen. Many [content operations](https://contentlexicon.com/content-operations) contain both, and a single node can hold an agent with freedom to choose its sources while the approvals around it stay fixed.\n\n## Questions people ask\n\n**Is graph engineering a new name for workflow automation?**\n\nPartly. Frameworks such as LangGraph were building agents this way before the term caught on. The difference lies in what sits inside the nodes. When a step is a model call or a whole agent, its output varies from run to run, so the routes have to handle uncertain results through retries, revision loops and escalation to a person.\n\n**What's the connection with knowledge graphs?**\n\nThey're different graphs. A [knowledge graph](https://contentlexicon.com/knowledge-graph) records facts and the relationships between them; an agent graph records steps and the routes between them. The two meet when a step queries a knowledge graph for context. Some writers also use \"graph engineering\" for building knowledge graphs, an older sense of the phrase.\n\n**Do content people need to write code to take part?**\n\nNot usually. Developers tend to implement the routes and state, but many of the design questions are editorial: which checks matter, what a reviewer needs to see, what happens when sources disagree and what follows publication. Those can be settled on a whiteboard before anyone opens a code editor.\n\n**Where does a team start?**\n\nA drawing of the current workflow as it runs in practice, including the checks people make without writing them down, is a useful first step. Marking which steps follow fixed rules, which need judgement and what each step needs to know turns that drawing into a brief a developer can build from. It often improves the human process too, before any agent is involved.\n\n## Learn more\n\n- [3 Years of Graph Engineering with LangGraph](https://www.langchain.com/blog/3-years-of-graph-engineering-with-langgraph) – LangChain's 2026 response to the term's arrival, conceding the buzzword while explaining when graphs help and when an open agent loop works better; written by the company behind LangGraph.\n- [Graph Engineering in the Era of LLM Agents](https://arxiv.org/abs/2608.21156) – a 2026 academic survey that organises the field around task organisation, agent coordination and runtime state, and argues that agent graphs will need shared definitions to hold together.\n- [Graph Engineering for AI Agents: Beyond the Single-Agent Loop](https://www.analyticsvidhya.com/blog/2026/07/graph-engineering) – a hands-on 2026 build of a planner, researcher, writer and evaluator pipeline, worth reading with an editor's eye for what its state leaves out.\n- [How the seven bridges of Königsberg spawned new math](https://www.scientificamerican.com/article/how-the-seven-bridges-of-koenigsberg-spawned-new-math) – a reader-friendly account of Euler’s puzzle and how reducing the city to landmasses and connections led to graph theory." }, { "term": "Guardrails", "alsoCalled": [], "slug": "guardrails", "definition": "Limits and checks intended to keep AI-assisted work within defined bounds.", "category": "AI and tooling", "related": [ "agentic-ai", "ai-governance", "ai-use-policy", "content-governance", "hallucination", "human-in-the-loop", "style-guide", "system-one-models", "system-prompt" ], "updated": "2026-10-11T11:52:00.000Z", "url": "https://contentlexicon.com/guardrails", "markdown": "https://contentlexicon.com/guardrails.md", "credit": "Guardrails, from the Content Lexicon by Contentious (https://contentlexicon.com/guardrails), CC BY 4.0", "body": "A read-only assistant can be prevented from changing records by its tool permissions. An instruction to use approved sources is different: it guides behaviour, but needs supporting retrieval controls and testing if exclusion matters. Both may be called guardrails; they don't provide the same assurance.\n\nGuardrails can sit at different points in the work: some limit what an AI agent can read, some filter what it produces, some constrain what it can do with the result and some pass uncertain cases to a person. A guardrail can be technical, such as a permission that blocks publishing, or procedural, such as a review stage no draft can skip. What makes it a guardrail, rather than a general ambition to be careful, is the specific failure it stands in the way of.\n\n## Why it matters for content\n\nStyle guidance, approval [workflows](https://contentlexicon.com/workflow) and publishing permissions already constrain content work. With AI, teams need to be clear about which limits are guidance, which are technical controls and which rely on a reviewer.\n\nA team might allow an AI agent to draft an answer from public guidance while withholding the ability to publish it. That creates a useful workspace for exploration without treating fluent wording as approval. Checks should cover the risks of the actual task, not an abstract promise to prevent every error.\n\n## How guardrails change the picture\n\nGuardrails change AI from a loose drafting toy into something that can be used responsibly inside [content operations](https://contentlexicon.com/content-operations). They make the difference between “ask it anything and hope” and a workflow where the tool has a defined role, clear limits and review points that match the level of risk.\n\nThey also expose weak human practice. If a team can't explain its [tone](https://contentlexicon.com/tone), source hierarchy, publishing rules or review standards, AI won't magically infer them. It may imitate the surface of existing content, including every inconsistency, outdated claim and fondness for phrases nobody would defend in daylight. Writing guardrails is therefore also a useful audit of the content system itself.\n\n## Questions people ask\n\n**Are guardrails just rules?**\n\nRules are part of it, but guardrails are broader. They include permissions, source limits, workflow checks, review stages, examples, escalation points and quality criteria. A rule says “don't do this”. A guardrail helps the whole system avoid doing it.\n\n**Do guardrails stop AI being creative?**\n\nNot if they are well designed. Good constraints often make creative work better because they clarify the space in which the work can happen. The aim isn't to make every output bland. Constraints can protect facts and permissions while leaving room to try different explanations. They reduce particular risks; they don't guarantee that an output is right.\n\n**Who should write them?**\n\nContent, product, legal, brand, data and operations people may all need a say, depending on the use case. Give each measure an owner who understands the work and the risk, alongside the people who configure it.\n\n**Do guardrails replace review?**\n\nNo. They reduce obvious risk and improve consistency, but they don't replace judgement. The more public, sensitive or consequential the content, the more human review still matters. Guardrails help reviewers spend less time catching avoidable nonsense and more time assessing whether it serves the reader.\n\n## Learn more\n\n- [NIST Generative AI Profile](https://www.nist.gov/publications/artificial-intelligence-risk-management-framework-generative-artificial-intelligence) – NIST’s 2024 voluntary cross-sector profile of AI RMF 1.0, addressing generative-AI risks and actions across the lifecycle.\n- [Building effective agents](https://www.anthropic.com/engineering/building-effective-agents) – Anthropic’s 2024 workflow-versus-agent guidance, with simple architectures and observable feedback; the page itself warns that implementation tooling has changed.\n- [LLM01:2025 Prompt Injection](https://genai.owasp.org/llmrisk/llm01-prompt-injection/) – the top risk in OWASP’s 2025 list for LLM applications, with mitigations including constrained behaviour, input and output filtering, least privilege and human approval for high-risk actions." }, { "term": "Hallucination", "alsoCalled": [ "Confabulation" ], "slug": "hallucination", "definition": "An AI-generated claim that is fabricated, factually wrong or unsupported by the source it purports to describe.", "category": "AI and tooling", "related": [ "ai-slop", "ai-use-policy", "guardrails", "llm", "rag", "system-one-models" ], "updated": "2026-10-10T18:02:00.000Z", "url": "https://contentlexicon.com/hallucination", "markdown": "https://contentlexicon.com/hallucination.md", "credit": "Hallucination, from the Content Lexicon by Contentious (https://contentlexicon.com/hallucination), CC BY 4.0", "body": "A hallucinating model might invent a statistic, cite a paper that doesn't exist, attribute a policy to the wrong organisation or describe a service you stopped offering in 2019. The answer may sound entirely assured, which makes the error harder to spot.\n\nA language model isn't mistaken in the way a person is mistaken; it's producing plausible continuations of text, and a fluent invention is often more plausible than an awkward truth.\n\n## Why it matters for content\n\nThe model, its training, the task and the material it's given all affect the risk of hallucination. [Content teams](https://contentlexicon.com/content-team) can improve the part they control: clear, current sources, unambiguous wording and fewer contradictory versions. These support grounded answers, but they don't guarantee them.\n\nErrors carry particular costs in eligibility, medical and legal information, prices, deadlines and safeguarding. In those areas, a confident wrong answer is worse than no answer at all.\n\n## How hallucination changes the picture\n\nIt changes what review is for. Traditional [editing](https://contentlexicon.com/editing) catches clumsy writing, inconsistency and errors of judgement. Reviewing AI-assisted content also means verifying things that read perfectly: that the quoted figure exists, that the cited source says what it's claimed to say, that the named person holds the role attributed to them. Fluency isn't evidence that an answer is true.\n\nIt also changes what counts as published content. If an assistant answers questions using your pages, [PDFs](https://contentlexicon.com/pdf), help articles and [knowledge base](https://contentlexicon.com/knowledge-base) entries, those sources can shape responses you haven't approved individually. Teams may inspect logged or sampled answers, but can't always review every interaction. Some of the work shifts upstream: fix the sources, retire contradictions and make the canonical version clear.\n\nRetrieval helps considerably – grounding answers in specific documents rather than general training – but it doesn't abolish the problem. Grounded systems still summarise, compress and occasionally stitch two unrelated passages into one plausible hybrid. High-stakes content needs layered controls: approved sources, checks against evidence, testing, appropriate specialist review and a route to abstain or escalate. Human review matters, but it can miss errors too.\n\n## Questions people ask\n\n**Will better models fix this?**\n\nResults vary by model and task. Better training, retrieval and evaluation can reduce errors, but a newer or more accurate model isn't automatically less likely to guess. Test the system on the work you need it to do, including whether it acknowledges missing evidence.\n\n**Is it our fault or the model's?**\n\nOften both, and the half you control is your content. Contradictory, outdated or missing sources can lead to retrieving or generating a misleading answer.\n\n**How do we spot hallucinations?**\n\nCheck the checkable: names, numbers, dates, quotations and citations. Ask for sources, then open them.\n\n**Does this mean we shouldn't use AI for content?**\n\nYou can use it, with controls proportionate to the consequences. Check factual claims against evidence and put specialist review or escalation where an error could cause harm.\n\n## Learn more\n\n- [Why language models hallucinate](https://openai.com/index/why-language-models-hallucinate/) – a clear account of why evaluation can reward guessing over admitting uncertainty.\n- [A survey of hallucination in large language models](https://arxiv.org/abs/2311.05232) – causes, detection, mitigation and open questions, including the limits of grounding answers in retrieved material.\n- [AI Hallucination Cases Database](https://www.damiencharlotin.com/hallucinations) – Damien Charlotin’s running record of court decisions worldwide in which parties relied on hallucinated AI content." }, { "term": "Head of content", "alsoCalled": [], "slug": "head-of-content", "definition": "A senior role responsible for all the content an organisation publishes, and for the people, standards and systems that produce and look after it.", "category": "Roles", "related": [ "centralised-governance", "content-governance", "content-manager", "content-operations-manager", "content-strategist", "content-strategy", "content-team", "senior-content-editor" ], "updated": "2026-10-11T06:24:00.000Z", "url": "https://contentlexicon.com/head-of-content", "markdown": "https://contentlexicon.com/head-of-content.md", "credit": "Head of content, from the Content Lexicon by Contentious (https://contentlexicon.com/head-of-content), CC BY 4.0", "body": "A head of content leads an organisation's content as a whole: the [website](https://contentlexicon.com/website), guidance, publications, newsletters, social channels and, increasingly, the words inside its products and services. They set standards, decide priorities, manage the people and budget that produce the content and answer to senior leadership for whether it works. Most combine an editorial or content background with the authority to make decisions that cross departmental lines.\n\nThe title is common in media, charities, membership bodies and technology companies, and its scope varies widely. In a small organisation the head of content may be a player-coach who still writes and edits; in a large one they may lead several teams of designers, editors, strategists and channel specialists. Related titles include content director, editor-in-chief and, at board level, chief content officer.\n\n## Why an organisation needs a head of content\n\nIn many organisations, content is produced in many places and owned in none. Communications runs news and campaigns, digital runs the website, policy writes guidance, fundraising or marketing writes appeals, and each team has its own idea of quality and [voice](https://contentlexicon.com/voice). The result is duplication, contradiction and a website nobody can fix because nobody has the authority to. A head of content gives the entire estate a single senior owner: someone who can decide what gets published and what gets retired, set standards everyone follows and make the case for investment in terms leadership understands.\n\nThe role also gives content a voice where decisions are made. Without one, content is often what gets asked for at the end of a project, after the service has been designed and the budget spent. A head of content can argue earlier for the audience, for the cost of maintaining what's published and for the systems content depends on. The need is usually clearest after a period of growth, a merger or a website rebuild, or when a strategist has set a direction and nobody senior is left to hold it.\n\n## How AI changes the role\n\nAI raises the stakes on decisions only a senior content leader can make. When leadership asks whether assistants can cut the content budget, someone needs to answer with evidence about where AI speeds work up and where it adds review, risk and maintenance. When a chatbot starts answering questions in the organisation's name, its answers are content, and the head of content is often the person accountable for whether they're accurate, consistent and in the right voice. That draws the role into procurement, risk and governance conversations that used to belong to technology teams.\n\nAI also changes the team a head of content leads. With drafting cheaper, they may move budget and headcount from production towards [editing](https://contentlexicon.com/editing), structuring and checking, and need to make that case before the savings are claimed elsewhere. The bigger danger is volume: when anyone can publish more, the organisation needs someone with the standing to say no. A job description might now ask for experience of leading teams through technological change, setting organisation-wide standards for AI-assisted content and working with legal, data and technology colleagues on risk.\n\n## Questions people ask\n\n**How is a head of content different from a head of communications?**\n\nA head of communications looks after what the organisation wants to say: its reputation, media relations, campaigns and internal messages. A head of content looks after what people need to find, understand and use, including the content that will never feature in a press release, such as guidance, help pages, service information and reference material. Communications sends messages out at particular moments; content has to serve people whenever they arrive. Many organisations need both, and the head of comms title often reflects an older model in which the website was one more channel for announcements.\n\n**Should we rename our head of communications?**\n\nOnly if the remit changes with the name. A new title can be a useful signal: when a head of communications becomes head of content, colleagues start to see the job as looking after everything the organisation publishes, including the guidance and service information that never made the news. A title on its own changes little, though. The role needs the authority, budget and team to match, and communications work, from press enquiries to crisis handling, still needs an owner.\n\n**How is a head of content different from a content strategist or content manager?**\n\nA strategist sets the direction and a [content manager](https://contentlexicon.com/content-manager) runs the system day to day; a head of content is accountable for both and for the people doing them. In smaller organisations one person may do all three. In larger ones the head of content usually manages strategists, managers and editors, and spends most of their time on priorities, budget, standards and relationships with other senior leaders.\n\n**Where should a head of content sit?**\n\nIn, or close to, the leadership team, whether the role reports to a director of communications, digital or engagement, or to the chief executive. What matters is that the reporting line gives authority across departments. A head of content placed inside a single team will tend to look after that team's content and little else.\n\n**Do we need a chief content officer?**\n\nPossibly, if content is a large operation in its own right, with several teams, brands or languages and a budget to match. Chief content officer is the board-level version of the role, found mostly in media businesses and larger companies. Many organisations are better served by a head of content with real authority. The abbreviation doesn't help either: CCO more often means chief communications officer, which brings back the very confusion the title was meant to resolve.\n\n## Learn more\n\n- [The New Chief Content Officer: Why the Job Description for the CCO Position Must Change](https://contentmarketinginstitute.com/career-development/the-new-chief-content-officer-why-the-job-description-for-the-cco-position-must-change) – the Content Marketing Institute's job description for the most senior content role, which argues that the function matters more than the title, whether that's chief content officer, vice president of content or head of content.\n- [Chief Content Officer in 2026: Content leadership when AI accelerates everything](https://www.toaststudio.com/en/articles/the-role-of-chief-content-officer/) – Toast Studio on content leadership as ownership of the content operating model, with a useful comparison of content, marketing, communications and digital leaders and a section on deciding where AI belongs.\n- [Head of Content and Channels job description](https://www.mssociety.org.uk/sites/default/files/2025-10/JD%20-%20Head%20of%20Content%20and%20Channels.pdf) – the MS Society's job description, which combines organisation-wide content strategy, editorial standards, governance and line management; one charity's version, but a realistic model for a non-profit." }, { "term": "Headless CMS", "alsoCalled": [], "slug": "headless-cms", "definition": "A content management system that stores content separately from the places where that content is presented.", "category": "Structure and systems", "related": [ "api", "backend", "cms", "create-once-publish-everywhere", "frontend", "separation-of-concerns", "structured-content" ], "updated": "2026-10-07T17:50:00.000Z", "url": "https://contentlexicon.com/headless-cms", "markdown": "https://contentlexicon.com/headless-cms.md", "credit": "Headless CMS, from the Content Lexicon by Contentious (https://contentlexicon.com/headless-cms), CC BY 4.0", "body": "In a traditional [CMS](https://contentlexicon.com/cms), the content and the [website](https://contentlexicon.com/website) [frontend](https://contentlexicon.com/frontend) are tightly coupled. In a headless CMS, the content is managed in one place and then delivered through an [API](https://contentlexicon.com/api) to whatever needs it: a website, an app, a kiosk, a chatbot or something else.\n\nThe term can sound more technical than it needs to. For [content teams](https://contentlexicon.com/content-team), it means content has to stand on its own. It can't rely on being hand-shaped inside one page template or one editor.\n\nMany vendors now avoid leading with the word _headless_: Sanity tends to call itself a content platform, and Contentful talks about content infrastructure. The idea hasn't gone away. The market has matured, and the question has moved on from “is it headless?” to whether the platform helps teams model, reuse and deliver content without every channel needing its own hand-built copy.\n\n## Why it matters for content\n\nA headless CMS encourages [structured content](https://contentlexicon.com/structured-content), reuse and clearer separation between content and presentation. That can make it easier to publish across several channels, redesign a frontend without rebuilding the content layer and maintain one source rather than several copied versions.\n\nThe case is strongest where the same content needs to appear in more than one place, where several channels are involved or where content is treated as a long-term asset rather than something written directly into one website page.\n\n## How AI changes the picture\n\nAI strengthens the case for headless systems, provided the content inside them is properly structured. Headless describes how content is delivered; structure describes how it's modelled. A headless CMS full of long rich-text fields is no easier for a machine to reuse than a web page. If content is broken into meaningful fields and managed independently of presentation, it becomes much easier to draft against, retrieve from and use in assistants, summaries and recommendations.\n\nSeparating source from output matters more, too. The more content is transformed for different uses, the more important it becomes to know what the canonical content is and where its structure lives.\n\n## Questions people ask\n\n**Is a headless CMS always better than a traditional CMS?**\n\nNot for every team. The extra flexibility can also add complexity. The right choice depends on the channels, team and publishing needs.\n\n**Do content people need to understand APIs to use one?**\n\nNot necessarily. They do need to understand structured content more clearly, because the system usually exposes the consequences of weak structure.\n\n**Does headless mean there's no website?**\n\nThere's usually still a website. It's built separately from the content management layer and pulls content from it.\n\n**Is a headless CMS only for big organisations?**\n\nSmaller teams use them too, but a headless CMS tends to be most useful when content is reused across channels or when flexibility matters enough to justify the setup.\n\n## Learn more\n\n- [Headless content management system](https://en.wikipedia.org/wiki/Headless_content_management_system) – a short neutral account of what headless means, including the drawback that setup takes more development effort.\n- [Headless CMS explained in one minute](https://www.contentful.com/headless-cms/) – covers headless versus decoupled systems and the create once, publish everywhere principle, before ending in Contentful's composable-platform pitch.\n- [How and why to move from WordPress to Astro](https://contentious.ltd/resources/how-and-why-to-move-from-wordpress-to-astro) – a practical account of leaving a monolithic CMS for a modern stack, with AI helping along the way." }, { "term": "Hub and spoke", "alsoCalled": [ "Federated publishing" ], "slug": "hub-and-spoke", "definition": "A federated governance model in which a central content team sets shared standards while local teams create and manage content.", "category": "Governance", "related": [ "author-experience", "centralised-governance", "content-governance", "content-operating-model", "content-team", "decentralised-governance", "silo" ], "updated": "2026-10-12T16:23:00.000Z", "url": "https://contentlexicon.com/hub-and-spoke", "markdown": "https://contentlexicon.com/hub-and-spoke.md", "credit": "Hub and spoke, from the Content Lexicon by Contentious (https://contentlexicon.com/hub-and-spoke), CC BY 4.0", "body": "Hub and spoke describes how many organisations divide content work: a small central team, the hub, supports and sets standards for the many teams across the organisation, the spokes, that publish in their own areas. It sits between [centralised governance](https://contentlexicon.com/centralised-governance), where one team controls most publishing, and [decentralised governance](https://contentlexicon.com/decentralised-governance), where each part of the organisation decides for itself. Most large organisations end up working this way whether or not they use the name, and much of their governance pain lives in the joins.\n\nThe hub usually provides standards and a [style guide](https://contentlexicon.com/style-guide), training, the publishing platform and templates, shared [components](https://contentlexicon.com/component), specialist help with research, [accessibility](https://contentlexicon.com/accessibility) and search, and [sign-off](https://contentlexicon.com/sign-off) for high-risk content. The spokes own the subject knowledge: planning content for their area, drafting it, keeping it accurate and retiring it when it's no longer needed. The name comes from transport networks, where routes radiate from a central hub.\n\n## Why it matters for content\n\nThe model keeps local knowledge and speed while giving the organisation enough common practice to behave like one organisation. Distributed publishing happens anyway. The communications team owns the homepage, service teams own guidance, human resources owns internal policy and enthusiastic colleagues own several pages everyone has forgotten. The governance problem is distribution without an agreement.\n\nThe usual failures come from one side doing too much or too little. A hub that reviews everything becomes a bottleneck, and spokes find ways round it, publishing in shared drives or on social channels instead. A hub that only issues rules becomes a distant authority whose standards nobody follows. Spokes without training or time drift apart in quality, [tone](https://contentlexicon.com/tone) and structure, producing duplicated pages and abandoned sections.\n\nThe model works when both sides have defined duties, the hub is funded to support rather than police, and spoke roles are part of people's jobs rather than an extra on top.\n\n## How AI changes the picture\n\nAI can strengthen the model by making shared guidance easier to apply. A central team might provide approved tools, [prompt](https://contentlexicon.com/prompt) patterns, quality checks and source collections while local teams use them for their own content.\n\nIt can also magnify inconsistency. If every department builds an assistant from different copies of policy and guidance, users receive different organisational truths. Federated AI needs common standards for sources, ownership, evaluation and escalation, with room for local subject knowledge.\n\n## Questions people ask\n\n**Is hub and spoke the same as a hybrid model?**\n\nOften, yes. Nielsen Norman Group's hybrid model, where a central team controls top-level pages and distributed editors publish within its guidelines, is one version of hub and spoke under another name.\n\n**When does a spoke need its own content specialist?**\n\nWhen it publishes enough, or carries enough risk, that occasional contributors can't keep up. Many organisations place trained [content designers](https://contentlexicon.com/content-designer) in the largest spokes and leave smaller ones with part-time editors supported by the hub.\n\n**Is hub and spoke cheaper than a central team?**\n\nNot automatically. It distributes work but still requires investment in training, support and oversight. Unfunded spokes tend to produce abandoned pages.\n\n**How do you know it's working?**\n\nLook for consistent quality, clear ownership, fewer duplicated pages and faster decisions. Also ask the spokes whether the hub helps them publish well or mostly sends them documents.\n\n## Learn more\n\n- [Content Management on Intranets: Centralized, Distributed, and Hybrid Models](https://www.nngroup.com/articles/intranet-content-management) – Maria Rosala and Anna Kaley's comparison of three ways to divide content ownership, with the benefits and drawbacks of each.\n- [Content Governance Models](https://www.upanup.com/article/content-governance-models) – an agency overview of four common models, including hub and spoke and a discipline-based matrix, and the kinds of organisation each suits.\n- [Establishing the Federated Hub and Spoke Governance Model in Microsoft 365](https://www.gravityunion.com/blog//m365-governance-hub-and-spoke) – a consultancy's example of central standards and delegated responsibility in Microsoft 365; adjacent infrastructure practice to adapt." }, { "term": "Human in the loop", "alsoCalled": [ "HITL" ], "slug": "human-in-the-loop", "definition": "A workflow where a person reviews, decides or approves part of an AI-assisted process instead of leaving it fully automated.", "category": "Operations", "related": [ "agentic-ai", "ai-governance", "ai-use-policy", "automated-testing", "content-governance", "guardrails", "sign-off", "system-one-models", "workflow" ], "updated": "2026-10-07T13:53:00.000Z", "url": "https://contentlexicon.com/human-in-the-loop", "markdown": "https://contentlexicon.com/human-in-the-loop.md", "credit": "Human in the loop, from the Content Lexicon by Contentious (https://contentlexicon.com/human-in-the-loop), CC BY 4.0", "body": "The loop might involve checking a draft, approving a classification, reviewing a summary, deciding whether something can be published or stepping in when a system is uncertain.\n\nThe loop is less about having a _human_ present than about where _judgement_ gets applied. Keeping someone involved at every stage can make a process needlessly slow. Removing people entirely lets mistakes reach readers unchecked. A good human-in-the-loop [workflow](https://contentlexicon.com/workflow) identifies the moments where context, responsibility and editorial sense matter most.\n\n## Why it matters for content\n\nContent work is full of decisions that look simple until something goes wrong. Is this claim accurate? Is the source current? Will this wording harm or exclude people? Does the answer follow policy, or merely sound as though it does? These aren't always tasks a model, rule or confidence score can settle safely.\n\nHuman review is especially important where content carries legal, financial, medical, safeguarding or reputational risk. It also matters when the work depends on subtle knowledge of an audience, organisation or situation. The aim is proportionate oversight: intensive checks for consequential material, lighter sampling for routine low-risk work and clear escalation when the system meets something outside its competence.\n\n## How AI changes the picture\n\nAI can draft, classify and rewrite content in volumes no manual team could match. That scale changes the review problem. Reading every generated word may cancel out the benefit, while reviewing nothing lets one bad source or instruction repeat across hundreds of outputs before anyone notices.\n\nA useful loop therefore needs more than a final approval box. It needs defined risk levels, reliable source material, criteria for review and a way for people to correct the system rather than repeatedly patch its outputs. Interfaces should show what the AI used, what it changed and where it's uncertain. Reviewers need permission to stop the process, not merely wave it through.\n\nHuman involvement also creates evidence. Corrections, rejections and escalations reveal where [prompts](https://contentlexicon.com/prompt), knowledge sources, policies or training are weak, and once someone starts logging them the same few causes tend to turn up again and again. If those signals are recorded and acted on, the loop improves the wider content system. If they disappear into private edits, people remain an expensive sticking plaster.\n\n## Questions people ask\n\n**Does human in the loop mean a person checks everything?**\n\nNo. It means deciding deliberately where human judgement belongs. High-risk outputs may need line-by-line approval, while low-risk tasks may need sampling, exception handling or periodic review.\n\n**Is human review automatically reliable?**\n\nNot by default. People skim, misunderstand and develop automation bias, the habit of trusting the machine's answer because it came from the machine. Reviewers need time, evidence and clear criteria.\n\n**Can confidence scores decide when to escalate?**\n\nThey can help, but confidence isn't the same as correctness or consequence. Escalation rules should consider the type of content, the audience and the harm an error could cause.\n\n**Who is responsible when AI-assisted content is wrong?**\n\nResponsibility remains with the organisation using and publishing it. A model can't hold editorial accountability, however fluent its apology may be.\n\n## Learn more\n\n- [NIST AI Risk Management Framework](https://www.nist.gov/itl/ai-risk-management-framework) – AI RMF 1.0 (2023), organised around governing, mapping, measuring and managing AI risk. As of October 2026, NIST is revising the framework as part of the US AI Action Plan.\n- [ICO human-review checklist](https://ico.org.uk/for-organisations/advice-and-services/audits/data-protection-audit-framework/toolkits/artificial-intelligence/human-review/) – the Information Commissioner's Office checklist for making human review of AI meaningful rather than a rubber stamp, with automation bias, training and caseloads covered. As of October 2026, the guidance is under review following the Data (Use and Access) Act.\n- [How to use AI for better content](https://contentious.ltd/resources/how-to-use-ai-for-better-content) – practical ways to use AI as an assistant, critic and extra pair of eyes rather than an unsupervised content tap." }, { "term": "Inclusive language", "alsoCalled": [], "slug": "inclusive-language", "definition": "Language that avoids exclusion, stereotype and unnecessary assumptions about the people being addressed.", "category": "Editorial craft", "related": [ "accessibility", "plain-language", "style-guide" ], "updated": "2026-10-11T06:25:00.000Z", "url": "https://contentlexicon.com/inclusive-language", "markdown": "https://contentlexicon.com/inclusive-language.md", "credit": "Inclusive language, from the Content Lexicon by Contentious (https://contentlexicon.com/inclusive-language), CC BY 4.0", "body": "Inclusive language means making deliberate choices about how gender, disability, ethnicity, age, class, sexuality, religion and other aspects of identity are referred to, and avoiding assumptions about who the reader is, what they know or what kind of life they lead.\n\nIn content work, inclusive language belongs in everyday editorial practice, not in a worthy side document nobody uses. It sits alongside [plain language](https://contentlexicon.com/plain-language), [accessibility](https://contentlexicon.com/accessibility) and style guidance because it shows up in ordinary decisions: examples, defaults, pronouns, metaphors, labels and what kinds of people the content seems to imagine.\n\n## Why it matters for content\n\nInclusive language helps more people recognise themselves in the content and avoid being needlessly pushed away by it. A form that offers only Mr, Mrs and Miss, or asks for a “Christian name”, tells some readers in one line that the page wasn't written with them in mind. Inclusive language also helps organisations avoid wording that's dated, patronising or casually exclusionary.\n\nService content, public-facing guidance, recruitment, policy and education carry the most weight, along with any material speaking to a broad or mixed audience. Internal writing counts just as much, because it shapes culture as well as communication.\n\n## How AI changes the picture\n\nModels tend to reflect patterns that were common in their training data, including stale assumptions and biased defaults. Ask for an example household on a benefits page and a model may reach for a married couple with two children; ask for a job advert and it may offer a “young, dynamic team”. Those defaults can pass into a draft without anyone in the team choosing them, and a busy editor can wave them through without noticing.\n\nExplicit guidance pays off here. Once it's written down and structured, it can steer the tools; if it stays vague, the machine falls back on the wider internet's habits rather than yours.\n\n## Questions people ask\n\n**Is inclusive language just about avoiding offence?**\n\nNo. It's also about clarity, accuracy, respect and making sure content doesn't exclude the people it's meant to help.\n\n**Does inclusive language mean using awkward official terms?**\n\nNot necessarily. It means choosing language deliberately, informed by how people describe themselves and by the context you're writing in.\n\n**Won't this change all the time?**\n\nSome guidance does change over time, which is one reason it needs active ownership rather than a static once-and-done policy.\n\n**Is inclusive language separate from accessibility?**\n\nThey aren't the same thing, but they overlap strongly. Both are concerned with whether content works for the people it's meant to serve.\n\n## Learn more\n\n- [Conscious Style Guide](https://consciousstyleguide.com/about/) – Karen Yin explains conscious language as contextual judgement, including how to weigh competing guidance.\n- [The Conscious Style Guide: A Flexible Approach to Language That Includes, Respects, and Empowers](https://karenyin.com/books/csg-book/) – Karen Yin's book on disagreement, changing norms and making choices thoughtfully.\n- [Inclusive language: words to use and avoid when writing about disability](https://www.gov.uk/government/publications/inclusive-communication/inclusive-language-words-to-use-and-avoid-when-writing-about-disability) – UK Disability Unit guidance on terms to use and avoid, updated in June 2026. A jurisdiction-specific reference, not a universal list of preferred terms." }, { "term": "Information architecture", "alsoCalled": [], "slug": "information-architecture", "definition": "The way information is grouped, labelled and connected so that people can find it and make sense of it.", "category": "Structure and systems", "related": [ "card-sorting", "content-ecosystem-map", "content-model", "findability", "paddle-swim-dive", "taxonomy", "wireframe" ], "updated": "2026-10-09T11:21:00.000Z", "url": "https://contentlexicon.com/information-architecture", "markdown": "https://contentlexicon.com/information-architecture.md", "credit": "Information architecture, from the Content Lexicon by Contentious (https://contentlexicon.com/information-architecture), CC BY 4.0", "body": "Information architecture is the way content is grouped, labelled and connected so that people can find what they need and understand what they've found. Abby Covert puts it more broadly as the way the parts of something are arranged to make it understandable as a whole. It's often mistaken for the navigation menu, but the menu is only the most visible part. A museum might organise the same collection by artist, period and material so visitors can explore by more than one route, and its floor plan, room names and object labels all belong to the same architecture. On a [website](https://contentlexicon.com/website), search, browsing and related links are different expressions of the structure, not substitutes for deciding how the information connects.\n\nThe term came from an architect. Richard Saul Wurman popularised it in his address to the American Institute of Architects conference in 1976, applying an architect's concern with organised space to making complex information clear, in work closer to what's now called information design. The web gave the term its working meaning: Louis Rosenfeld and Peter Morville, who came from library science, published Information Architecture for the World Wide Web in 1998 (the \"polar bear book\", after its cover) and built the practice around organisation, labelling, navigation and search systems. Covert's How to Make Sense of Any Mess (2014) took the idea back beyond websites to anyone arranging information for other people.\n\n## Why it matters for content\n\nContent that can't be found is, from the reader's point of view, content that doesn't exist. Information architecture is the difference between an estate people can navigate confidently and one they endure, and it decides whether related content behaves like a coherent body of knowledge or a set of unconnected pages that happen to share a logo. It also sets context before anyone reads a word of body text. As Jorge Arango points out, a navigation bar listing current accounts, savings, credit cards and mortgages tells you you're on a bank's website without saying so.\n\nGood architecture uses the language of the people using it rather than the organisation's internal vocabulary. Internal labels make readers translate the organisation's structure before they can find an answer. Sometimes that structure is the subject, as when someone wants a particular department's contact details, so the test is the need in front of you rather than a blanket rule against the organisation chart. During a redesign or migration, test the proposed grouping before moving everything into it. An organisational restructure is a reason to review the architecture; it isn't automatically a reason to reproduce the new hierarchy on the site.\n\n## How AI changes the picture\n\nAI hasn't abolished the need for structure, despite hopeful claims that search and assistants make navigation irrelevant. It has changed where structure pays off. Clear hierarchies, consistent labels, sensible [taxonomies](https://contentlexicon.com/taxonomy) and canonical pages all help AI systems retrieve the right material and represent it accurately, though a tidy hierarchy doesn't guarantee a correct generated answer. For readers, search changes most, moving from lists of pages towards synthesised answers. Navigation changes less. A chat box assumes people know what to ask for and what things are called, which is often what they don't know when they arrive, and a menu that rebuilt itself for every visitor would lose the shared context a stable one provides.\n\nAI also makes the grouping work faster. It can cluster large content sets, propose candidate categories, spot orphan pages, suggest alternative labels and check whether existing labels match the words people search for. On large estates that's a real gain, but the output is a set of hypotheses for testing rather than a finished structure.\n\nAI can help compare the consequences of proposed structures, but it doesn't carry the authority to settle organisational disputes. Information architecture decisions are frequently political: whose content appears in the top navigation, which team's language wins, what gets demoted to a subsection. Dan Brown notes that every structural choice sets a precedent, which is why one team's place in a \"quick links\" panel soon becomes every team's request. Those decisions need people who can carry the consequences, ideally with evidence from [card sorting](https://contentlexicon.com/card-sorting) or testing rather than seniority.\n\n## Questions people ask\n\n**How do you test an information architecture?**\n\nCard sorting and tree testing are the standard methods, alongside watching people attempt real tasks. Card sorting explores how people group things and is most useful early, while tree testing checks whether they can find things in a proposed hierarchy before any design exists. Neither is automatically cheap, and both depend on well-designed tasks and a sensible participant sample.\n\n**How is information architecture different from a taxonomy?**\n\nA taxonomy is one [component](https://contentlexicon.com/component) of it: the controlled set of terms used to classify content. Information architecture also covers navigation, page relationships, labels, URLs and search, and the decisions about how all of these fit together.\n\n**Do good search and AI assistants remove the need for structure?**\n\nThey depend on it. Search works better on well-structured, well-labelled content, and assistants retrieve more reliably from it. Many people still browse to orient themselves before searching, particularly when they don't yet know the right words.\n\n**Should a website mirror the organisation chart?**\n\nOnly where the organisation is what people are looking for, such as identifying who runs a service. For finding help by circumstance or topic, a different grouping usually works better. Test the routes instead of assuming one model fits every task.\n\n## Learn more\n\n- [A Brief History of Information Architecture](https://journalofia.org/volume3/issue2/03-resmini/) – Andrea Resmini and Luca Rosati trace the term from 1960s IBM papers and Wurman to Rosenfeld and Morville, separating the information design and library science traditions behind it.\n- [How to Make Sense of Any Mess](https://www.howtomakesenseofanymess.com/) – the free online edition of Covert's book, which treats information architecture as a skill for anyone and works through intent, structure and taxonomy in short lessons.\n- [Everyday Information Architecture](https://everydayia.com/) – the site for Lisa Maria Marquis's practical book (first published under the name Lisa Maria Martin), with its introduction and links to buy it in print or as an ebook. The book covers sitemaps, wayfinding and taxonomies for people who don't call themselves information architects, with a strong thread on the ethics of how information is organised.\n- [Information Architecture Lenses](https://medium.com/eightshapes-llc/information-architecture-lenses-8a3929a106dd) – Dan Brown on questioning structural decisions from different angles, including the precedent each one sets and how navigation teaches people what a product does.\n- [The Impact of AI on Front-end Information Architecture](https://jarango.com/2024/03/31/the-impact-of-ai-on-front-end-information-architecture) – Jorge Arango on how AI is likely to change search, navigation and labelling, and why he's wary of menus that rebuild themselves for each user." }, { "term": "Jargon", "alsoCalled": [], "slug": "jargon", "definition": "Specialist language that can make communication efficient for insiders and difficult for other readers.", "category": "Editorial craft", "related": [ "plain-language", "readability" ], "updated": "2026-10-14T16:17:00.000Z", "url": "https://contentlexicon.com/jargon", "markdown": "https://contentlexicon.com/jargon.md", "credit": "Jargon, from the Content Lexicon by Contentious (https://contentlexicon.com/jargon), CC BY 4.0", "body": "Among people who share the same knowledge, “[content model](https://contentlexicon.com/content-model)” is quicker than explaining the whole concept every time. The problem begins when the audience has not been let in on the code.\n\nJargon includes technical terms, fashionable business language, internal project names and acronyms. Some terms are necessary because they name something precise. Others make ordinary ideas sound grander and the speaker harder to question. The test isn't whether a word is specialised, but whether it helps this audience understand and act.\n\n## Why it matters for content\n\nUnexplained jargon makes readers work harder and can exclude the people who most need the information. It assumes knowledge, hides responsibility and creates false distance between an organisation and the people it serves. In instructions or services, that can lead to mistakes rather than mere irritation.\n\n[Plain language](https://contentlexicon.com/plain-language) doesn't mean removing every technical term. Sometimes people need to learn the official name to complete a task, discuss a diagnosis or understand a right. Explain it on first use, connect it to familiar language and use it consistently afterwards.\n\nInternal jargon matters too. When teams use different names for the same [component](https://contentlexicon.com/component), audience or process, misunderstandings become requirements and then expensive software. A [glossary](https://contentlexicon.com/glossary) and [word list](https://contentlexicon.com/word-list) can turn private shorthand into shared terminology.\n\n## How AI changes the picture\n\nAI can identify obscure terms, suggest plain alternatives and adapt explanations for different levels of knowledge. It can also reproduce corporate jargon and introduce AI jargon of its own. Checking a proposed alternative with the relevant source reduces the risk that replacing a specific term makes the sentence easier and the instruction wrong.\n\nClear [prompts](https://contentlexicon.com/prompt) and style guidance should define the audience, ban known clichés and require unfamiliar terms to be explained. Teams still need to check that simplification has not removed technical precision or replaced one unfamiliar phrase with another.\n\nJargon also affects retrieval. If users search with everyday words while source content uses only internal terminology, people and AI systems may miss the answer. Good [metadata](https://contentlexicon.com/metadata) and synonyms help bridge the gap without forcing readers to speak like the organisation chart.\n\n## Questions people ask\n\n**Is all specialist language jargon?**\n\nNot necessarily. A precise term may be essential within a field. It becomes a problem when it's unnecessary, unexplained or used with the wrong audience.\n\n**Are acronyms always bad?**\n\nNo, but they can often concentrate jargon and exclude people. Spell them out on first use unless the shortened form is definitely more familiar to the audience.\n\n**Does plain language oversimplify complex ideas?**\n\nIt shouldn’t. Good plain language preserves the idea while removing avoidable difficulty. Complexity deserves explanation, not camouflage.\n\n**Can a glossary solve jargon?**\n\nIt helps, but readers should not need to consult one for every sentence. Explain essential terms where they appear and remove the decorative ones.\n\n## Learn more\n\n- [Politics and the English Language](https://www.orwellfoundation.com/the-orwell-foundation/orwell/essays-and-other-works/politics-and-the-english-language) – Orwell’s sharp, enduring argument against stale phrases, inflated diction and language that conceals thought.\n- [Terminology management](https://www.rws.com/glossary/terminology-management) – a multilingual terminology perspective from an AI solutions company, including consistent definitions and managed usage.\n- [The Effects of Jargon on Processing Fluency, Self-Perceptions, and Scientific Engagement](https://journals.sagepub.com/doi/abs/10.1177/0261927X20902177) – Shulman and colleagues’ 2020 experiments showing that jargon disrupts processing and engagement even when definitions are provided." }, { "term": "Knowledge base", "alsoCalled": [], "slug": "knowledge-base", "definition": "A structured collection of useful knowledge that people and systems can search, link, maintain and learn from.", "category": "Structure and systems", "related": [ "content", "content-inventory", "knowledge-graph", "mcp", "metadata", "rag", "taxonomy" ], "updated": "2026-10-10T17:43:00.000Z", "url": "https://contentlexicon.com/knowledge-base", "markdown": "https://contentlexicon.com/knowledge-base.md", "credit": "Knowledge base, from the Content Lexicon by Contentious (https://contentlexicon.com/knowledge-base), CC BY 4.0", "body": "A knowledge base might hold policies, guidance, research, decisions, product information, examples, definitions or project knowledge.\n\nA useful knowledge base has a clear purpose, an audience, a recognisable structure and people responsible for keeping it trustworthy. A neglected knowledge base does not preserve organisational memory so much as preserve several competing memories and leave the reader to referee.\n\n## Why it matters for content\n\nA good knowledge base helps people answer questions, complete tasks and make decisions without hunting through inboxes or asking the same colleague again. It reduces duplicated writing and makes important knowledge less dependent on who happens to be available. For customers, it can make support faster and more consistent. For internal teams, it can make policies, decisions and working methods easier to find and apply.\n\nIts value depends on content practice: clear scope, useful [information architecture](https://contentlexicon.com/information-architecture), consistent [metadata](https://contentlexicon.com/metadata), readable entries, links between related ideas and visible ownership. Maintenance matters as much as publication. Search cannot rescue inaccurate content, and tidy categories cannot rescue a policy nobody has reviewed since three reorganisations ago.\n\nA knowledge base should be treated as a product rather than a storage project. That means understanding what people seek, measuring whether they find it, learning from failed searches and retiring material that no longer earns its place.\n\n## How AI changes the picture\n\nAI makes knowledge bases more operational. Instead of asking people to browse a hierarchy or guess a keyword, an assistant can retrieve relevant material, summarise it and answer a question in natural language. This can make large collections much easier to use, particularly when the answer spans several sources.\n\nIt also invites a different kind of use. A product designer could explore earlier decisions, compare the reasons behind two approaches and return to the recorded evidence as a new question develops. A well-kept knowledge base preserves the difference between an agreed decision, an abandoned proposal and an unresolved discussion, and an assistant can help someone explore those differences without silently settling them.\n\nIt also raises the stakes of weak content. An AI system can distribute an outdated instruction faster and more confidently than a person stumbling across an old page. Clear canonical sources, dates, ownership and relationships therefore become essential. Content needs enough structure for systems to distinguish policy from discussion, current guidance from history and evidence from opinion.\n\nAI can help classify new material, suggest links, identify duplication and flag likely gaps or stale pages. It should not decide which organisational knowledge is true. Human owners still need to resolve conflicts, approve consequential changes and decide what belongs in the collection. The machine can tend the paths; someone must still know where they lead.\n\n## Questions people ask\n\n**Is a knowledge base the same as an intranet?**\n\nNot necessarily. An intranet is a wider internal digital environment. A knowledge base is a maintained collection designed around finding and using knowledge, and it may be public, internal or embedded in another service.\n\n**Do we need a taxonomy before we begin?**\n\nYou need enough structure to prevent immediate chaos, but not a perfect universal scheme. Start with real user questions and improve the [taxonomy](https://contentlexicon.com/taxonomy) as patterns emerge.\n\n**Can AI organise the knowledge base for us?**\n\nIt can suggest labels, links and clusters. People still need to agree purpose, authority, ownership and the meaning of important categories.\n\n**What should not go in a knowledge base?**\n\nUnresolved chat, duplicated files and material with no continuing value should not enter automatically. Capture useful knowledge, not every trace of work that produced it.\n\n## Learn more\n\n- [As We May Think](https://www.theatlantic.com/magazine/archive/1945/07/as-we-may-think/303881) – Vannevar Bush's 1945 proposal for associative trails through recorded knowledge: a historical perspective on linking ideas rather than merely storing documents.\n- [Knowledge management best practices for 2025](https://digitalworkplacegroup.com/knowledge-management-best-practices) – Digital Workplace Group on governance, knowledge graphs and semantic search.\n- [KCS v6 Practices Guide](https://library.serviceinnovation.org/KCS/KCS_v6/KCS_v6_Practices_Guide) – the Consortium for Service Innovation’s Knowledge-Centered Service method, in which support teams capture, structure, reuse and improve knowledge as part of solving each request." }, { "term": "Knowledge graph", "alsoCalled": [], "slug": "knowledge-graph", "definition": "A network of connected concepts, entities and relationships.", "category": "AI and tooling", "related": [ "knowledge-base", "metadata", "rag", "schema", "structured-content", "taxonomy" ], "updated": "2026-10-14T21:02:00.000Z", "url": "https://contentlexicon.com/knowledge-graph", "markdown": "https://contentlexicon.com/knowledge-graph.md", "credit": "Knowledge graph, from the Content Lexicon by Contentious (https://contentlexicon.com/knowledge-graph), CC BY 4.0", "body": "A knowledge graph represents information as entities, such as people, organisations, places, services or documents, connected by named relationships. Each entity has a stable identifier, so the same thing can be recognised wherever it appears. Each relationship has a type, such as \"is provided by\", \"is governed by\" or \"replaces\". People and software can then follow connections rather than only search for matching words.\n\nGoogle popularised the term in 2012 when it launched its Knowledge Graph, describing a shift from matching keywords to understanding \"things, not strings\". The panels of facts beside many search results draw on it.\n\nA knowledge graph might link a parking permit service to the council that provides it, the guidance that explains it and the policy that governs it. Relational [databases](https://contentlexicon.com/database) can represent such relationships too. A graph is a modelling choice, and its value depends on which questions the connections help answer.\n\n## Why it matters for content\n\n[Content teams](https://contentlexicon.com/content-team) often meet knowledge graphs without the name. Structured data such as [schema](https://contentlexicon.com/schema).org markup describes entities and relationships on a web page, such as an organisation and its address or an event and its date, and search engines add those facts to their own graphs. A [content model](https://contentlexicon.com/content-model) with references between types, linking an article to its author and topic, is a graph in miniature.\n\nA [taxonomy](https://contentlexicon.com/taxonomy) is a natural starting point. It already gives concepts agreed labels and a hierarchy. Adding identifiers and other relationship types, such as \"related to\" or \"is a step in\", moves it towards a graph.\n\nContent estates contain relationships that pages often hide. A knowledge graph can connect terms, owners, evidence, audience segments and related guidance across channels. That supports discovery, reuse and impact analysis when something changes.\n\nGraphs need consistent identifiers, schemas and governance. If the same organisation appears under three names or relationships are vague, the structure becomes impressively connected confusion. Not every content problem needs a graph. They help most where relationships are numerous, important and hard to represent in a simple hierarchy.\n\n## How AI changes the picture\n\nAI systems can use knowledge graphs to retrieve context, constrain answers and follow explicit relationships. An assistant could show which recorded policy a guidance item depends on. That makes a connection traceable when the system exposes it, but doesn't establish that either item is current.\n\nAI can also propose entities and links from documents, speeding up graph creation. Human review remains important because inferred relationships may be ambiguous or wrong.\n\nGraphs and language models complement one another. The graph provides recorded claims and connections, while the model handles flexible language. Those claims still need reliable sources, and neither replaces content ownership.\n\n## Questions people ask\n\n**Is a knowledge graph the same as a database?**\n\nIt can be stored in a database or assembled from several sources. The graph model emphasises entities and named relationships. Relational databases can represent those relationships too, using a different model.\n\n**Is a taxonomy a knowledge graph?**\n\nNot usually. A taxonomy is a hierarchy or classification. It can form part of a richer graph containing many relationship types.\n\n**Do we need one for AI?**\n\nNot always. Start with the information problem. A graph helps when explicit relationships improve retrieval, reasoning or maintenance.\n\n**Who maintains it?**\n\nContent, data and subject experts, with shared governance for entities, relationships and sources.\n\n## Learn more\n\n- [Introducing the Knowledge Graph: things, not strings](https://blog.google/products-and-platforms/products/search/introducing-knowledge-graph-things-not) – Google's 2012 announcement, explaining the move from matching keywords to modelling real-world entities and their relationships.\n- [Knowledge graphs](https://www.turing.ac.uk/research/interest-groups/knowledge-graphs) – the Alan Turing Institute on applications across knowledge management, search, recommendations and AI.\n- [Ontologies, semantics and knowledge graphs](https://neo4j.com/blog/knowledge-graph/ontologies-in-neo4j-semantics-and-knowledge-graphs/) – a practical tour through the meaning layer that turns connected data into more than a network diagram." }, { "term": "KPIs", "alsoCalled": [], "slug": "kpis", "definition": "Key performance indicators: the measures used to judge whether content is achieving what it needs to achieve.", "category": "Research and evidence", "related": [ "content-health-check", "content-operating-model", "content-strategy" ], "updated": "2026-10-14T21:06:00.000Z", "url": "https://contentlexicon.com/kpis", "markdown": "https://contentlexicon.com/kpis.md", "credit": "KPIs, from the Content Lexicon by Contentious (https://contentlexicon.com/kpis), CC BY 4.0", "body": "For content, useful KPIs depend on purpose. A service page, campaign, [knowledge base](https://contentlexicon.com/knowledge-base) and [email](https://contentlexicon.com/email) sequence should not all be judged by the same numbers.\n\nA metric becomes a KPI because it indicates something central to success, not because it happens to be available in an analytics dashboard. Page views are easy to count. Whether someone understood an eligibility rule and completed the right task may be harder, but considerably more interesting.\n\n## Why it matters for content\n\nGood KPIs connect content activity with user and organisational outcomes. They help teams decide what to improve, stop or investigate. A service might track successful completion and avoidable support contacts. A knowledge base might examine successful searches, answer usefulness and repeated gaps. A newsletter might care about sustained readership and meaningful action rather than raw list size.\n\nNo single measure tells the whole story. Quantitative data shows patterns while research helps explain them. A high exit rate might indicate failure, or it might mean the page answered the question perfectly. Teams need a small set of complementary indicators and enough context to interpret them.\n\nPoor KPIs distort behaviour. If success is measured by volume, people produce volume. If clicks dominate, headlines become bait. Measures should reward the outcome the content exists to support, with safeguards for quality, [accessibility](https://contentlexicon.com/accessibility) and unintended harm.\n\n## How AI changes the picture\n\nAI can help analyse large sets of comments, support queries and search logs, revealing themes that traditional dashboards miss. It can connect qualitative signals with quantitative patterns and make performance evidence easier for teams to explore.\n\nIt can also generate convincing but unreliable interpretations. Models may infer causes from correlation, flatten minority experiences or summarise feedback without preserving important context. Analytical outputs need traceable data, explicit definitions and human scrutiny.\n\nAI content creates additional questions for measurement. Speed and volume aren't enough. Teams may need to track factual errors, unsupported claims, escalation rates, correction effort, source coverage and whether people can challenge an answer. The measure should reflect the quality of the experience, not merely the efficiency of the machine.\n\n## Questions people ask\n\n**What is the difference between a metric and a KPI?**\n\nA metric is any measure. A KPI is one selected as particularly important for judging progress towards an agreed goal. Useful dashboards contain both, but not every number deserves capital letters.\n\n**How many KPIs should we have?**\n\nUsually a small enough set to guide decisions. A service team might start with task completion, avoidable support contacts and serious content errors, then keep diagnostic measures beneath them. That's an illustrative set, not a universal quota. Add diagnostic measures underneath rather than promoting every available metric.\n\n**Are page views a useful KPI?**\n\nSometimes, when reach is the goal. Often they need context from task completion, quality, engagement or user research to mean anything useful.\n\n**How often should KPIs change?**\n\nKeep definitions stable enough to see trends, but review them when goals, services or user behaviour change. A measure that no longer supports a decision has become decoration.\n\n## Learn more\n\n- [Choosing content marketing goals and KPIs](https://nytlicensing.com/latest/methods/choosing-content-marketing-goals-and-kpis/) – NYT Licensing’s commercial content-marketing perspective linking goals with measures; not a complete framework for service or editorial performance.\n- [Vanity Metrics vs. Actionable Metrics – Guest Post by Eric Ries](https://tim.blog/2009/05/19/vanity-metrics-vs-actionable-metrics) – Eric Ries’s 2009 distinction between numbers that make you feel good and numbers that show cause and effect and help decide what to do next.\n- [Why content quality matters more than traffic](https://contenthealthcheck.com/learn/quality-not-traffic) – a challenge to proxy metrics, with a practical case for explicit quality criteria." }, { "term": "Linting", "alsoCalled": [ "Linter", "Prose linting" ], "slug": "linting", "definition": "Checking text automatically against a set of written rules and flagging every place that breaks one.", "category": "AI and tooling", "related": [ "ai-slop", "automated-testing", "continuous-integration", "copy-editing", "style-guide", "word-list" ], "updated": "2026-10-12T13:44:00.000Z", "url": "https://contentlexicon.com/linting", "markdown": "https://contentlexicon.com/linting.md", "credit": "Linting, from the Content Lexicon by Contentious (https://contentlexicon.com/linting), CC BY 4.0", "body": "Linting is a long-established habit in software development. A linter reads code without running it and flags anything that breaks a set of rules, from a variable that's never used to a construction the compiler accepts but that's likely to cause trouble. Stephen Johnson wrote the first, Lint, at Bell Labs in 1978 and named it after the fluff a dryer filter collects. Linters now run routinely in programmers' editors and again before code is merged, so a team's conventions get checked every time without anyone having to remember them.\n\nThe idea transfers neatly to writing, and documentation teams, who often keep their work alongside code already, have been quickest to adopt it. Prose linters such as Vale and proselint compare a draft with a set of rules and report what they find: a banned word, an em dash where house style wants an en dash, a heading in title case, a sentence over forty words. Rules range from simple patterns (this spelling, never that one) to counts and measures such as sentence length or reading level. A linter flags rather than fixes. Some propose a correction, but the decision stays with whoever is writing.\n\nSpellcheckers and [grammar](https://contentlexicon.com/grammar) checkers have done something similar for decades, but they apply the general rules of a language. Linting comes into its own with the rules particular to an organisation: its house terms, its [punctuation](https://contentlexicon.com/punctuation) choices, the phrases its [style guide](https://contentlexicon.com/style-guide) has ruled out. It also differs from [automated testing](https://contentlexicon.com/automated-testing), which checks whether things work (links resolve, required fields are filled, pages build), and from a copy edit, where a person reads for meaning as well as consistency. A linter checks text against rules, and only the rules someone has written down.\n\n## Why it matters for content\n\nWriting a rule down tightly enough for a machine to apply shows what a style guide means. “Avoid [jargon](https://contentlexicon.com/jargon)” can't be linted; a list of forty terms with their plain alternatives can. The exercise sorts guidance into three kinds: rules a pattern can check reliably, such as spellings, punctuation and house terms; rules where a pattern can only nominate candidates, such as whether a year belongs in a citation or whether “they” is ambiguous; and rules that stay with judgement, such as whether a paragraph earns its place. Sorting them often exposes guidance that different people had been reading in different ways.\n\nLinting also moves a rule to the moment of writing. A flag in the editor costs a few seconds while the draft is open; the same point raised in review costs a round of comments, and an editor correcting “e-mail” for the hundredth time is attention spent on nothing. Severity needs as much thought as coverage, because a linter that blocks publishing on every warning trains people to switch it off. GitLab's documentation team fails a build only on errors and leaves warnings and suggestions as information for the [writer](https://contentlexicon.com/writer). And a linter enforces a bad rule as faithfully as a good one. If the guide is wrong, the linter carries the mistake into every page it checks.\n\n## How AI changes the picture\n\nLanguage models can apply rules that patterns can't. Asked whether “they” is ambiguous in a sentence, whether a year in a citation changes how a reader would treat the source, or whether a paragraph has slipped into a sales register, a model can give a usable answer across thousands of pages. That brings judgement-shaped guidance within reach of automated checking, provided the rule comes with a test the model can apply and examples of passing and failing cases. Without them, a model applies its own reading of the rule, which may not be the editor's.\n\nGenerated text tends to fail in patterned ways (favourite words, stock sentence shapes, a reflexive rhythm of contrasts), which makes linting a natural first check on AI-assisted drafts. Open rule sets for spotting those patterns already exist, and a draft can be run against one before a person reads it. A clean pass is a minimum, though. A draft can avoid every listed pattern and still say nothing.\n\nModel-based checks bring a failure mode that patterns don't have. A pattern rule does what it says and nothing more; a model interprets the rule, and can misread it, deleting a whole sentence when asked to trim one detail or reporting a check it didn't run. Model checks are safest when they produce visible flags and proposed changes instead of silent edits, with someone sampling what they pass as well as what they flag.\n\n## Questions people ask\n\n**Is a linter the same as a style guide?**\n\nA linter holds the part of a style guide that can be written as automated checks. The guide still carries the reasoning, examples and judgement calls a linter can't, and people need it to understand why a rule exists. Keeping the two in step is ongoing work: when the guide changes, the rules have to change with it, or the linter goes on enforcing last year's decisions.\n\n**Should linting block publishing?**\n\nFor rules where a breach is unambiguous and costly, yes: a trademark, a legally required phrase, markup that breaks the page. Most style rules work better as warnings a writer can see and overrule with a reason. If people override a rule regularly, it probably needs rewriting or downgrading.\n\n**What makes a good first rule?**\n\nA correction editors make again and again. If the same fix turns up in review after review, it's a candidate. A dozen reliable rules that people trust do more than two hundred that bury every draft in flags.\n\n**Can a linter check voice?**\n\nPartly. Patterns can catch the words and constructions a [voice](https://contentlexicon.com/voice) avoids, and a model can judge whether a passage drifts from examples of the voice done well. Neither can tell whether the writing sounds like someone with something to say, which is still a job for a reader.\n\n## Learn more\n\n- [Vale](https://vale.sh/) – an open-source prose linter in which each team writes its own rules, from banned terms to sentence-length limits, and runs them in an editor or a publishing pipeline.\n- [Vale documentation tests](https://docs.gitlab.com/development/documentation/testing/vale/) – GitLab's working setup, with three levels of result of which only errors fail a build, and its own guidance on when a rule is too subjective or too noisy to enforce.\n- [proselint](https://github.com/amperser/proselint) – a contrasting approach: in place of house rules, a ready-made set of checks drawn from usage guides and editors including Bryan Garner, George Orwell and Mary Norris." }, { "term": "LLM", "alsoCalled": [ "Large language model" ], "slug": "llm", "definition": "Large language model: an AI system trained to predict and generate language, used for tasks such as drafting, summarising and answering questions.", "category": "AI and tooling", "related": [ "agentic-ai", "context-engineering", "context-window", "hallucination", "prompt", "rag", "system-one-models" ], "updated": "2026-10-10T17:04:00.000Z", "url": "https://contentlexicon.com/llm", "markdown": "https://contentlexicon.com/llm.md", "credit": "LLM, from the Content Lexicon by Contentious (https://contentlexicon.com/llm), CC BY 4.0", "body": "LLM stands for large language model. The same model can help someone compare several reports, ask a follow-up question or try another way of explaining an idea, through a language interface rather than a separate tool for each task.\n\nDespite the conversational surface, an LLM isn't a person, [database](https://contentlexicon.com/database) or source of truth. It generates an answer from patterns in its training and the context it has been given. That answer can be useful, subtle and apparently certain while still being wrong. Fluency is one of the technology’s strengths and one of its more mischievous disguises.\n\n## Why it matters for content\n\nLLMs are unusually relevant to [content teams](https://contentlexicon.com/content-team) because language is both the interface and much of the material being processed. They can help explore alternatives, reduce repetitive work, summarise large bodies of text and make specialist knowledge easier to query. Used carefully, they can give people more time for research, judgement and [editing](https://contentlexicon.com/editing).\n\nThey also change the conditions in which content is made. It becomes cheap to produce plausible prose at volume, which increases the need for strong sources, editorial standards and deliberate restraint. More words don't automatically create more value. An organisation that can't govern ten hand-written pages won't be rescued by generating ten thousand.\n\nContent people also have expertise that AI work needs: audience understanding, information structure, [voice](https://contentlexicon.com/voice), [taxonomy](https://contentlexicon.com/taxonomy), governance and the ability to distinguish a well-shaped sentence from a useful answer. Adopting an LLM involves technical, organisational and content decisions. Content specialists help decide what information the system should use, whom it serves and how its answers should be judged.\n\n## How LLMs change the picture\n\nA reader can ask for a different level of detail, compare passages or explore a document in another language without the publisher preparing every possible version. For example, a staff member reading technical guidance could ask for an explanation in their working language, then follow up on a term they don't understand. The answer still needs checking against the source, especially where an instruction has safety or legal consequences.\n\nThis changes the relationship between a publication and its reader: the reader can take part in shaping the explanation rather than only choosing a page. Teams need to decide which kinds of adaptation they can support, what evidence the system should show and where a fixed, approved answer is necessary.\n\nLLMs also make provenance more important. Teams need to know what source supports a claim, what information was available to the model and what should happen when sources disagree. Human review remains essential for consequential content, but review should be supported by evidence rather than asking someone to judge polished prose from instinct alone.\n\n## Questions people ask\n\n**Does an LLM understand what it writes?**\n\nNot in the human sense. It models patterns and relationships in data extremely well, but apparent understanding doesn't guarantee grounded knowledge, intention or responsibility.\n\n**Is an LLM the same as a chatbot?**\n\nNo. An LLM is the underlying model. A chatbot is a product or interface that may combine a model with instructions, tools, memory, retrieval and safety controls.\n\n**Why do LLMs make things up?**\n\nThey're built to generate likely continuations, not to retrieve truth automatically. If evidence is missing or the task invites speculation, a plausible invention can fit the pattern. Better sources, constraints and checking reduce the risk but don't abolish it.\n\n**Will LLMs replace content specialists?**\n\nThey will change the work and automate parts of it. They don't remove the need to decide what should exist, whom it serves, which sources are trustworthy and whether the result is responsible and useful.\n\n## Learn more\n\n- [The Illustrated Transformer](https://jalammar.github.io/illustrated-transformer/) – Jay Alammar's 2018 visual explanation of attention and the original encoder–decoder Transformer.\n- [Attention Is All You Need](https://arxiv.org/abs/1706.03762) – the 2017 paper that introduced the Transformer architecture, demonstrated through machine translation.\n- [A jargon-free explanation of how AI large language models work](https://arstechnica.com/science/2023/07/a-jargon-free-explanation-of-how-ai-large-language-models-work) – Timothy B. Lee and Sean Trott’s 2023 Ars Technica explainer on word vectors, transformers and next-word prediction, written for non-specialists." }, { "term": "Localisation", "alsoCalled": [], "slug": "localisation", "definition": "Adapting content and experience for another language, region or culture.", "category": "Operations", "related": [ "content-inventory", "content-maintenance" ], "updated": "2026-10-14T21:41:00.000Z", "url": "https://contentlexicon.com/localisation", "markdown": "https://contentlexicon.com/localisation.md", "credit": "Localisation, from the Content Lexicon by Contentious (https://contentlexicon.com/localisation), CC BY 4.0", "body": "Localisation is the work that makes a product or page feel as if it was made for the place it's used, whatever language or culture it started in. A faithful translation can still be a poor local experience: the words may be correct while the payment method is unavailable, the joke meaningless or the instruction legally wrong. Localisation asks whether the whole thing works here.\n\nThree related terms are easy to mix up. Translation changes the language. Localisation adapts everything else as well. Internationalisation comes first: designing content, templates and systems so they can be localised without being rebuilt, for example by allowing for longer words, right-to-left scripts and different date formats.\n\nWhat gets localised goes well beyond the text: date, time, number and address formats; currencies and prices; units of measurement; images, colours and icons that carry different meanings; examples, names and cultural references; legal requirements such as consumer rights, privacy notices and age limits; and [voice](https://contentlexicon.com/voice) and [tone](https://contentlexicon.com/tone), since formality and directness land differently from one culture to another.\n\nIn Wales, public bodies work under the Welsh Language Standards, which require them not to treat Welsh less favourably than English. That makes bilingual publishing a test of maintenance as much as translation. The Welsh version needs to be as current, complete and easy to find as the English one, rather than a translation that lags weeks behind, so teams that publish both at once tend to build translation into the content [workflow](https://contentlexicon.com/workflow) from the start.\n\n## Why it matters for content\n\nLocalisation affects inclusion, trust and task completion. People shouldn't receive a second-rate version because they use another language or live in another region. Planning for localisation early lets [content models](https://contentlexicon.com/content-model), layouts and workflows accommodate variation rather than treating it as a late export job.\n\nMaintenance is critical. Source content changes, but translated versions are often updated later or not at all. Teams need relationships between variants, visible status and clear responsibility for review. A changed delivery restriction, for example, needs to reach the source page and every affected language version.\n\nThe working tools make consistency possible at scale. A [glossary](https://contentlexicon.com/glossary) or terminology [database](https://contentlexicon.com/database) records approved translations of key terms. Translation memory stores previously translated sentences so they can be reused, which keeps wording consistent and cuts the cost of updates. A localisation [style guide](https://contentlexicon.com/style-guide) covers tone, formality and formatting for each language. Translators and local reviewers also need the purpose, audience and product behaviour, not isolated strings.\n\n## How AI changes the picture\n\nAI can speed up translation, terminology checks and first-pass adaptation. It can make support for more languages affordable, particularly when paired with approved glossaries and translation memory.\n\nFluency doesn't guarantee accuracy or cultural fit. Models may alter names, legal meaning or tone, and performance varies between languages, often falling for languages with less training data, Welsh among them. High-risk content needs qualified human review and testing with people who use the local version.\n\nStructured source content helps machines and people track variants. Clear fields, locale [metadata](https://contentlexicon.com/metadata) and shared terminology reduce drift, while change notifications can alert reviewers to affected variants. Someone still needs to act on them.\n\n## Questions people ask\n\n**Does every language version need the same content?**\n\nNot always. The core facts and rights should match, but examples, emphasis and even page structure can differ where local needs differ. No version should be missing something its readers need.\n\n**Should we write the source content differently?**\n\nYes. Clear, consistent source content with fewer idioms and explicit context is easier to localise without losing meaning.\n\n**Can AI replace translators?**\n\nIt can support routine work, but human expertise is still essential for nuance, consequence and quality, particularly in legal, medical or public-service content.\n\n**When should localisation begin?**\n\nDuring research and design, before layouts, fields and workflows make language variation expensive.\n\n## Learn more\n\n- [Localization vs. Internationalization](https://www.w3.org/International/questions/qa-i18n) – W3C's explanation of how the two differ, with a list of what localisation can involve beyond translation, from date formats and currency to legal requirements.\n- [Don’t alienate your user](https://prototypr.io/post/dont-alienate-your-user-a-primer-for-internationalisation-localisation) – a visual, practical primer on designing beyond one language and cultural context.\n- [Bilingual technology toolkit: good user experience](https://www.welshlanguagecommissioner.wales/media/vokjtkka/20220622-dg-s-pecyn-cymorth-prosiectau-tgch.pdf) – the Welsh Language Commissioner's case study of the Welsh Government's checklist for bilingual digital services, written for people commissioning technology rather than building it." }, { "term": "Longform", "alsoCalled": [ "Long-form" ], "slug": "longform", "definition": "Content that takes as long as it needs to – essays, features, reports, deep dives – as opposed to the soundbite and the listicle.", "category": "Editorial craft", "related": [ "content-type", "readability", "storytelling" ], "updated": "2026-10-06T15:16:00.000Z", "url": "https://contentlexicon.com/longform", "markdown": "https://contentlexicon.com/longform.md", "credit": "Longform, from the Content Lexicon by Contentious (https://contentlexicon.com/longform), CC BY 4.0", "body": "Longform is not defined only by word count, though word count tends to be involved. It is defined by depth, structure and sustained attention.\n\nThe persistent myth that nobody reads online is, technically speaking, cobblers. People read very long things when those things are valuable, readable and repay the time. What they don't read is padding, waffle or a page that mistakes length for substance.\n\n## Why it matters for content\n\nLongform matters because some ideas need room. Complex services, research findings, policy issues, product thinking, case studies and cultural arguments often need context, narrative, evidence and explanation. Short content can orientate, answer or [prompt](https://contentlexicon.com/prompt). Longform can build understanding.\n\nFor [content teams](https://contentlexicon.com/content-team), longform creates both opportunity and responsibility. It can show expertise, earn trust, support search, help sales conversations, deepen a brand’s intellectual position and give readers something useful. But it also asks more of structure, [editing](https://contentlexicon.com/editing), [accessibility](https://contentlexicon.com/accessibility) and design. A long article without signposting isn't a deep read; it's a corridor with no lights.\n\nDigital longform can combine prose, photography, [audio](https://contentlexicon.com/audio) and visual evidence, giving readers different routes through a substantial subject. Someone exploring an environmental report might begin with a local story, examine a map or follow the methods behind a finding. Those elements need a purposeful relationship: a recording might convey a witness's experience, while a chart lets readers inspect the evidence. They're part of the explanation, not decoration for a long article.\n\n## How AI changes the picture\n\nAI changes both the production and consumption of longform. It can help organise notes, summarise source material, test structure, generate outlines and find repetition, which helps most in the messy middle where ideas need arranging. But it can also produce long, fluent, strangely weightless prose that fills space without earning it.\n\nDiscovery and reuse change too. A strong longform piece may be summarised by AI tools, cited in answers or used as source material for internal knowledge systems. A companion experience could also let a reader ask about an unfamiliar reference, compare claims within the piece or follow a source without losing their place in the main argument. That would support reading rather than replace it; it isn't something every publication already provides. Structure and source links help keep the exchange connected to what the piece says. The piece still needs to reward the reader who stays; otherwise it's only raw material wearing literary trousers.\n\n## Questions people ask\n\n**How long is longform?**\n\nThere's no fixed threshold. Some people start using the term at around 1,000 words, others much later. What counts is whether the subject has been given enough space to be explained properly, without padding or compression that damages understanding.\n\n**Does longform work on the web?**\n\nYes, when it's well structured and has something to say. People read novels, investigations, essays and extensive guides on screens. They also skim, search and return later. Digital longform needs clear headings, summaries, navigation and design that respects different reading behaviours.\n\n**Is longform bad for accessibility?**\n\nNot inherently. Poorly structured longform can be hard to use, but well-structured longform can be accessible and helpful. Headings, summaries, descriptive links, readable [typography](https://contentlexicon.com/typography), plain structure and sensible page design all matter.\n\n**Can AI write longform?**\n\nAI can help draft or assemble longform, but good longform still needs a human point of view, editorial judgement, accurate sources and rhythm. A model will produce 3,000 words on request; deciding which 3,000 words the subject needs is still an editor's job.\n\n## Learn more\n\n- [Snow Fall: The Avalanche at Tunnel Creek](https://www.nytimes.com/projects/2012/snow-fall/) – The New York Times's influential 2012 example of digital longform storytelling.\n- [Scrolling and attention](https://www.nngroup.com/articles/scrolling-and-attention/) – Nielsen Norman Group eyetracking on how attention drops down a long page, and why the top matters.\n- [Longing for longer](https://contentious.ltd/thoughts/longing-for-longer) – on turning worthy reports that nobody reads into digital stories people might stay with." }, { "term": "Lorem ipsum", "alsoCalled": [], "slug": "lorem-ipsum", "definition": "Scrambled Latin used as placeholder text, so that a layout can be judged before the real words exist.", "category": "Editorial craft", "related": [ "agile", "content-design", "design-system", "pair-writing" ], "updated": "2026-10-11T06:25:00.000Z", "url": "https://contentlexicon.com/lorem-ipsum", "markdown": "https://contentlexicon.com/lorem-ipsum.md", "credit": "Lorem ipsum, from the Content Lexicon by Contentious (https://contentlexicon.com/lorem-ipsum), CC BY 4.0", "body": "Lorem ipsum is dummy text that fills a layout with something that looks like writing but can't be read. Its even mix of short and long words lets designers judge typefaces, spacing and the overall shape of a page without anyone being distracted by meaning. It turns up in mock-ups, templates and slide decks and, with depressing regularity, on live [websites](https://contentlexicon.com/website) where nobody got round to replacing it.\n\nThe words are a mangled passage from Cicero's _De finibus bonorum et malorum_ (45 BC), a treatise on ethics. \"Lorem ipsum\" is what remains of \"dolorem ipsum\", pain itself, from a sentence observing that nobody pursues pain for its own sake. [Typography](https://contentlexicon.com/typography) manuals had used a different passage of Cicero as specimen text since the 18th century, but the version in use today comes from Letraset's dry-transfer sheets of the 1960s, reportedly supplied by the printing historian James Mosley. Aldus brought it into PageMaker in the 1980s, typed in by hand from one of those sheets, and from there it spread into word processors and [content management systems](https://contentlexicon.com/cms). Even its origin story has been filled in without much care. The claim, repeated across countless websites, that printers have used it since the 1500s began as a guess in a 1994 letter, which its author, the Latin scholar Richard McClintock, has since admitted had no evidence behind it.\n\n## Why it matters for content\n\nReal content shapes design. A heading establishes hierarchy, a button enables an action and an error message helps someone recover. Placeholder text reduces all of these to rectangles, encouraging teams to treat content as material poured into a finished container. A design filled with meaningless text can't show whether its wording helps anyone complete a task. It assumes tidy headings, convenient lengths and content that never needs to explain an error, meet a legal requirement or accommodate a long Welsh place name.\n\nLorem ipsum also stands for a sequence: design first, words later. Placeholder text does little harm in a process where content and design develop together, and a great deal in one where the words arrive last and must squeeze into whatever space is left. Using representative content early exposes the difficult cases (long names, missing data, translations, warnings, competing priorities) and lets stakeholders judge whether an experience works rather than approving an attractive arrangement of grey bars. When final copy doesn't exist yet, provisional content clearly labelled as draft is enough to test the proposition and the structure. And lorem ipsum that survives onto a live page is rarely a typo: it's evidence that nobody was responsible for the words, and readers draw that conclusion quickly.\n\n## How AI changes the picture\n\nAI can produce realistic placeholder content in seconds, which is more useful than lorem ipsum when the examples are grounded in the product. It's particularly good at stress-testing a layout: the longest plausible product name, a heading translated into German, an error message for the worst case rather than the convenient one. Used that way, generated text challenges a design instead of flattering it.\n\nIt also removes lorem ipsum's one safety feature. Nobody mistakes scrambled Latin for finished copy, whereas fluent generated text looks ready to publish. It can be approved in a review as though someone had written it deliberately, set expectations that the real content can't meet or slip into production without fact-checking or ownership. Draft status and source need to stay visible, whether through labelling, highlighting or a rule that generated placeholders never reach a live environment.\n\nThe comparison runs the other way too. Much of what's now called [AI slop](https://contentlexicon.com/ai-slop) does the same job as lorem ipsum, filling space rather than saying anything, except that it can be read. The questions [content designers](https://contentlexicon.com/content-designer) have long asked of placeholder text apply to it directly: what is this for, who is it for and who decided it should be here?\n\n## Questions people ask\n\n**Is lorem ipsum ever useful?**\n\nYes, for the job it was made for. Comparing typefaces, spacing or the texture of a page works better when meaning isn't competing for attention. It's poor material for designing or testing a content experience.\n\n**Does it matter that the text is Latin?**\n\nIt can. Lorem ipsum approximates the rhythm of English and other Western European languages reasonably well, but says nothing about German compound words, right-to-left scripts or languages that need more space after translation. For a multilingual service, placeholder text in the wrong language hides the constraints that matter most.\n\n**What if the final content doesn't exist yet?**\n\nWrite realistic provisional copy based on known needs and mark it as draft. Include the awkward cases rather than only the ideal one.\n\n**Should content be finished before design begins?**\n\nNo. Content and design should develop together, each revealing constraints and possibilities in the other.\n\n## Learn more\n\n- [Lorem Ipsum is Killing Your Designs](https://www.smashingmagazine.com/2010/01/lorem-ipsum-killing-designs) – Kyle Fiedler's case for meaningful content during design, arguing that dummy copy can conceal real constraints.\n- [In Defense of Lorem Ipsum](https://karenmcgrane.com/in-defense-of-lorem-ipsum/) – Karen McGrane argues that separating content and design is the underlying problem, and intentional placeholder use can help some design reviews; not an endorsement for user testing.\n- [The Strange History of Lorem Ipsum](https://www.openculture.com/2026/06/the-strange-history-of-lorem-ipsum.html) – Colin Marshall's piece on Emily Zhang's video investigation, from the Letraset meeting where a single dummy text was first proposed to the Aldus creative director who typed it into PageMaker by hand." }, { "term": "Machine-readable", "alsoCalled": [], "slug": "machine-readable", "definition": "Represented in a form software can parse and process.", "category": "AI and tooling", "related": [ "api", "copyright-and-licensing", "markdown", "schema", "structured-content", "style-guide" ], "updated": "2026-10-14T21:11:00.000Z", "url": "https://contentlexicon.com/machine-readable", "markdown": "https://contentlexicon.com/machine-readable.md", "credit": "Machine-readable, from the Content Lexicon by Contentious (https://contentlexicon.com/machine-readable), CC BY 4.0", "body": "Machine-readable content is stored in a form that software can reliably find, extract and process without a person interpreting it first. The structure is explicit: a date is marked as a date, a price as a price and a heading as a heading.\n\nIt's a range rather than a yes-or-no quality. A scanned [PDF](https://contentlexicon.com/pdf) is an image of text, so software sees pixels until character recognition guesses at the words. A text-based PDF exposes the words but little of their structure. Semantic HTML marks up headings, lists, tables and links. Structured fields in a [CMS](https://contentlexicon.com/cms) separate an event's title, date, location and booking link. Data formats such as JSON and CSV, and markup such as [schema](https://contentlexicon.com/schema).org, present information in a standard shape that any program can read.\n\nMachine-readable doesn't mean unfriendly to people. Human-readable describes content a person can read and understand, and the best content serves both. [Structured content](https://contentlexicon.com/structured-content) is the practice of breaking content into defined, reusable parts, which is one route to machine [readability](https://contentlexicon.com/readability). Open data is narrower: data published under an open licence, ideally in machine-readable, non-proprietary formats, so anyone can reuse it. Tim Berners-Lee's five-star scheme for open data runs from a document on the web under an open licence to data linked to other data.\n\n## Why it matters for content\n\nStructure supports reuse, validation, search and automation. Software can display the same information across channels, check required fields and filter or connect related items. That reduces copying and makes updates more dependable.\n\n[Metadata](https://contentlexicon.com/metadata), schemas and consistent terminology all contribute. An event listed in a PDF timetable can be read by a person but has to be retyped to appear in a calendar feed. The same event in structured fields can feed the [website](https://contentlexicon.com/website), an app and search results at once, and a change to the date updates all three.\n\nThe work requires editorial judgement. Teams must decide which distinctions matter and avoid reducing nuanced content to fields because a [database](https://contentlexicon.com/database) prefers them.\n\n## How AI changes the picture\n\nAI can interpret unstructured prose, but clear structure improves reliability. It can help a configured retrieval system find a date or recognise a labelled source. Authority, and the difference between an instruction and an example, also depend on the system's rules, not format alone.\n\n[Style guides](https://contentlexicon.com/style-guide), [content models](https://contentlexicon.com/content-model) and governance rules can also become machine-readable, so a connected assistant can consult maintained guidance during drafting. That needs an integration. A structured file nobody connects doesn't change behaviour.\n\nStructure doesn't guarantee truth. Machines can reuse an outdated fact perfectly. Ownership and maintenance remain essential.\n\n## Questions people ask\n\n**Is HTML machine-readable?**\n\nIt can be, especially when semantic elements and structured data express meaning rather than visual formatting alone.\n\n**Do we need to turn everything into data?**\n\nNo. Structure recurring information where it improves reuse, checking or discovery. Keep prose where meaning depends on argument and context.\n\n**Is machine-readable the same as accessible?**\n\nThey overlap but aren't identical. Semantic structure helps assistive technology, while accessible content also depends on language, interaction and presentation.\n\n**Can AI understand unstructured content anyway?**\n\nIt can work with prose, although results depend on the task and system. Explicit structure makes recurring fields easier to extract and test but it isn't always the best form for an argument or story.\n\n## Learn more\n\n- [Schema.org for developers](https://schema.org/docs/developers.html) – a closer look at the machine-readable definitions themselves, including JSON-LD.\n- [5-star Open Data](https://5stardata.info/en/) – examples, costs and benefits for each step of Tim Berners-Lee's scheme, from a document on the web to linked data in open formats.\n- [From guidance to infrastructure](https://voicetoneandstyle.com/learn/from-guidance-to-infrastructure) – why static PDFs are giving way to structured guidance that both human teams and AI tools can draw on." }, { "term": "Markdown", "alsoCalled": [], "slug": "markdown", "definition": "A lightweight way to write structured plain text using simple marks for headings, links, lists and emphasis.", "category": "AI and tooling", "related": [ "context-engineering", "git", "machine-readable", "structured-content" ], "updated": "2026-10-07T09:06:00.000Z", "url": "https://contentlexicon.com/markdown", "markdown": "https://contentlexicon.com/markdown.md", "credit": "Markdown, from the Content Lexicon by Contentious (https://contentlexicon.com/markdown), CC BY 4.0", "body": "A hash makes a heading, hyphens make bullets, square brackets and parentheses make a link. The text stays readable even before anything renders it prettily.\n\nThat combination of [readability](https://contentlexicon.com/readability) and structure is why Markdown has become common in documentation, notes, developer tools, publishing systems and AI [workflows](https://contentlexicon.com/workflow).\n\n## Why it matters for content\n\nMarkdown makes content more portable, more durable and easier to reuse. Because it's plain text, it isn't locked inside one tool's formatting model: a Markdown file written in 2004 still opens in any text editor, which is more than can be said for plenty of word-processor files from the same year. Because it has explicit structure, it's easier for [websites](https://contentlexicon.com/website), apps and other systems to transform or render consistently. For [content teams](https://contentlexicon.com/content-team), that can make guidance, drafts, documentation and [knowledge-base](https://contentlexicon.com/knowledge-base) material easier to maintain.\n\nIt earns its keep where content needs to move between tools, be versioned cleanly, be published in more than one place or be used by both people and systems. It's less about aesthetic control and more about separating words and structure from presentation.\n\n## How AI changes the picture\n\nMarkdown can give an assistant clear headings, lists and links without the visual formatting of a page. Converting a 40-page [PDF](https://contentlexicon.com/pdf) [style guide](https://contentlexicon.com/style-guide) into well-structured Markdown may make its rules easier to extract and retrieve. That depends on the quality of the conversion and the retrieval system: a change of format alone won’t make the assistant find the right section.\n\nThe older argument for plain text gets stronger, too. The more content is drafted, transformed and passed through systems, the more useful it becomes to have a clean canonical form that humans and machines can both handle.\n\n## Questions people ask\n\n**Is Markdown the same as HTML?**\n\nNo. HTML is a fuller markup language for describing web pages in detail. Markdown is a simpler writing format that can often be converted into HTML. Different tools support slightly different flavours, such as CommonMark and GitHub Flavored Markdown, so check before you rely on the fancier marks.\n\n**Do I need to be technical to use Markdown?**\n\nThe basics are small and easy to learn. Most people can pick up headings, lists and links in a few minutes.\n\n**When should we use Word or Google Docs instead?**\n\nUse them when their collaboration, commenting or formatting features suit the work. Markdown is especially useful when portability, version control or structured publishing matters more than rich visual formatting while drafting.\n\n**Is Markdown only for developers?**\n\nNot at all. It's widely used for documentation, notes, publishing workflows and knowledge management as well as for software work.\n\n## Learn more\n\n- [Getting started with Markdown](https://www.markdownguide.org/getting-started/) – an explanation of syntax, rendering and portability, followed by tool suggestions.\n- [Markdown syntax](https://www.markdownguide.org/basic-syntax) – a reference for headings, lists, links and emphasis, with compatibility notes for different processors.\n- [Markdown](https://daringfireball.net/projects/markdown/) – John Gruber's original 2004 page, with the thinking behind the format: plain text that should be publishable as it is." }, { "term": "MCP", "alsoCalled": [ "Model Context Protocol" ], "slug": "mcp", "definition": "An open protocol for connecting AI applications to sources of information and tools.", "category": "AI and tooling", "related": [ "agentic-ai", "api", "knowledge-base", "machine-readable" ], "updated": "2026-10-12T17:27:00.000Z", "url": "https://contentlexicon.com/mcp", "markdown": "https://contentlexicon.com/mcp.md", "credit": "MCP, from the Content Lexicon by Contentious (https://contentlexicon.com/mcp), CC BY 4.0", "body": "MCP is an open standard that defines how AI applications connect to outside systems such as document stores, [databases](https://contentlexicon.com/database) and business tools. Instead of building a bespoke integration for every pairing of AI tool and system, developers build to one shared specification, and anything that follows it can connect. Anthropic introduced it in November 2024, and it has since been adopted more widely, including by OpenAI and Google.\n\nThe parts have plain jobs. An MCP client is the part of an AI application, such as a chat assistant or coding tool, that makes the connection. An MCP server is a small program that sits in front of a system and describes what it offers. Servers can offer tools, which are actions the model can take, such as searching records or creating a draft. They can offer resources, which are information the model can read, such as files or a [style guide](https://contentlexicon.com/style-guide). And they can offer [prompts](https://contentlexicon.com/prompt), which are reusable templates for common tasks. The protocol standardises how all of these are described and used. It doesn't decide what a system contains or whether it's trustworthy.\n\n## Why it matters for content\n\nMCP makes live content systems available to AI assistants while they work. A style guide, terminology list, [knowledge base](https://contentlexicon.com/knowledge-base) or product catalogue can become a source the assistant consults during drafting, rather than a document pasted into a prompt and gradually forgotten. When the style guide changes, every connected assistant sees the new version.\n\nThat raises the value of well-structured, authoritative content. A team could let an assistant look up the current style rule, check a draft against it and prepare an amendment in the same workspace. The benefit comes from maintained sources and well-defined actions, not from the connection alone.\n\nWriting is part of the design. Each tool and resource has a name and description that the model reads to decide when to use it. Vague descriptions lead to wrong choices, so content people should help write and test them.\n\nPermissions need the same care. A connection exposes whatever its server provides, within the access the application and server enforce. A read-only connection to a published style guide carries little risk. One that can edit or publish needs tighter limits, logging and human confirmation for consequential actions. The protocol supports these controls but doesn't enforce any organisation's policy.\n\n## How MCP changes the picture\n\nMCP moves AI from isolated chat towards agents that work inside organisational systems, finding information and taking actions across several tools in one task. The more an agent can do, the more its sources and permissions matter.\n\nContent an agent reads can also carry instructions. A web page, [email](https://contentlexicon.com/email) or document containing hidden text such as \"ignore previous instructions\" can steer an agent that has access to other tools, a risk known as prompt injection. Teams need to test how tools are selected, how errors are handled, where answers came from and what happens when a source is hostile. The protocol supplies a route, but it doesn't make every use of that route safe.\n\n## Questions people ask\n\n**Is MCP a database?**\n\nNo. It's a protocol that lets AI applications connect to systems that may contain data or actions.\n\n**Does MCP give an AI access to everything?**\n\nNo. A connection exposes the capabilities its server provides, within the access the application and server enforce. Grant only what's needed, and don't treat the protocol name as assurance that every permission boundary is correct.\n\n**Is MCP only for developers?**\n\nBuilding a server is technical, but content people should shape source quality, tool descriptions, permissions and review.\n\n**Does MCP replace APIs?**\n\nNo. An MCP server often sits on top of an existing [API](https://contentlexicon.com/api) and presents its capabilities in a consistent form for AI applications.\n\n## Learn more\n\n- [What is MCP?](https://modelcontextprotocol.io/docs/2026-07-28/getting-started/intro) – the official introduction under the 2026-07-28 documentation version, explaining the connection model; version context matters for technical details.\n- [MCP security best practices](https://modelcontextprotocol.io/docs/2026-07-28/tutorials/security/security_best_practices) – the more consequential reading on consent, token handling, scope and attack surfaces.\n- [Introducing the Model Context Protocol](https://www.anthropic.com/news/model-context-protocol) – the original announcement and the problem the standard was created to solve." }, { "term": "Merge conflict", "alsoCalled": [], "slug": "merge-conflict", "definition": "A clash between changes that the version-control system cannot combine automatically.", "category": "AI and tooling", "related": [ "single-source-of-truth", "versioning" ], "updated": "2026-10-14T22:09:00.000Z", "url": "https://contentlexicon.com/merge-conflict", "markdown": "https://contentlexicon.com/merge-conflict.md", "credit": "Merge conflict, from the Content Lexicon by Contentious (https://contentlexicon.com/merge-conflict), CC BY 4.0", "body": "A merge conflict happens when two sets of changes to the same material can't be combined without a human decision. Version-control systems merge most concurrent changes automatically, but they stop when two people have changed the same lines differently, or when one person has edited material another has deleted.\n\n[Git](https://contentlexicon.com/git) shows a conflict by pausing the merge and marking the disputed passage inside the file. The lines between `<<<<<<<` and `=======` hold one version. The lines between `=======` and `>>>>>>>` hold the other. Someone with context resolves it by keeping one side, combining the two or writing something new, then deleting the markers and completing the merge. Platforms such as GitHub and most code editors present the same choice side by side.\n\n## Why it matters for content\n\nThe content version is familiar. Two editors open the same paragraph of an eligibility page in a system without page locking. One updates the income threshold. The other rewrites the paragraph for clarity, working from the old figure. Whoever saves last wins, and the threshold change disappears without anyone noticing. Version control would have stopped at that paragraph and asked someone to choose. Many content systems overwrite without warning. Page locking, which lets only one person edit at a time, prevents this at the cost of making others wait.\n\nSome conflicts hide where no tool will flag them. Two copies of a policy, one on the [website](https://contentlexicon.com/website) and one in a [PDF](https://contentlexicon.com/pdf), can drift apart without ever passing through the same merge. Martin Fowler calls the software equivalent a semantic conflict: changes that combine cleanly as text but contradict each other in meaning. These are harder to spot and depend on testing, review and a clear canonical source.\n\nPrevention starts with that canonical source, visible ownership and version history. Teams should know who's [editing](https://contentlexicon.com/editing) important material and how related versions, including translations and PDFs, are connected.\n\nConflict isn't bad in itself. Two changes may reflect legitimate needs. Making the disagreement visible lets someone with context resolve it deliberately.\n\n## How AI changes the picture\n\nAI systems combine information from many sources, creating an invisible form of merging. When sources disagree, a model may pick one or blend them rather than stop and report the conflict.\n\nConfigure retrieval to prefer the appropriate current source and make important disagreements visible. Dates and status help when used correctly, but a newer date doesn't settle policy authority by itself.\n\nAI can propose a resolution for a literal merge conflict and compare documents for contradictory claims. It can't decide which owner has authority unless the organisation has already defined that.\n\n## Questions people ask\n\n**Is duplication always a merge conflict?**\n\nNo. Duplication can lead to contradictions, but a literal merge conflict happens only when changes are combined and the system can't resolve them. Independent copies can disagree without ever producing that technical event.\n\n**Can a CMS prevent conflicts?**\n\nIt can lock pages, show who's editing, warn about simultaneous changes and keep history. It can't reconcile separate copies in other systems without governance.\n\n**Who decides the winning version?**\n\nThe relevant [content owner](https://contentlexicon.com/content-owner) or subject authority, informed by [user needs](https://contentlexicon.com/user-needs) and evidence.\n\n**Should old versions be deleted?**\n\nRetire them from active use while keeping history where records or accountability require it.\n\n## Learn more\n\n- [Merge conflicts](https://docs.github.com/en/pull-requests/reference/merge-conflicts) – GitHub's explanation of changes it cannot combine automatically, including concurrent edits and edit-versus-delete cases.\n- [Basic Branching and Merging](https://git-scm.com/book/en/v2/Git-Branching-Basic-Branching-and-Merging) – a chapter of the free Pro Git book showing what conflict markers look like in a file and how to resolve them.\n- [Semantic Conflict](https://martinfowler.com/bliki/SemanticConflict.html) – Martin Fowler's 2011 note on changes that merge cleanly as text but clash in meaning, and why integrating often helps." }, { "term": "Message architecture", "alsoCalled": [], "slug": "message-architecture", "definition": "A short, prioritised list of the qualities and ideas an organisation wants its content to convey.", "category": "Foundations", "related": [ "channel-strategy", "content-principles", "content-strategy", "core-content-strategy", "storytelling", "voice" ], "updated": "2026-10-11T06:25:00.000Z", "url": "https://contentlexicon.com/message-architecture", "markdown": "https://contentlexicon.com/message-architecture.md", "credit": "Message architecture, from the Content Lexicon by Contentious (https://contentlexicon.com/message-architecture), CC BY 4.0", "body": "A message architecture is a short, prioritised list of the qualities and ideas an organisation wants its content to convey. It's usually written as an outline: three to five main communication goals in order of importance, each with a few supporting terms that pin down what it means. A housing advice charity might lead with \"practical\" (clear next steps, no [jargon](https://contentlexicon.com/jargon)), followed by \"on your side\" and then \"expert\", so that when a page can't do everything, being useful wins over sounding authoritative.\n\nThe words aren't meant to appear in the copy. A message architecture isn't a set of slogans, a mission statement or a list of values; it guides decisions across topics and channels while leaving [writers](https://contentlexicon.com/writer) room to find the right language. The approach comes from the [content strategist](https://contentlexicon.com/content-strategist) Margot Bloomstein, who set it out in Content Strategy at Work (2012).\n\n## Why it matters for content\n\nOrganisations often want to say everything at once. A message architecture creates hierarchy, so when a page, campaign or interface can't carry every idea, teams can decide what leads and what stays in support. It also reaches beyond writers. Designers and information architects can use it to choose labels and patterns, and [SEO](https://contentlexicon.com/seo) specialists to keep keyword choices in character.\n\nIt connects strategy with [voice](https://contentlexicon.com/voice) and [tone](https://contentlexicon.com/tone). If content should communicate expertise and warmth, writers and reviewers can check whether the evidence, structure and language produce those impressions, which is more useful feedback than \"make it punchier\". It gives a [content audit](https://contentlexicon.com/content-audit) something to measure against beyond accuracy and freshness, and it helps with choosing formats: if one goal is being approachable, an annual report may not be the best way to show it. The priorities should come from research, organisational purpose and stakeholder discussion.\n\n## How AI changes the picture\n\nA message architecture makes good guidance for AI drafting, in [system prompts](https://contentlexicon.com/system-prompt), briefs and review checklists, because it names what content should convey without dictating the words. Models tend to drift towards a generic register of helpful, innovative and friendly, which is the opposite of a set of deliberate choices. Spelling out what the organisation isn't, along with examples and counterexamples, helps keep generated drafts away from that middle ground.\n\nModels may also reduce qualities to surface features. \"Warm\" becomes exclamation marks and \"expert\" becomes jargon, when the qualities live in what the content says and how it treats the reader. Review should look at substance, evidence and the relationship with the audience, not just word choice. AI can help at scale by tagging a body of content against the architecture's goals to show where a quality is missing or overplayed, but those are judgements to check against the content itself.\n\n## Questions people ask\n\n**Is message architecture the same as key messages?**\n\nKey messages are particular claims, such as a campaign's three main points. A message architecture sits above them and shapes many messages, so the key messages for any campaign should express its goals in that campaign's terms.\n\n**How is message architecture different from a tone of voice?**\n\nA message architecture says what impressions content should leave and in what order of importance. Tone of voice describes how the writing sounds while doing it. Voice guidelines often grow out of a message architecture, but they aren't the same thing.\n\n**How is a message architecture put together?**\n\nThe usual method is a card sort. Stakeholders work through around a hundred adjectives, sorting them into who the organisation is now, who it isn't and who it wants to be. The \"isn't\" pile is set aside, and the rest are clustered and ranked until three to five goals survive, defined in as much detail as the team using them needs. The sorting is as valuable as the result, because it makes people argue about priorities before those arguments reach individual drafts.\n\n**How do you know a message architecture is working?**\n\nTest content with the people it's for and ask what qualities or ideas they perceive, rather than whether colleagues approve of the words. Audits against the architecture over time show whether it's shaping new content or sitting in a drawer.\n\n## Learn more\n\n- [Content Strategy at Work: Margot Bloomstein on Deriving Design from Content at MOO](https://www.printmag.com/branding-identity-design/content-strategy-at-work-margot-bloomstein-on-deriving-design-from-content-at-moo) – an excerpt from Bloomstein’s book, published by PRINT, tracing MOO's message architecture from its order confirmation email through its copy, imagery and brand guidelines.\n- [Three Steps to Developing a Brand-driven Content Strategy](https://www.contentstrategy.at/lecture-reports/how-a-brand-driven-content-strategy-makes-brand-values-great-again-4) – a student report from FH Joanneum's content strategy master's, covering the card sort, a sample message architecture and how it feeds a content audit.\n- [Brand Message Architecture](https://www.oho.com/blog/brand-message-architecture) – the agency OHO on adapting the card sort for university websites, with examples of how messages shaped homepages and a clear account of what a message architecture isn't." }, { "term": "Metadata", "alsoCalled": [], "slug": "metadata", "definition": "Information about content that helps it be found, filtered, understood, reused and managed.", "category": "Structure and systems", "related": [ "alt-text", "backend", "content-inventory", "findability", "knowledge-base", "review-date", "schema", "taxonomy" ], "updated": "2026-10-11T06:14:00.000Z", "url": "https://contentlexicon.com/metadata", "markdown": "https://contentlexicon.com/metadata.md", "credit": "Metadata, from the Content Lexicon by Contentious (https://contentlexicon.com/metadata), CC BY 4.0", "body": "Metadata can include title, author, date, topic, audience, status, format, owner, tags and other descriptive fields that help a system or a person understand what something is. Metadata often sits behind the scenes, but it shapes how content is organised, found, filtered, reused and governed.\n\nIn practice, metadata is one of the things that turns a pile of files or pages into something workable. Without it, content is hard to sort, search or maintain.\n\n## Why it matters for content\n\nMetadata helps content be found, reused and kept under control. It lets a reader narrow 400 reports down to the three about housing, tells a team what it has and who owns each piece, feeds reporting and makes content easier to manage over time. It also connects closely to other structural work such as [taxonomy](https://contentlexicon.com/taxonomy), content modelling and [information architecture](https://contentlexicon.com/information-architecture).\n\nThe need grows with scale and with the number of people managing the content. A handful of pages can be held in memory; hundreds or thousands can't.\n\nA photograph shows the everyday version: its date, location, creator and rights information can help someone find it, understand its context and establish whether it can be reused. An assistant could help navigate those fields, but the description and permissions still need a reliable source.\n\n## How AI changes the picture\n\nOnce AI is reading your content, metadata becomes operational. Labels, dates, types, statuses and ownership fields affect what gets retrieved, summarised, recommended or transformed. If nobody fills in the “last reviewed” field, it becomes much harder to establish what's current, though dates and revision statements may exist elsewhere. And a [review date](https://contentlexicon.com/review-date) isn't proof that a claim still holds true.\n\nAI can help suggest metadata, but it also raises the stakes on getting it right. Once machines are acting on content rather than storing it, metadata stops being admin and starts shaping behaviour.\n\n## Questions people ask\n\n**How does metadata relate to taxonomy?**\n\nA taxonomy is a controlled vocabulary for classifying things. Metadata fields such as topic or audience often draw their values from it, so the two work closely together without being the same thing.\n\n**Is metadata only for machines?**\n\nNo. Machines use it heavily, but people benefit too whenever they search, filter, sort or try to understand a content estate.\n\n**Do small sites need metadata?**\n\nYes, though less of it. Even a small set of content benefits from basic fields such as owner, date, type or topic.\n\n**What happens if metadata is inconsistent?**\n\nSearch, reporting, reuse and governance all get weaker. Inconsistent metadata is one of the least visible ways to make a content system unreliable.\n\n## Learn more\n\n- [Taxonomy 101](https://www.nngroup.com/articles/taxonomy-101/) – useful context on controlled vocabularies and classification.\n- [DCMI Metadata Terms](https://www.dublincore.org/specifications/dublin-core/dcmi-terms/) – a shared vocabulary that shows why descriptions, rights and provenance need agreed meanings across systems; technical reference rather than introductory prose.\n- [Understanding Metadata: What is Metadata, and What is it For?](https://www.niso.org/publications/understanding-metadata-2017) – Jenn Riley's National Information Standards Organization primer on the types of metadata, the main standards and what metadata is used for. A 49-page PDF, this is from 2017 but is still useful." }, { "term": "Microcopy", "alsoCalled": [], "slug": "microcopy", "definition": "The short pieces of interface text, such as labels, hints and error messages, that help people complete tasks.", "category": "Editorial craft", "related": [ "call-to-action", "content", "content-design", "conversational-design", "frontend", "tone", "ux-writer" ], "updated": "2026-10-11T06:25:00.000Z", "url": "https://contentlexicon.com/microcopy", "markdown": "https://contentlexicon.com/microcopy.md", "credit": "Microcopy, from the Content Lexicon by Contentious (https://contentlexicon.com/microcopy), CC BY 4.0", "body": "Microcopy is the small text in an interface that helps people act: button labels, form hints, error messages, confirmations, empty states, tooltips and the line under a password field explaining what it needs. Each piece is short, but together these words carry much of what a product tells people about what to do next and what just happened. Their meaning depends on what the system does, so microcopy has to be designed alongside the flow rather than added at the end.\n\nJoshua Porter popularised the term in a 2009 blog post, “Writing Microcopy”. He described a checkout form whose transactions kept failing because people entered an address that didn't match their card; one added line, “Be sure to enter the billing address associated with your credit card”, made the errors go away. Kinneret Yifrah's Microcopy: The Complete Guide (2017) later gave the practice a book-length treatment.\n\n## Why it matters for content\n\nMicrocopy is where content and behaviour meet most directly. A clear label can reduce uncertainty; a specific error can help someone recover without contacting support; a confirmation can prevent a second payment attempt when it explains what happened and what comes next. A button that says “Apply” and one that says “Submit” may imply different next steps even when they look identical, so decide what the action does before choosing the words. Testing the effect of changes rather than assuming better wording guarantees completion is the safe approach.\n\nMicrocopy is also where [tone](https://contentlexicon.com/tone) is tested hardest. [Voice](https://contentlexicon.com/voice) is easy to maintain in an article and difficult in a failed payment message, where breeziness is insulting and coldness is worse. Microcopy shows whether a voice and tone guide works under pressure.\n\nFor money, eligibility or personal data, ambiguity can carry serious consequences. A cancellation message needs to distinguish ending a service now from stopping its next renewal. A consent checkbox needs to say what someone is agreeing to, not just that they agree.\n\n## How AI changes the picture\n\nAI can generate plausible microcopy instantly, which is both helpful and risky. Drafting twelve variants of an error message is tedious work a machine can do well. But microcopy depends on system context: what failed, what the person can do about it, what happens next. A model that doesn't know those things will write something confident, tidy and wrong.\n\nInterface language is also increasingly generated. As products add assistants and conversational flows, many of the words doing the work are responses composed on the spot rather than fixed labels. Documented voice, tone and style rules can guide that generation as well as human writing, if the system is configured to use them, though guidance doesn't enforce permissions or guarantee compliant answers.\n\nAI can support consistency checks across a product: finding fourteen different phrasings for the same action, spotting labels that contradict the help text, flagging [jargon](https://contentlexicon.com/jargon) in field hints. Review what it flags, since different wording sometimes reflects a different action.\n\n## Questions people ask\n\n**Is microcopy the same as UX writing?**\n\nMicrocopy is the text; [UX](https://contentlexicon.com/ux) writing is the discipline of producing it alongside design and research. [UX writers](https://contentlexicon.com/ux-writer) also work on flows, naming and content structure, so the job reaches beyond the words on any one screen.\n\n**Should designers or content people write it?**\n\nEither can, provided they're involved while the flow is being designed. Microcopy written after the screens are finished can only tidy up problems the words might have prevented.\n\n**Does it need a style guide entry?**\n\nYes. Button conventions, error message patterns and how to address the user are the kinds of decisions a [style guide](https://contentlexicon.com/style-guide) exists to settle once. Patterns with worked examples tend to help more than abstract rules.\n\n**Can microcopy be tested?**\n\nThrough task-based research or a controlled experiment. For a label change, check whether people understand the action as well as whether completion changes. Reliable comparisons may need enough participants, traffic and time, so they're not always cheap or quick.\n\n## Learn more\n\n- [Writing Microcopy](https://web.archive.org/web/2010/http://bokardo.com/archives/writing-microcopy/) – Joshua Porter's Bokardo post, via the Internet Archive, with the billing address story and other examples of small copy making a large difference.\n- [What is UX writing?](https://uxcontent.com/what-is-ux-writing) – microcopy placed in its proper context as interface design.\n- [5 Microcopy & UX Writing Best Practices We Can Learn from Lemonade](https://blog.prototypr.io/5-microcopy-ux-writing-best-practices-we-can-learn-from-lemonade-e26f1d2adea9) – a design account of Lemonade’s interface wording; concrete observations, not a controlled demonstration that those words caused performance gains.\n- [Error message guidelines](https://www.nngroup.com/articles/error-message-guidelines/) – evidence-led guidance on making failure states useful." }, { "term": "Microsite", "alsoCalled": [], "slug": "microsite", "definition": "A small, focused website or section created for a campaign, product, audience or specific purpose.", "category": "Structure and systems", "related": [ "channel-strategy", "cms", "content-strategy", "website" ], "updated": "2026-10-11T06:25:00.000Z", "url": "https://contentlexicon.com/microsite", "markdown": "https://contentlexicon.com/microsite.md", "credit": "Microsite, from the Content Lexicon by Contentious (https://contentlexicon.com/microsite), CC BY 4.0", "body": "Microsites usually sit beside the main [website](https://contentlexicon.com/website) rather than inside the normal navigation, though the boundary can be blurry. A microsite gives a piece of content, a proposition or a story its own space.\n\nThat can be useful; it can also be a way to create a new governance problem. The crucial question is whether it has a clear job, a clear audience and someone willing to look after it once the launch is over.\n\n## Why it matters for content\n\nMicrosites matter because they change the content context. A campaign landing page, product explainer, knowledge hub or public report may need a different rhythm, structure and visual treatment from a main organisational site. A microsite gives that content room. People get a focused route through it, without the main site's navigation, news and promotions competing for their attention.\n\nThe risk is fragmentation. Every microsite creates more content to maintain, more analytics to interpret, more governance to define and more places where old messages can lurk. If the microsite repeats information from the main site, the organisation also has to decide which version is canonical. Otherwise, someone will eventually update one and forget the other.\n\n## How AI changes the picture\n\nThe lifecycle question gets sharper once AI is involved. A microsite isn't only read by people clicking through from a campaign. It may also be read, summarised and cited by search engines, AI assistants and internal tools. If its content is out of date, disconnected from the main site or unclear about ownership, it can go on feeding last year's prices and a closed helpline number to those tools long after the main site has corrected them.\n\nWell-structured microsites gain value, too. A tightly focused site with clear headings, good [metadata](https://contentlexicon.com/metadata), stable URLs and explicit source information can become useful input for summaries, proposals, help content and internal knowledge systems. Clear structure and maintenance also help readers, while metadata and explicit source information help systems interpret the content.\n\n## Questions people ask\n\n**What's the difference between a microsite and a landing page?**\n\nMostly size. A landing page is usually a single page designed around one action or campaign. A microsite is usually a small collection of pages or a self-contained experience. In practice, people use the terms loosely.\n\n**When is a microsite a good idea?**\n\nWhen the content needs its own identity, structure or way in. That might be a major campaign, a public report, a product, a research project or a specialist audience. If the content could live happily inside the main site, put it there.\n\n**What's the main content risk?**\n\nThe main risk is orphaning. Microsites are often funded and loved at launch, then ignored. They can quickly become places where old claims, old branding, old contact details and old priorities remain publicly available long after the main site has moved on.\n\n**Who should own a microsite?**\n\nSomeone with responsibility beyond launch day. Ownership should cover content accuracy, analytics, [accessibility](https://contentlexicon.com/accessibility), technical maintenance, redirects and eventual retirement. Without that person, the site stays live and unchecked until someone outside the organisation spots a mistake on it.\n\n## Learn more\n\n- [The Rise of the Subsite](https://www.nngroup.com/articles/the-rise-of-the-subsite/) – Jakob Nielsen's 1996 argument for connected subsites, with a 2003 addition warning about microsite fragmentation.\n- [What is a microsite?](https://contently.com/2015/04/09/what-the-hell-is-a-microsite-and-why-do-i-need-one/) – a distinction between campaign sites and ongoing publications and the different resourcing they require." }, { "term": "Paddle, swim, dive", "alsoCalled": [], "slug": "paddle-swim-dive", "definition": "A way to organise one topic at different, useful depths, so people can get an overview, more detail or the full evidence without being sorted into fixed audience segments.", "category": "Structure and systems", "related": [ "audience-segmentation", "content-design", "content-model", "findability", "front-loading", "information-architecture", "pdf", "plain-language", "structured-content", "website" ], "updated": "2026-10-11T06:25:00.000Z", "url": "https://contentlexicon.com/paddle-swim-dive", "markdown": "https://contentlexicon.com/paddle-swim-dive.md", "credit": "Paddle, swim, dive, from the Content Lexicon by Contentious (https://contentlexicon.com/paddle-swim-dive), CC BY 4.0", "body": "A paddle experience gives a concise, complete overview or immediate action. Swim might add explanation and practical guidance. Dive provides the evidence, complexity and specialist detail for people who need it.\n\nThe framework is about moving between depths. It doesn't rank readers, and it doesn't insist on three levels. Each experience should be rewarding in its own right and offer a clear, inviting route into deeper material. The same person might paddle through a topic today and dive into it next month. Context determines the depth, not a fixed audience label.\n\n## Why it matters for content\n\nOrganisations often arrange content around audience segments: for parents, for professionals, for trustees. Real people rarely stay inside those boxes. Dividing a topic this way can produce several versions of the same facts, owned by different teams and drifting apart over time. Paddle, swim, dive keeps the topic together while giving people different ways into it.\n\nThe framework also protects substantial material from becoming either overwhelming or invisible. A research report buried in a [PDF](https://contentlexicon.com/pdf) may contain the best evidence an organisation has, but few people will find it from the surrounding pages. A complete paddle layer can orient the reader, a swim layer can explain the implications and a well-signposted dive can lead to the report itself. Depth is about the commitment asked of the reader, not word count or format: a two-minute video can be paddle, while a dense transcript can still be swim. Once you have the terms, the common failures are easy to name: the marketing site that's all paddle, the research institute that's all dive with no way in, and the dataset nobody can reach from the topic page it belongs to.\n\n## How AI changes the picture\n\nAn assistant could offer a short explanation, then follow a reader’s question into practical guidance or the underlying evidence. Someone exploring a research finding might first ask what it means, then ask how the study reached it. Paddle, swim and dive content can support that change of depth without the publisher preparing a separate answer to every possible question.\n\nThe layers should agree on their shared facts and make their relationships clear. Descriptive headings, links and [metadata](https://contentlexicon.com/metadata) help readers and retrieval systems recognise connected views of one topic. They don’t guarantee that an assistant selects the current or appropriate source. A generated summary still needs editorial judgement about what a paddle reader must understand, including any qualification that changes the meaning of the finding.\n\n## Questions people ask\n\n**Does everyone have to start by paddling?**\n\nSearch, links and AI answers often drop people straight into swim or dive material, so the structure needs several entry points, with the surrounding depths easy to find from each one.\n\n**How does it relate to progressive disclosure?**\n\nIt's a close relative. Progressive disclosure, from interface design, hides advanced or less-used options until someone asks for them, and the journalist's [inverted pyramid](https://contentlexicon.com/front-loading) puts the most important information first. Paddle, swim, dive is Contentious's framework for applying the same instinct to whole topics: each depth has to work as a complete experience, so paddle should satisfy someone who never goes further.\n\n**Does dive mean a long report or PDF?**\n\nFormat and length are weak measures of depth. Dive content asks for more knowledge, attention or time because the subject requires it. It might be a dataset, a technical standard, a worked case study or an interactive tool.\n\n**Should websites label content as paddle, swim and dive?**\n\nUsually not. They're planning terms for [content teams](https://contentlexicon.com/content-team). People should experience clear summaries, sensible detail and well-signposted evidence without having to learn the framework first.\n\n## Learn more\n\n- [Progressive Disclosure](https://www.nngroup.com/articles/progressive-disclosure) – Jakob Nielsen’s account of revealing secondary options when needed, including the risks of poorly signposted or overly deep interfaces.\n- [Writing and editing: Structuring content](https://service-manual.ons.gov.uk/content/writing-for-users/structuring-content) – Office for National Statistics guidance on frontloading, the inverted pyramid and linking people to supporting detail without repeating it.\n- [Taking a thematic approach](https://contentious.ltd/thoughts/taking-a-thematic-approach) – Contentious on organising related material around what people care about rather than internal structures." }, { "term": "Pair writing", "alsoCalled": [], "slug": "pair-writing", "definition": "Two people writing together, usually combining subject knowledge with content craft.", "category": "Editorial craft", "related": [ "content-design", "editing", "subject-matter-expert", "writer" ], "updated": "2026-10-15T10:03:00.000Z", "url": "https://contentlexicon.com/pair-writing", "markdown": "https://contentlexicon.com/pair-writing.md", "credit": "Pair writing, from the Content Lexicon by Contentious (https://contentlexicon.com/pair-writing), CC BY 4.0", "body": "Pair writing is a way of producing content in which two people work on one draft at the same time, talking through each decision as they go. The usual pairing puts someone who knows the subject alongside someone who knows how to write for the audience, but two [writers](https://contentlexicon.com/writer), a writer and a designer, or a policy lead and a lawyer can pair just as well. It can happen around one screen or in a shared document on a call.\n\nWhat sets it apart from ordinary collaboration is timing. Instead of a draft travelling back and forth through tracked changes and comment threads, questions about meaning, evidence and wording get settled while the sentence is being written. The idea borrows from pair programming, popularised through Extreme Programming in the late 1990s, where one person types and the other watches direction and quality. In [content teams](https://contentlexicon.com/content-team) a common arrangement is for the subject expert to hold the keyboard while the [content designer](https://contentlexicon.com/content-designer) suggests plainer wording, so the expert stays in charge of accuracy and sees why each change is made.\n\n## Why it matters for content\n\nReview cycles drag because each round surfaces something nobody raised earlier: a term the audience won't know, a caveat the expert considers essential, a step that only makes sense to insiders. Pairing brings those conversations forward. The content designer can ask what a rule means for someone applying for the first time and get an answer on the spot, and the expert can stop a simplification that has become inaccurate before it reaches [sign-off](https://contentlexicon.com/sign-off).\n\nIt also changes the working relationship. Experts who've paired tend to trust the final wording because they shaped it, which can shorten approval and cut down late requests to add a paragraph. The method suits complex, sensitive or contested topics best. It works less well when neither person has authority to decide or the facts are still unknown, because the session turns into a list of questions for someone else.\n\nSessions go better with a clear purpose, the source material to hand and a time limit, because more than an hour or so of concentrated drafting is tiring. Neither person should end up as a typist taking dictation.\n\n## How AI changes the picture\n\nA model can sit in as a second participant, and much of what makes pair writing work also makes writing with AI work. In both cases there's one shared draft and a running conversation about what each sentence is for, with decisions made as the words go down rather than at the end. The habits pair writers have learned carry across almost unchanged: agree the purpose before starting, keep the source material to hand, stop when concentration goes and keep asking what the reader needs.\n\nThe roles carry over too. A model often makes a good driver, producing wording quickly while the person navigates by setting the brief, judging accuracy and deciding what stays. It can also work the other way round, with the person drafting and the model reading over their shoulder, suggesting plainer phrasings, checking terms against an approved [glossary](https://contentlexicon.com/glossary) or flagging a sentence that's harder to read than the audience needs. As in any pair, the danger is one side sliding into taking dictation. Someone who accepts every suggestion has stopped pairing, and so has someone who uses the model only to tidy commas. When two people pair, a model can still sit in as a third participant, perhaps producing a draft for them to critique together, which often reveals what both had assumed but not said.\n\nA model can also act as fact-checker and verifier, which helps with one of the things that can stall a human pair: a fact nobody in the room knows. When asked, an AI agent can research a topic, check a draft's facts and figures, and find the references and verified data behind a claim, often while the session is still running. Its checking needs checking in turn. Follow its references back to the source rather than trusting the summary, and keep track of which claims came from an expert, which from sources the model found and which from the model alone, because a fluent suggestion accepted mid-conversation can be hard to trace later.\n\nPair writing also shows where the comparison stops. A model can't push back from first-hand experience of the subject, and it can't share responsibility for what's published. The person carries that part, which makes the conversational habits matter more: asking for alternatives rather than taking the first version, and telling the model what the audience does and doesn't know.\n\n## Questions people ask\n\n**Who should hold the keyboard?**\n\nOften the subject expert, because typing keeps them engaged and in control of accuracy, but there's no fixed rule. Switching partway through can help when one person is flagging.\n\n**Is pair writing slower?**\n\nIt ties up two people for the length of the session, which can look expensive. Set that against the full cost of the alternative – drafting, several review rounds, clarifying [emails](https://contentlexicon.com/email) and rework – and pairing often comes out ahead for anything complex.\n\n**How do you persuade a subject expert to try it?**\n\nStart with the time it saves them. Most experts have sat through enough review rounds to welcome anything that cuts them short, and an hour of pairing is easier to agree to than another week of tracked changes. Offer one session on something that's already stuck, such as a page that keeps coming back from sign-off, and suggest they take the keyboard so they can see that nothing goes in without their agreement. Much of the wariness comes from having had their words changed in the past without explanation. Pairing deals with that directly, because every change is discussed while they're in the room, and an expert who has seen why a sentence was simplified is far less inclined to put the [jargon](https://contentlexicon.com/jargon) back.\n\n**What happens when the two people disagree?**\n\nIt's usually useful, because a disagreement about wording tends to be a disagreement about meaning or audience that would otherwise have surfaced three review rounds later. It helps to agree at the start who has the final say on what: the subject expert on accuracy, the content designer on whether the audience will understand it. When neither can settle a point, evidence from user research, search data or a quick test with real readers does better than a longer argument, and asking a model to write the sentence both ways can make the difference easier to see. Anything still unresolved can be parked with a note so the session doesn't lose half an hour to one sentence.\n\n## Learn more\n\n- [How to get started with pair writing](https://uxdesign.cc/how-to-get-started-with-pair-writing-8fdf3a38d69e) – Ove Dalen's practical lessons from more than a decade of using the method in product teams.\n- [How writing in pairs improves content](https://digitalblog.ons.gov.uk/2021/01/26/how-writing-in-pairs-improves-content) – an account from the Office for National Statistics of rewriting a careers page through remote pair writing.\n- [On Pair Programming](https://martinfowler.com/articles/on-pair-programming.html) – Birgitta Böckeler and Nina Siessegger's guide to the software practice pair writing borrows from, including the driver and navigator roles and the benefits and challenges of working in pairs.\n- [A Real Writer's Guide to Co-Writing with AI](https://medium.com/@tracyantonioli/a-real-writers-guide-to-co-writing-with-ai-cf0a49595ac7) – Tracy Antonioli's step-by-step account of writing an article with AI as a thinking partner, from brain-dumping ideas and testing structures to checking research and keeping every final call for herself.\n- [From Pen to Prompt: How Creative Writers Integrate AI into their Writing Practice](https://arxiv.org/html/2411.03137v4) – a 2025 academic report by Alicia Guo, Amy Zhang, Jeffrey Heer and colleagues at the University of Washington, based on interviews with writers on how they use AI in their work. Participants variously describe AI as an assistant, an actor they direct, a co-writer, a junior writer, a ghostwriter, an editor, a muse and a tool." }, { "term": "PDF", "alsoCalled": [ "Portable Document Format" ], "slug": "pdf", "definition": "A fixed-layout document format often used for reports, forms and downloads – useful sometimes, overused often.", "category": "Structure and systems", "related": [ "accessibility", "content-inventory", "paddle-swim-dive", "structured-content" ], "updated": "2026-10-07T13:35:00.000Z", "url": "https://contentlexicon.com/pdf", "markdown": "https://contentlexicon.com/pdf.md", "credit": "PDF, from the Content Lexicon by Contentious (https://contentlexicon.com/pdf), CC BY 4.0", "body": "PDF stands for Portable Document Format. It was designed to preserve the appearance of a document across devices and software, making it useful for print-ready material, signed records, designed publications and files whose layout must remain fixed.\n\nThat strength can become a weakness on the web. A fixed page designed for A4 paper is often awkward on a phone, harder to navigate, less adaptable and more difficult to maintain than an ordinary web page. [Content teams](https://contentlexicon.com/content-team) should treat PDF as a deliberate format choice, not the ceremonial final form of anything that once passed through Microsoft Word.\n\n## Why it matters for content\n\nPDFs often contain important material: reports, policies, forms, research and guidance. If that content is hard to search, read or use, the format creates a barrier around precisely the information people came for. Long documents need proper headings, bookmarks, links and a logical reading order. Images need alternatives and forms need labelled controls.\n\nMaintenance is another concern. A PDF can be downloaded, copied and linked from elsewhere long after a newer version appears. Updating the [website](https://contentlexicon.com/website) may leave old files circulating without visible context. Teams need clear dates, version information, canonical locations and a retirement process.\n\nA web page is usually better for material that changes regularly, supports an online task or needs to work fluidly across devices. PDFs earn their place when the bounded document itself matters – for example, a printable report, formal record or designed publication – and when an accessible version can be produced and maintained.\n\n## How AI changes the picture\n\nAn assistant can help someone locate the assumptions behind a finding, compare sections of a report or ask for an explanation of an unfamiliar term. The document becomes easier to question as well as summarise. Answers need references back to the relevant pages, especially where a table, qualification or footnote changes the meaning. Source-based tools such as NotebookLM show one direction this reading experience can take; they don't establish that every PDF can be interpreted reliably.\n\nPDF extraction needs care. Scanned pages may require optical character recognition, complex layouts can scramble reading order and tables may lose their relationships. Teams should preserve structured source content where possible and check extracted text before relying on it. Help with reading or identifying likely [accessibility](https://contentlexicon.com/accessibility) problems doesn't make the underlying file accessible.\n\nDates, versions, titles and [metadata](https://contentlexicon.com/metadata) help systems select the right file. Where the same facts exist in web pages, PDFs and internal documents, a canonical source should be explicit. Readers should be able to tell which version an answer draws on and return to it to check the detail.\n\n## Questions people ask\n\n**Are PDFs inaccessible?**\n\nNot inherently, but accessible PDFs require proper structure and testing. Untagged, scanned or badly exported files can be very difficult for screen-reader and keyboard users.\n\n**When should we use a web page instead?**\n\nPrefer HTML when content changes often, supports a task, needs responsive layout or should be easy to search, link and reuse. Offer a PDF as an additional format where printing or fixed presentation adds value.\n\n**Can we make a PDF accessible after it has been designed?**\n\nOften, but remediation can be laborious and imperfect. Accessibility is easier when the source document has correct headings, reading order, link text and alternatives from the start.\n\n**Can AI safely summarise a PDF?**\n\nIt can help, but check that the text was extracted correctly and verify consequential claims against the document. A confident summary cannot restore a missing column or unreadable scan.\n\n## Learn more\n\n- [Creating accessible PDFs](https://accessibility.huit.harvard.edu/pdf) – practical guidance and a frank comparison with the accessibility and maintainability of HTML.\n- [Create and verify PDF accessibility](https://helpx.adobe.com/acrobat/using/create-verify-pdf-accessibility.html) – detailed steps for tagging, checking and repairing documents.\n- [Longing for longer](https://contentious.ltd/thoughts/longing-for-longer) – Contentious considers how reports can become immersive, navigable digital stories, without making length an end in itself." }, { "term": "Persona", "alsoCalled": [], "slug": "persona", "definition": "A research-based representation of a group of people with shared goals, behaviour or needs.", "category": "Research and evidence", "related": [ "audience-segmentation", "content-brief", "core-content-strategy", "tone", "user-needs" ], "updated": "2026-10-15T08:16:00.000Z", "url": "https://contentlexicon.com/persona", "markdown": "https://contentlexicon.com/persona.md", "credit": "Persona, from the Content Lexicon by Contentious (https://contentlexicon.com/persona), CC BY 4.0", "body": "A persona turns research findings into a character: one named, described person who stands in for many real ones. It usually sets out their context and motivation, the tasks they're trying to complete and what gets in their way. The fiction gives a team someone specific to design and write for, instead of an abstract \"user\" who can be bent to fit any argument.\n\nA persona isn't one research participant, a marketing segment or a demographic sketch. Names, photographs and implausibly specific coffee orders add memorability, not validity; a persona's value comes from the evidence underneath and how directly it informs decisions. Personas entered software design through Alan Cooper, who role-played a user drawn from his interviews while designing a project management tool in the early 1980s and later set out the method in The Inmates Are Running the Asylum, though marketers had been describing archetypal customers for decades before.\n\n## Why it matters for content\n\nPersonas help [content teams](https://contentlexicon.com/content-team) see that different groups need different things from the same page. A first-time applicant needs reassurance and plain steps; a returning professional wants the reference table and nothing else. Named, evidence-based characters make those differences easier to discuss and prioritise than \"our audience\" or the preferences of whoever is in the meeting.\n\nThe details that earn their place are the ones that change content decisions: what someone already knows, what they're trying to get done, what's at stake for them, what device and setting they're in and what might stop them. Age, job title or family situation belong only when they affect those things. Personas also need maintaining. Built once and pinned to a wall, they harden into polite stereotypes and hide the people who don't fit the dominant pattern.\n\nSome researchers argue for dropping the character altogether. Indi Young's thinking styles, for example, group people by how they reason towards one goal, with no demographics, on the grounds that demographics rarely explain behaviour. For content work that lens is often more useful, because it ties directly to the decision someone is making on the page.\n\n## How AI changes the picture\n\nGenerative AI can produce a persona, complete with backstory and stock photo, in seconds, and that ease is the risk. A model asked for \"a typical small business owner\" returns a plausible average of everything it has read, which is fiction dressed up as insight. Synthetic users like this can help brainstorm questions to ask in research, but they aren't evidence, and treating them as evidence lets invented needs steer real decisions.\n\nAI is more useful with the research itself: clustering interview notes, tagging support tickets by goal or pulling recurring frustrations out of hundreds of survey responses. A researcher can then build personas from patterns they've checked against the source. Personas can also become test [prompts](https://contentlexicon.com/prompt), such as asking a model to read a page as a nervous first-time applicant. That can surface gaps, provided everyone remembers it's simulating the persona, not a person.\n\n## Questions people ask\n\n**How many personas do we need?**\n\nEnough to cover differences that change a content decision, and usually fewer than teams expect. A small service might need only two if the main contrast is between occasional applicants and the staff processing their applications. Add another when research reveals a need the existing ones miss.\n\n**Should personas include demographics?**\n\nOnly where they affect the task or experience. Someone's age might matter for a pension page and be irrelevant to a parking permit. Behaviour, context and existing knowledge usually tell a [writer](https://contentlexicon.com/writer) more than a biography.\n\n**What's the difference between a persona and an audience segment?**\n\nA segment is a grouping, often defined by shared characteristics for targeting or analysis. A persona takes a pattern from research and gives it a face and a story, so it's easier to keep in mind while writing. The two can coexist, with one segment represented by one or more personas.\n\n**Are personas still worth making?**\n\nThey are when they're grounded in research, used in real decisions and updated as behaviour changes. They're much less useful when they exist mainly as a deliverable. If nobody can point to a decision a persona has changed, it isn't doing its job.\n\n## Learn more\n\n- [The origin of personas](https://web.archive.org/web/2010/http://www.cooper.com/journal/2008/05/the_origin_of_personas.html) – Alan Cooper's account, via the Internet Archive, of improvising his first persona while designing software in 1983 and later formalising the method in client projects.\n- [7 Core Ideas about Personas and The User Experience](https://measuringu.com/personas-ux/) – Jeff Sauro's overview of what personas are for, including their earlier roots in marketing and the work of Angus Jenkinson alongside Cooper's.\n- [Explanations – Thinking Styles](https://indiyoung.com/explanations-thinking-styles/) – Indi Young on grouping people by how they think about one goal, free of demographics, as an alternative to conventional personas." }, { "term": "Plain language", "alsoCalled": [ "Plain English" ], "slug": "plain-language", "definition": "Writing that helps people understand what they need quickly and clearly, the first time they read it.", "category": "Editorial craft", "related": [ "accessibility", "front-loading", "inclusive-language", "jargon", "paddle-swim-dive", "readability" ], "updated": "2026-10-11T06:03:00.000Z", "url": "https://contentlexicon.com/plain-language", "markdown": "https://contentlexicon.com/plain-language.md", "credit": "Plain language, from the Content Lexicon by Contentious (https://contentlexicon.com/plain-language), CC BY 4.0", "body": "Plain language doesn't mean sounding childish, stripping out complexity or pretending difficult subjects are simple. It's about expressing the substance clearly, using words, structure and emphasis that help people find and grasp what matters.\n\nIn practice, that usually means familiar words, direct sentences, clear hierarchy, good examples and an order that reflects the reader's task rather than the [writer](https://contentlexicon.com/writer)'s train of thought. It's as much about structure as wording.\n\n## Why it matters for content\n\nClear writing makes content easier to use. It reduces the effort needed to understand instructions, compare options, make decisions and complete tasks. It's especially valuable where people are stressed, unfamiliar with the subject, short on time or trying to act rather than browse.\n\nIt also improves editorial quality. Writing clearly forces a team to decide what it means, what the reader needs first and which details are carrying their weight. That makes plain language useful for public-facing copy and for internal guidance, product content, policy and technical material alike.\n\n## How AI changes the picture\n\nGenerative tools are very good at producing vague, padded, plausible-sounding copy: the benefits letter that says “you may be eligible for support” and never says how much, or by when. If a team doesn't have clear standards for directness, AI can multiply that mush quickly. Clear source material, explicit [voice](https://contentlexicon.com/voice) and [tone](https://contentlexicon.com/tone) guidance and strong editorial standards all make better AI output more likely.\n\nThere's a second reason to care: machines retrieve, summarise and transform clear content more reliably than ambiguous content. Plain language is still primarily for people, but it often improves machine use as well.\n\n## Questions people ask\n\n**Are shorter words the answer?**\n\nPartly. Shorter words can help, but plain language is about clarity. A familiar long word can be clearer than an awkward short one if it's the right word.\n\n**Does plain language mean dumbing things down?**\n\nNo. It means respecting the reader enough to be clear. Complex subjects still need complexity where the subject demands it; they don't need unnecessary difficulty on top.\n\n**Can technical or specialist content be plain?**\n\nYes. You may still need specialist terms, but you can define them, order information sensibly and avoid making the surrounding language harder than it needs to be.\n\n**Can a readability score tell me if something is plain?**\n\nNot on its own. [Readability](https://contentlexicon.com/readability) tools can be useful signals, but they can't tell whether the structure makes sense, whether the key point arrives soon enough or whether the reader can act on what they've read.\n\n## Learn more\n\n- [Use clear language](https://guidance.publishing.service.gov.uk/writing-to-gov-uk-standards/writing-guidelines/clear-language/) – the UK government's plain English rules for its own website, with words to avoid and limits on sentence and paragraph length.\n- [Plain language guide series](https://digital.gov/guides/plain-language) – practical guidance on writing for understanding, from the US government.\n- [Plain Language Is for Everyone, Even Experts](https://www.nngroup.com/articles/plain-language-experts/) – Hoa Loranger's Nielsen Norman Group article on why specialists and professionals also prefer clear, jargon-free information." }, { "term": "Prompt", "alsoCalled": [], "slug": "prompt", "definition": "An instruction or request given to an AI system.", "category": "AI and tooling", "related": [ "agentic-ai", "content-brief", "context-engineering", "context-window", "llm", "skill", "system-prompt" ], "updated": "2026-10-11T12:33:00.000Z", "url": "https://contentlexicon.com/prompt", "markdown": "https://contentlexicon.com/prompt.md", "credit": "Prompt, from the Content Lexicon by Contentious (https://contentlexicon.com/prompt), CC BY 4.0", "body": "A prompt gives an AI system a starting point for generating a response or taking action. It can take the form of a question, a command or a fuller brief containing background information, examples and requirements. In systems that accept more than text, a prompt can also include an image or an [audio](https://contentlexicon.com/audio) recording, such as a photograph accompanied by a request to describe it.\n\nThe system interprets the prompt alongside any other context available to it, including earlier exchanges and instructions set by the application. A follow-up such as “make it shorter” depends on that context for its meaning. A prompt influences the result, but it isn't a fixed command that guarantees the same response every time.\n\n## Why it matters for content\n\nPrompts matter because they shape what AI systems produce. For [content teams](https://contentlexicon.com/content-team), that includes drafting, rewriting, summarising, classifying, checking, transforming and comparing content. A vague prompt usually creates vague output. A stronger prompt makes the task clearer, sets boundaries and gives the system better material to work with.\n\nFor a repeated task, preserve the prompt alongside examples and the checks used to judge it. At that point, prompting stops being an individual trick and becomes part of a repeatable content [workflow](https://contentlexicon.com/workflow). The prompt is effectively a small piece of operational content: it encodes standards, sources, audience, [tone](https://contentlexicon.com/tone) and process.\n\nPrompts also reveal whether a team can describe its own work clearly. If you can't explain what good looks like to a machine, there's a fair chance the human brief was foggy too. That makes prompt-writing (or, more grandly, “prompt engineering”) a useful discipline, because it makes the team’s assumptions visible.\n\n## How prompts change the picture\n\nA prompt becomes part of an interaction rather than only a brief to a [writer](https://contentlexicon.com/writer). Readers may ask follow-up questions or request a different explanation, so the system needs to retain relevant context and distinguish a new request from quoted source material.\n\nFor repeated work, test prompts on varied examples, keep versions and record failures. A model update or a changed source can affect behaviour even if the wording stays the same. Permissions and review belong to the wider system; instructions alone don't enforce them.\n\n## Questions people ask\n\n**Is prompting just about finding the right wording?**\n\nWording matters alongside source material, context and review criteria. A request can be beautifully phrased and still ask for a claim the evidence doesn't support.\n\n**Do prompts need to be long?**\n\nNot necessarily. Some tasks need little more than a sentence. Others need a fuller brief with context, examples, constraints and output format. Length isn't the goal; sufficient information is.\n\n**Can one good prompt be reused?**\n\nOften yes, especially for repeated tasks. But reusable prompts should be tested against real examples and updated when the work, sources or standards change.\n\n**Is prompt engineering still a useful term?**\n\nSometimes, but many teams now find it more useful to think in terms of context, workflow and source design rather than wording alone. The prompt is one part of the system, not the whole machine.\n\n## Learn more\n\n- [Prompt engineering](https://developers.openai.com/api/docs/guides/prompt-engineering) – OpenAI’s API guidance on instructions, context and evaluating behaviour; examples and recommended techniques are product-specific.\n- [Effective context engineering for AI agents](https://www.anthropic.com/engineering/effective-context-engineering-for-ai-agents) – Anthropic’s account of managing context across agent work, including retrieval and longer-running tasks; a broader perspective than prompt wording alone.\n- [Getting started with AI: Good enough prompting](https://www.oneusefulthing.org/p/getting-started-with-ai-good-enough) – Ethan Mollick’s 2024 case for treating AI like an infinitely patient new colleague who forgets everything, and for saving elaborate prompt engineering for prompts used at scale." }, { "term": "Proofreading", "alsoCalled": [], "slug": "proofreading", "definition": "The final check for typos, formatting mistakes and small errors before publication.", "category": "Editorial craft", "related": [ "automated-testing", "copy-editing", "editing", "style-guide" ], "updated": "2026-10-11T06:28:00.000Z", "url": "https://contentlexicon.com/proofreading", "markdown": "https://contentlexicon.com/proofreading.md", "credit": "Proofreading, from the Content Lexicon by Contentious (https://contentlexicon.com/proofreading), CC BY 4.0", "body": "Proofreading checks content after the words, structure and meaning have been agreed. It looks for errors that have survived [editing](https://contentlexicon.com/editing) or crept in along the way: misspellings, inconsistent capitals, wrong numbers, broken links, misplaced headings, missing [alt text](https://contentlexicon.com/alt-text) and lines that wrap badly. It's done on the version people will see – the rendered web page, the typeset [PDF](https://contentlexicon.com/pdf), the [email](https://contentlexicon.com/email) as it lands – because layout creates errors the raw text never showed.\n\nThe limits of the job matter as much as the checks. A proofreader doesn't rewrite sentences, restructure an argument or verify facts unless that's been agreed. If a proof reveals something bigger, such as a paragraph that contradicts the policy it describes, the problem goes back to the right editor rather than being slipped into the final pass. The name comes from printing, where a proof was a trial impression pulled from set type so corrections could be marked before the full run.\n\nA title can be right in the draft and still break awkwardly across two lines on a phone. A link can point to the [staging](https://contentlexicon.com/staging) site once the page goes live. Neither shows up when reviewing a document.\n\n## Why it matters for content\n\nSmall errors have disproportionate costs. In an interface, a wrong label or a dead link can stop someone finishing a task. In a report, a transposed figure can change a decision. Even when the meaning survives, typos signal carelessness, and readers tend to extend that judgement to everything else on the page, fairly or not.\n\nProofreading also protects the work done earlier. Editing and approval settle what the content should say; proofreading confirms that's what got published. A short checklist keeps it focused: names, dates, numbers, links, headings, [punctuation](https://contentlexicon.com/punctuation), alt text and consistency with the [style guide](https://contentlexicon.com/style-guide). Agree the scope in advance so the final check doesn't turn into an unplanned rewrite the day before launch.\n\n## How AI changes the picture\n\nSpelling and [grammar](https://contentlexicon.com/grammar) checkers have been around for decades. Language models extend them to style and consistency across thousands of pages, which makes them good at a first sweep: every \"e-mail\" on a site that uses \"email\", every heading in title case when the house style is [sentence case](https://contentlexicon.com/sentence-case), every inconsistent spelling of a product name.\n\nTheir weak spot is the error that looks right. A model reading for fluency will pass a word that's spelt correctly but wrong for the context, accept a plausible date that's a year out and leave a link that resolves to the wrong page. It may also \"fix\" deliberate choices, like contractions or a fragment used for effect, unless the style guide is part of its instructions.\n\nGenerated text raises the stakes. Fluent prose makes errors harder for people to spot, so AI drafts need proofreading at least as careful as human ones, with names, numbers and links checked against their sources rather than read for sense.\n\n## Questions people ask\n\n**How is proofreading different from copy editing?**\n\n[Copy editing](https://contentlexicon.com/copy-editing) works on the draft, improving clarity, consistency, grammar and style within an agreed scope. Proofreading comes afterwards, on the final layout, and is limited to correcting errors. Some teams combine the two for short pieces, but they call for different kinds of attention.\n\n**Why is it so hard to spot your own typos?**\n\nBecause you know what you meant. Reading your own text, the brain supplies the intended words and skims past what's on the page. Changing the font, printing it out, reading it aloud or leaving it overnight makes the text unfamiliar enough to see. A second reader is better still.\n\n**Is spellcheck enough?**\n\nIt catches misspellings and not much else. It won't tell you that a link goes to the wrong page, a figure doesn't match the source table or a heading has been styled as body text. Those need someone looking at the rendered page.\n\n**Do proofreaders still use correction marks?**\n\nSome do. The standard symbols for marking up proofs, set out in British and international standards, are now mostly applied as stamps or annotations on PDFs. For web content, comments in a shared document or a ticket against the live page are more common. Either works if every correction is unambiguous to whoever makes it.\n\n## Learn more\n\n- [What is proofreading?](https://www.ciep.uk/resource/what-is-proofreading.html) – the Chartered Institute of Editing and Proofreading on the proofreader's role as the final quality check, including the work that falls outside it, such as rewording and fact-checking.\n- [What's Up With That: Why It's So Hard to Catch Your Own Typos](https://www.wired.com/2014/08/wuwt-typos/) – Nick Stockton's Wired piece, in which psychologist Tom Stafford explains why writers read for meaning and miss their own slips, and how making text look unfamiliar helps.\n- [New standard for proof reading](https://committee.iso.org/sites/tc130/home/resources/articles/content-left-area/previous-articles/proof-reading-symbols.html) – David Penfold of ISO's printing technology committee on how proof correction marks developed from the British standard into a 2016 international version covering Chinese and Japanese, and how they're now used on PDF proofs." }, { "term": "Pull request", "alsoCalled": [ "Merge request" ], "slug": "pull-request", "definition": "A proposal to merge changes, keeping the differences, discussion and checks together for review.", "category": "AI and tooling", "related": [ "content-governance", "copy-editing", "git" ], "updated": "2026-10-15T14:14:00.000Z", "url": "https://contentlexicon.com/pull-request", "markdown": "https://contentlexicon.com/pull-request.md", "credit": "Pull request, from the Content Lexicon by Contentious (https://contentlexicon.com/pull-request), CC BY 4.0", "body": "A pull request asks for a set of changes on one [branch](https://contentlexicon.com/branch) of a repository to be merged into another, usually the main version. It gathers the changed material, an explanation, review comments and the results of automated checks in one place, so the work can be reviewed before it becomes part of the shared source. GitLab calls the same thing a merge request.\n\nMost pull requests follow the same sequence. Someone creates a branch, a separate line of work, and makes their changes there. They open a pull request describing what changed and why. The platform shows a diff, the line-by-line difference between the branch and the main version. Reviewers comment on specific lines, suggest edits or request changes, and the author updates the branch in response. Automated checks run on each update. When the required approvals and checks are in place, someone merges the branch and the change becomes part of the main version.\n\n## Why it matters for content\n\nMost [content teams](https://contentlexicon.com/content-team) already review changes, usually through tracked changes or suggestion mode in Word and Google Docs. That handles the edit itself well, showing what changed and gathering comments in place. The weaker point is what happens next. The approved text is copied into a [CMS](https://contentlexicon.com/cms), sometimes picking up fresh edits on the way, and the record of who agreed to what stays in a document that rarely gets opened again.\n\nA pull request closes that gap, because the change that was reviewed is the change that gets published. The discussion and approval stay attached to the content after it goes live, so anyone asking later why a page changed can find out. Related edits can travel together: a new product name across forty pages is reviewed as one change, published at once and reversed at once if needed. Automated checks can test links, [alt text](https://contentlexicon.com/alt-text) and house style on every update, and a preview shows the page as readers will see it.\n\nDocs-as-code teams got there first, since their documentation already lived beside the product code, but the approach suits any content where accuracy matters, from product pages and policies to interface copy. Git-based content management systems such as Decap CMS open and merge pull requests behind ordinary save and publish buttons, so [writers](https://contentlexicon.com/writer) needn't see a branch. Where content stays in a conventional CMS, the habits still transfer: a stated reason for each change, a clear approver and a record that outlasts the conversation.\n\n## How AI changes the picture\n\nAI can suggest style changes, summarise a large change and point reviewers towards likely risks before a person reads it. Link and [accessibility](https://contentlexicon.com/accessibility) tests may use ordinary automation instead. Neither an AI review nor a passing test establishes that the whole change is sound.\n\nAutomated review should support accountability rather than replace it. Models can miss context or object to deliberate wording. A named person still decides whether the content is accurate, appropriate and safe to publish.\n\nAI-generated changes also need provenance. When an agent opens a pull request, reviewers should see the source, instructions and scope, rather than judging polished output without knowing how it arose.\n\n## Questions people ask\n\n**Does every change need two people?**\n\nNo. Match review to risk. Minor spelling fixes may be self-approved, while legal or high-impact changes need specialist scrutiny.\n\n**What should a reviewer check?**\n\nAccuracy, [user need](https://contentlexicon.com/user-needs), meaning, dependencies, accessibility, consistency and possible breakage, rather than personal preference. A changed price needs its source checked. A changed warning needs to be read in the context where it appears.\n\n**Is a pull request just an approval workflow?**\n\nIt does more than a yes-or-no gate, because it keeps the change, context, discussion and evidence together and preserves them after the merge.\n\n**Can AI approve a change?**\n\nIt can run checks and recommend action. Organisational accountability should stay with an authorised person.\n\n## Learn more\n\n- [Pull request reviews](https://docs.github.com/en/pull-requests/reference/pull-request-reviews) – the mechanics of commenting on exact changes, requesting revisions and recording approval.\n- [Editing workflows through pull requests](https://idratherbewriting.com/2015/09/10/edits-through-pull-requests) – Tom Johnson's documentation-review account, including the learning curve for reviewers using branches and reused source files.\n- [Docs as Code](https://www.writethedocs.org/guide/docs-as-code/) – the wider philosophy of bringing versioning, review and shared tools into documentation work." }, { "term": "Punctuation", "alsoCalled": [], "slug": "punctuation", "definition": "Marks that help show sentence structure, relationships and emphasis.", "category": "Editorial craft", "related": [ "copy-editing", "proofreading", "sentence-case", "style-guide", "word-list" ], "updated": "2026-10-11T12:40:00.000Z", "url": "https://contentlexicon.com/punctuation", "markdown": "https://contentlexicon.com/punctuation.md", "credit": "Punctuation, from the Content Lexicon by Contentious (https://contentlexicon.com/punctuation), CC BY 4.0", "body": "Full stops, commas, colons, dashes, brackets and quotation marks show boundaries, relationships, pauses and emphasis. They don't reproduce speech exactly, and their interpretation can vary by context. A comma can change which words belong together, not merely where a reader pauses.\n\nPunctuation also carries house style. Teams need shared choices about dashes, quotation marks, lists, capitalisation after colons and the Oxford comma. These choices rarely have one morally correct answer; they do have consequences for clarity and consistency.\n\n## Why it matters for content\n\nPunctuation changes meaning. Commas can identify which information is essential, apostrophes distinguish possession from plurals and a badly placed full stop can detach an instruction from its condition. In digital content, punctuation also affects scannability, screen-reader pauses and the rhythm of short interface messages.\n\nA [style guide](https://contentlexicon.com/style-guide) should settle the recurring decisions so [writers](https://contentlexicon.com/writer) don't relitigate them page by page. Consistency keeps attention on the meaning rather than unexplained changes in convention.\n\nRules should serve readers rather than produce typographical theatre. Dense semicolons may suit a legal argument but hinder a short instruction. Exclamation marks don't make unclear guidance friendlier. Lists need parallel structure and enough punctuation to show whether their items are fragments or sentences.\n\n## How AI changes the picture\n\nAI can apply house punctuation at scale, find inconsistencies and transform copy between conventions. It's particularly useful when a team has documented choices clearly and supplied examples. “Fix the punctuation” is a less reliable instruction than a style guide that states which marks to use and why.\n\nModels can also introduce US English conventions into non-US content, excessive dashes or a suspicious abundance of perfectly balanced clauses. Automated changes may alter meaning, especially in legal, technical or safety-critical content. A final check should consider the sentence, not merely whether the chosen marks match a pattern.\n\nPunctuation appears in [prompts](https://contentlexicon.com/prompt), structured fields and [machine-readable](https://contentlexicon.com/machine-readable) text too. A curly quotation mark may be elegant in prose but troublesome in code or data. [Content teams](https://contentlexicon.com/content-team) need to know when punctuation is expressive language and when it's part of a technical format that must remain exact.\n\n## Questions people ask\n\n**What is the difference between a hyphen, en dash and em dash?**\n\nA hyphen joins words. An en dash commonly marks ranges and, in British style, interruptions with spaces around it. An em dash often marks interruptions without spaces in American style. An em dash by itself doesn't establish AI authorship; follow the relevant style convention rather than treating a mark as a detector, or necessarily a bad thing.\n\n**Do we need the Oxford comma?**\n\nUse the convention your style guide sets, but always add a comma when it prevents ambiguity. Consistency matters, though meaning matters more.\n\n**Should bullet points end with full stops?**\n\nUse full stops when the items are complete sentences. Fragments usually need none. Keep the structure consistent within the list.\n\n**Can punctuation improve accessibility?**\n\nClear punctuation supports comprehension and can help assistive technologies pause sensibly. It can't rescue long, tangled sentences, so edit the language as well as the marks.\n\n## Learn more\n\n- [The Punctuation Guide](https://www.thepunctuationguide.com/) – A reference index for individual marks, with a separate British-versus-American-style comparison; regional conventions still vary by publication.\n- [University of Southampton punctuation guide](https://cdn.southampton.ac.uk/assets/imported/transforms/content-block/UsefulDownloads_Download/3ADEB0E8DABE4098A43CF45F8C915D2A/punctuation_guide.pdf) – a welcome reminder that punctuation is an art as well as a logical system.\n- [Commas, colons and capitalisation](https://contentious.ltd/resources/commas-colons-and-capitalisation) – a short tour of reliable online style guides for settling everyday language decisions." }, { "term": "RACI", "alsoCalled": [ "RACI matrix", "Responsibility assignment matrix" ], "slug": "raci", "definition": "A responsibility matrix that identifies who is responsible, accountable, consulted and informed for each task or decision.", "category": "Governance", "related": [ "content-owner", "sign-off", "stakeholder-mapping" ], "updated": "2026-10-11T06:25:00.000Z", "url": "https://contentlexicon.com/raci", "markdown": "https://contentlexicon.com/raci.md", "credit": "RACI, from the Content Lexicon by Contentious (https://contentlexicon.com/raci), CC BY 4.0", "body": "RACI stands for responsible, accountable, consulted and informed. A RACI matrix lists tasks or decisions on one axis and roles on the other, then marks how each role is involved.\n\nResponsible people do the work. The accountable person owns the result and makes the final call. Consulted people are asked for their view; informed people are kept up to date without being asked to decide. The letters are simple; agreeing them can be bracing.\n\n## Why it matters for content\n\nContent work crosses subject experts, [writers](https://contentlexicon.com/writer), designers, legal teams, service owners and publishers. A lot of content disputes come down to a missing A: a pricing page collects six rounds of comments from product, legal and sales, and none of them is sure they're the one who signs it off. Too many Cs cause the opposite problem, turning every decision into a consultation and making [sign-off](https://contentlexicon.com/sign-off) take three weeks. RACI makes both problems visible.\n\nWhere RACI earns its keep is recurring work, such as publishing guidance, reviewing product copy or responding to a factual error. The matrix should cover decisions as well as tasks. “Approve a change to eligibility wording” often matters more than “attend review meeting”.\n\nFor content work, the biggest gain often comes from the conversation the grid forces rather than the grid itself. Once a team has argued through who holds the A on a campaign page, people start carrying the distinction into work nobody has mapped. A writer knows whose call the headline is. A founder knows their comment is a C, not a veto. A policy lead stops treating “I wasn't consulted” and “I wasn't told” as the same complaint. If the shared understanding of who answers for what outlasts the matrix, you’re winning.\n\n## How AI changes the picture\n\nAI introduces new responsibilities without removing the old ones. Someone selects the tool, someone provides sources, someone reviews output and someone remains accountable for what reaches the audience. “The model did it” is not a fifth RACI role.\n\nAutomation can perform part of a responsible role, such as flagging broken links or drafting [metadata](https://contentlexicon.com/metadata). Accountability still belongs to a person with authority and enough understanding to challenge the result. [Prompts](https://contentlexicon.com/prompt), [system prompts](https://contentlexicon.com/system-prompt) and the sources a tool draws on need rows of their own, because someone should be accountable for both the standing instructions and the drafts produced from them.\n\n## Questions people ask\n\n**How many accountable people should there be?**\n\nOne for each task or decision. Cornell's guidance insists on a single A, and shared accountability has a way of becoming nobody's.\n\n**Can one person be responsible and accountable?**\n\nYes. In a small team, the same person may do the work and own the outcome.\n\n**Does a RACI need to cover every task?**\n\nNo. Most teams get further by starting with the decisions that stall: sign-off on a fundraising appeal, retiring a product page, taking down help content that's gone out of date. Twenty rows people use will do more than two hundred nobody opens.\n\n**Are there alternatives to RACI?**\n\nSeveral, with confusingly similar acronyms. RASCI adds S for support, for people who help without owning the work. DACI (driver, approver, contributors, informed) suits work where the sticking point is a decision, not delivery. Bain & Company's RAPID (recommend, agree, perform, input, decide) splits decisions more finely: useful when the stakes are high, though more machinery than a small team usually needs.\n\n## Learn more\n\n- [RACI and RASCI Definitions](https://it.cornell.edu/it-service-management/raci-and-rasci-definitions) – Cornell University's definitions and practical guidance, including one accountable role per item.\n- [DACI Decision-Making Framework](https://www.atlassian.com/wac/team-playbook/plays/daci) – Atlassian's team playbook for running a decision with DACI, including why the approver is one person.\n- [RAPID® Decision Making](https://www.bain.com/insights/rapid-decision-making/) – Bain & Company's explainer on its RAPID model and the five roles it assigns in a decision." }, { "term": "RAG", "alsoCalled": [ "Retrieval-augmented generation" ], "slug": "rag", "definition": "An approach in which an AI system retrieves relevant material and uses it as context when generating an answer.", "category": "AI and tooling", "related": [ "context-window", "geo", "hallucination", "knowledge-base", "knowledge-graph", "llm", "single-source-of-truth", "structured-content" ], "updated": "2026-10-07T06:41:00.000Z", "url": "https://contentlexicon.com/rag", "markdown": "https://contentlexicon.com/rag.md", "credit": "RAG, from the Content Lexicon by Contentious (https://contentlexicon.com/rag), CC BY 4.0", "body": "RAG stands for retrieval-augmented generation. Instead of relying only on the model's training data or general memory, a RAG system retrieves passages that appear relevant, adds them to the [prompt](https://contentlexicon.com/prompt) and asks the model to answer from those.\n\nIn practice, this is how many \"ask our docs\" tools, internal assistants and [website](https://contentlexicon.com/website) chatbots work. The value isn't that the model becomes magically trustworthy. It's that the model is at least trying to answer from the right shelf.\n\n## Why it matters for content\n\nRAG turns content quality into part of AI quality. If source material is current, structured, well-labelled and deduplicated, answers tend to improve. If the source material is contradictory, out of date or badly organised, the system can retrieve the wrong thing very efficiently. That makes [content operations](https://contentlexicon.com/content-operations) central to whether a RAG system works.\n\nRAG has two distinct stages: finding material and writing from it. Source documents are usually split into chunks and indexed; when someone asks a question, the system finds chunks that appear relevant and puts them in front of the model. Headings, self-contained sections and enough context to identify what a passage refers to all affect what gets found. A perfectly written answer can't rescue a poor retrieval result.\n\nIt's most useful where an organisation has a body of material people need answers from: policies, service information, product documentation, guidance, research, support content or internal know-how. And it earns its keep where answers need to reflect current organisational knowledge rather than generic web knowledge.\n\n## How RAG changes the picture\n\nWhat changes for [content teams](https://contentlexicon.com/content-team) isn't only the tooling but the level of consequence. Duplicates, stale information, weak [metadata](https://contentlexicon.com/metadata) and ambiguous ownership stop being tidy-up issues and can affect the answers the system gives. RAG creates a practical reason to care about governance, retirement and single sources of truth, because the model works from the material retrieved, which may not be the source you intended.\n\nTake a college that uploads this year's fees [PDF](https://contentlexicon.com/pdf) to its assistant's index and leaves last year's in place. A prospective student asks what the course costs. Without a rule favouring approved, current versions, retrieval may surface the old document, and the assistant may quote it fluently, with a citation that makes it look checked. The old PDF was never taken down.\n\n## Questions people ask\n\n**What's the difference between RAG and a chatbot?**\n\nA chatbot is an interface. RAG is one way of giving that chatbot access to source material before it answers.\n\n**Does RAG stop hallucinations?**\n\nIt can reduce them, but it doesn't remove them. A model can still misunderstand, overstate or combine information badly. RAG improves the chances of grounding an answer; it doesn't guarantee truth.\n\n**Do we need perfect content before using RAG?**\n\nYou don't need perfect content, but you do need to know what mess you're feeding it. RAG often reveals content problems that were already there: duplicates, stale information, contradictory guidance and missing metadata.\n\n**Is RAG only for technical teams?**\n\nTechnical teams build the system, but content teams often decide whether the source material is clear, current and structured enough to retrieve from sensibly.\n\n## Learn more\n\n- [Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks](https://arxiv.org/abs/2005.11401) – the original 2020 paper that introduced the term.\n- [Retrieval-Augmented Generation for Large Language Models: A Survey](https://arxiv.org/abs/2312.10997) – a map of the research since, including retrieval and evaluation challenges.\n- [A practical guide to RAG](https://www.comet.com/site/blog/retrieval-augmented-generation/) – Comet's technical guide to indexing, chunking, hybrid retrieval and evaluation; its later sections demonstrate Comet's Opik tool." }, { "term": "Readability", "alsoCalled": [], "slug": "readability", "definition": "How easy content is to read and process for its intended audience.", "category": "Editorial craft", "related": [ "accessibility", "content-health-check", "front-loading", "plain-language", "user-testing" ], "updated": "2026-10-11T06:36:00.000Z", "url": "https://contentlexicon.com/readability", "markdown": "https://contentlexicon.com/readability.md", "credit": "Readability, from the Content Lexicon by Contentious (https://contentlexicon.com/readability), CC BY 4.0", "body": "Readability is shaped by sentence length, vocabulary, paragraph structure, heading hierarchy, explanation of unfamiliar terms and the overall amount of effort a reader has to spend to understand what is being said.\n\nReadability scores such as Flesch-Kincaid can be useful signals, but they aren't the whole story. A text can score well and still be confusing, or score badly because it's dealing carefully with a difficult subject for the right audience. Readability is ultimately about fit between the writing and the reader.\n\n## Why it matters for content\n\nReadable content reduces cognitive load. It helps people understand instructions, compare options, make decisions and keep going without unnecessary friction. It's especially important where people are stressed, tired, unfamiliar with the subject, reading on a phone or trying to complete a task rather than browse for pleasure.\n\nProblems blamed on content being \"too long\" are often problems of readability: poor sequencing, heavy noun phrases, unexplained [jargon](https://contentlexicon.com/jargon) or paragraphs that make the reader work too hard. A council tax letter can score well on Flesch-Kincaid and still bury the payment deadline in paragraph four.\n\n## How AI changes the picture\n\nGenerative tools can produce text that is grammatically fluent but effortful to read: padded, abstract, repetitive or vague. Ask a drafting tool to expand a two-line instruction and you may get three paragraphs of smooth, general prose that a readability formula won't complain about, because the sentences are short and the words are common. The reader still has to dig for the instruction.\n\nAI can also help with first-pass simplification or flagging dense passages, but readability still needs editorial judgement. A machine can count syllables; it can't fully judge whether the structure helps the reader do the thing they came to do.\n\n## Questions people ask\n\n**How does readability relate to plain language?**\n\nThey overlap strongly, but readability is about ease of reading and processing, while [plain language](https://contentlexicon.com/plain-language) is a broader commitment to clarity and usability.\n\n**Can a readability score tell me if something is good?**\n\nA score can offer a clue, not a verdict. Content Health Check, for example, treats readability as one of 15 quality criteria, judged alongside things no formula can see, such as whether the content is accurate, up to date and useful.\n\n**Does readable mean short?**\n\nNot necessarily. Long content can be readable if it's well structured and written with care.\n\n**Should all content aim for the same reading level?**\n\nThe right level depends on the audience, subject and stakes. A guide for specialist clinicians can assume more than a leaflet for patients.\n\n## Learn more\n\n- [Legibility, Readability, and Comprehension: Making Users Read Your Words](https://www.nngroup.com/articles/legibility-readability-comprehension/) – Jakob Nielsen's distinction between whether people can see the words, read them, and understand them, with advice for each level.\n- [Jargon and Readability in Plain Language Summaries of Health Research](https://pubmed.ncbi.nlm.nih.gov/39805102/) – an analysis of 1,241 NIHR summaries using readability and jargon measures.\n- [Communicating health research with plain language](https://journals.sagepub.com/doi/10.1177/00469580251357755) – recommendations and examples drawn from researchers' and communication specialists' experience." }, { "term": "Refactoring", "alsoCalled": [], "slug": "refactoring", "definition": "Restructuring something without changing what it does or says, so the next change is easier.", "category": "AI and tooling", "related": [ "content-maintenance", "information-architecture" ], "updated": "2026-10-15T14:34:00.000Z", "url": "https://contentlexicon.com/refactoring", "markdown": "https://contentlexicon.com/refactoring.md", "credit": "Refactoring, from the Content Lexicon by Contentious (https://contentlexicon.com/refactoring), CC BY 4.0", "body": "Refactoring is a change to the internal structure of something that leaves its observable behaviour intact. The term comes from software, where Martin Fowler's book Refactoring (1999) set out the discipline: make one small structural change at a time, check that nothing visible has changed, then make the next.\n\nThe aim is to keep things easy to change. Code that has been patched and extended for years gathers duplication, tangled logic and workarounds that nobody wants to touch. Ward Cunningham called this [technical debt](https://contentlexicon.com/technical-debt): each shortcut makes the next change slower and riskier. Refactoring pays the debt down. Renaming a confusing variable, breaking a long function into smaller ones or removing a duplicated block changes nothing a user can see, but it makes the code easier to understand, test and extend. Fowler's argument is that a clean structure is what lets a team keep adding features at a steady pace, rather than slowing down as the system ages.\n\n## Why it matters for content\n\nRearranging the sections of a single page is ordinary [editing](https://contentlexicon.com/editing). The parallel with software holds better at a larger scale: a [website](https://contentlexicon.com/website), a wider content ecosystem or the system that produces the content. There, the structure is the [information architecture](https://contentlexicon.com/information-architecture), the [content model](https://contentlexicon.com/content-model) and [taxonomy](https://contentlexicon.com/taxonomy), and the templates and [components](https://contentlexicon.com/component) that pages are built from. Over time, sites absorb campaign [microsites](https://contentlexicon.com/microsite), departmental sections and pages written for long-forgotten reorganisations until every change feels risky. Refactoring might merge duplicated pages, regroup content around what people need rather than how the organisation is arranged, or split a catch-all body field into separate fields so the same content can be reused.\n\nBehaviour, in content, is what people are told and how they reach it. That covers meaning, URLs and inbound links, search terms and downstream uses such as reused components and translations. A restructure that keeps every sentence but breaks the link from a printed letter has changed behaviour. So the software rule applies: make one change, such as merging two pages, check that links, redirects and components still work, then make the next. Small steps keep any problem close to its cause, and they let refactoring happen during ordinary maintenance rather than waiting for a relaunch.\n\nThe line between structural and substantive change matters for governance too. If meaning changes, subject experts may need to review it. Even a purely structural move can change who finds a piece of guidance, so set the level of review by what the change could affect.\n\n## How AI changes the picture\n\nAI can cluster a site's pages by topic, find duplicated or near-duplicate content and suggest how a content model might be split or simplified. It can help teams compare structures and carry out routine transformations across many pages.\n\nModels may also alter meaning while claiming to reorganise. When two pages are merged, a shortened clause, a dropped exception or a smoothed qualification can change policy. AI-assisted refactoring needs a comparison against the source and a list of protected facts, such as figures, deadlines and conditions, that must survive unchanged.\n\n[Structured content](https://contentlexicon.com/structured-content) makes refactoring safer. Clear fields, version history and automated checks help teams see what moved and whether links, [metadata](https://contentlexicon.com/metadata) and downstream uses still work.\n\n## Questions people ask\n\n**How is refactoring different from rewriting?**\n\nRefactoring improves structure while preserving meaning. Rewriting may change argument, [tone](https://contentlexicon.com/tone), evidence or purpose. If the subject expert would need to approve the result again, it's probably a rewrite.\n\n**Can a page be split without changing it?**\n\nThe information can be preserved, but splitting changes the reading experience. Redirect the old URL, keep anchor links working where possible and check that warnings and conditions still reach people following either route.\n\n**When should we refactor?**\n\nWhen structure makes maintenance, reuse or understanding harder than it needs to be. Small, regular steps during routine updates are usually safer than saving everything for a relaunch.\n\n**Can AI refactor content safely?**\n\nIt can help, but compare the result with the source, especially where wording carries legal or technical meaning.\n\n## Learn more\n\n- [Refactoring](https://martinfowler.com/books/refactoring.html) – Martin Fowler's book on improving internal structure through small, behaviour-preserving changes, first published in 1999.\n- [The Boy Scout Rule](https://www.informit.com/articles/article.aspx?p=1235624\\&seqNum=6) – the memorable instruction to leave the code cleaner than you found it.\n- [Refactoring Design and Incremental Development](https://www.uxmatters.com/mt/archives/2012/04/refactoring-design-and-incremental-development.php) – Peter Hornsby's UXmatters column applying refactoring principles to design work, from documenting changes to looking for commonality." }, { "term": "Retention and archiving", "alsoCalled": [], "slug": "retention-and-archiving", "definition": "The rules for how long content is kept and how material is preserved, retired from current use or deleted.", "category": "Governance", "related": [ "content-audit", "content-lifecycle", "storytelling", "versioning" ], "updated": "2026-10-10T17:01:00.000Z", "url": "https://contentlexicon.com/retention-and-archiving", "markdown": "https://contentlexicon.com/retention-and-archiving.md", "credit": "Retention and archiving, from the Content Lexicon by Contentious (https://contentlexicon.com/retention-and-archiving), CC BY 4.0", "body": "Retention and archiving cover the end of the [content lifecycle](https://contentlexicon.com/content-lifecycle), a stage often represented in diagrams and neglected in calendars.\n\nUnpublishing takes content off the public site. Archiving preserves material and its context for historical, legal or reference use, with public or restricted access as appropriate. Deleting removes the record, subject to applicable retention duties and legal holds. Retention schedules cover records whether or not they're published. GOV.UK's withdrawal and history mode are examples of marking historical content while keeping it public, rather than definitions every organisation must follow.\n\n## Why it matters for content\n\nKeeping everything live can feel safe, but it has costs. Old content goes on being read and acted on: last year's prices, a returns policy [PDF](https://contentlexicon.com/pdf) from 2019, an appeal for an emergency that closed months ago. It competes with current pages in search, contradicts newer guidance and makes every audit, redesign and migration longer, because each page kept is a page someone should be checking. Where it holds personal details, such as a customer's or volunteer's story, keeping it can also run into privacy law: the GDPR and similar laws elsewhere limit how long personal data can be held.\n\nThe other common pattern is the retention debate that never ends. Without agreed criteria every page has a champion, every retirement becomes a negotiation and the content stays up while the meeting is rescheduled. It ends in the same place as keeping everything, with more meetings.\n\nRemoving content carelessly can have costs too: broken links, a lost public record, a breached retention duty. An old policy announcement may need to remain publicly accessible as a historical record, clearly marked and linked to its replacement. A policy agreed in advance settles most cases before anyone gets attached to a page, and saves the argument for the few that deserve one.\n\n## How AI changes the picture\n\nExternally invisible isn't the same as invisible to machines. An internal assistant with access to the shared drive or intranet will happily quote the archive, and a banner saying “no longer current” helps a reader but may not survive being summarised. Pages taken off the site can also linger in search indexes and in the sources an assistant was given before they came down.\n\nSo retention rules need technical consequences: whether archived material stays eligible for retrieval, how its status shows up in [metadata](https://contentlexicon.com/metadata) and what replaces it as the canonical source. Otherwise yesterday's policy can return as today's answer with excellent [grammar](https://contentlexicon.com/grammar).\n\n## Questions people ask\n\n**When should content be preserved rather than deleted?**\n\nPreserve it when it has historical, legal or reference value, or someone may want to build on it. Depending on its purpose and sensitivity, that record may stay public or move to a restricted archive. Content published in error, consolidated elsewhere or risky to leave up may need unpublishing, ideally with a redirect or an explanation.\n\n**Can archived content come back?**\n\nYes, and it's one of the best reasons to archive rather than delete. A launch page from 2021 can be the starting point for this year's campaign, and last year's research summary can be updated instead of rewritten. That only works if the archive is findable, which is why the context stored with it matters as much as the content.\n\n**Is deletion ever appropriate?**\n\nYes, particularly for personal data that's no longer needed, duplicates and material with no reason to be kept. Check any applicable records duties or legal holds, and make the deletion policy cover copies and backups as well as live content.\n\n**Who decides?**\n\nThe [content owner](https://contentlexicon.com/content-owner) makes the editorial recommendation, within the organisation's records and legal policy. Sensitive cases need the relevant specialists, not a lone [CMS](https://contentlexicon.com/cms) administrator with a tidy streak.\n\n## Learn more\n\n- [Retire outdated content](https://www.gov.uk/guidance/content-design/gov-uk-content-retention-and-withdrawal-archiving-policy) – GOV.UK guidance distinguishing withdrawal, unpublishing and history mode.\n- [Unpublishing, deleting and archiving: how and when](https://insidegovuk.blog.gov.uk/2014/04/09/unpublishing-deleting-and-archiving-how-and-when/) – a 2014 explanation of government publishing as a public historical record, not current instructions for Publisher.\n- [When Online Content Disappears](https://www.pewresearch.org/data-labs/2024/05/17/when-online-content-disappears) – Pew Research Center’s 2024 study finding that 38% of web pages that existed in 2013 were no longer accessible a decade later." }, { "term": "Review date", "alsoCalled": [], "slug": "review-date", "definition": "The point by which a named owner must check that a piece of content remains accurate, useful and fit to stay live.", "category": "Governance", "related": [ "content-decay", "content-lifecycle", "content-maintenance", "content-owner", "metadata" ], "updated": "2026-10-15T08:25:00.000Z", "url": "https://contentlexicon.com/review-date", "markdown": "https://contentlexicon.com/review-date.md", "credit": "Review date, from the Content Lexicon by Contentious (https://contentlexicon.com/review-date), CC BY 4.0", "body": "A review date is a scheduled checkpoint attached to a page, document or asset. When it arrives, the person responsible looks at the content and decides what happens next: confirm it, update it, merge it with something else, archive it or take it down. It's usually stored as [metadata](https://contentlexicon.com/metadata) alongside the owner, publication date and last-updated date, and many [content management systems](https://contentlexicon.com/cms) can send a reminder when it falls due.\n\nIt's different from an expiry date, which takes content offline automatically. A review date leaves the content live and asks for a decision, and systems that support both keep them as separate fields. It's also where [content maintenance](https://contentlexicon.com/content-maintenance) becomes practical: maintenance covers why content needs looking after, while a review date names the person and the day. Without both, the field becomes a box nobody ticks.\n\n## Why it matters for content\n\nContent goes out of date at different speeds. Emergency guidance may need checking within days, fees whenever they change and a background explainer once a year. Review dates let teams match the effort to how likely change is and how much harm an outdated page could do. Some organisations use fixed cycles tied to their calendar, such as reviews before each academic term. Others set a date per item based on risk, or trigger a review when a known event happens, like an annual rates change.\n\nThey also make ownership concrete. \"Finance owns this page\" doesn't tell anyone what to do. \"Aisha reviews it by 15 October, after the new rates are announced\" can go in a calendar and be checked. Overdue dates are one of the quickest ways to find orphaned content, because pages whose owners have left without a handover pile up past their review dates. A date only works if missing it has consequences: a reminder, a visible overdue list and someone with the authority to chase.\n\n## How AI changes the picture\n\nScheduling was never the hard part; most CMSs can send a reminder. AI helps with the part owners put off, which is the review itself. An agent can open an overdue page, compare it with the policy, price list or source document it depends on, and draft a summary of passages that may now be wrong, links that have broken and figures that no longer match. A request to review 40 pages becomes a short list of specific questions.\n\nIt shouldn't make the decision. A model that finds no differences hasn't shown the page is right, only that nothing changed in the sources it was given; a new regulation or a renamed team can make content wrong without touching anything it checked. The owner still signs off, and the review record should show what the agent checked and what the person confirmed. AI also makes review dates more necessary. When content can be produced in bulk, the number of pages needing an owner and a date can grow faster than the number of people available to check them.\n\n## Questions people ask\n\n**How often should content be reviewed?**\n\nIt depends on how fast the subject changes and what goes wrong if the page is out of date. Many teams set a default, such as every six or twelve months, and shorten it for high-risk or high-traffic pages. For content tied to a known change, like a new academic year or a price review, an event-based trigger usually works better than an interval.\n\n**Should every page have a review date?**\n\nEvery managed page needs someone deciding its future, but the interval can vary widely. Pages kept as records, such as news stories or past announcements, may need little more than an occasional check that they're clearly dated and still findable.\n\n**What happens when the date passes?**\n\nIdeally the owner gets a reminder beforehand, the item shows as overdue afterwards and there's a defined route for escalation if nothing happens. Without those, overdue dates measure neglect rather than prevent it.\n\n**Isn't the last-updated date enough?**\n\nThey record different things. A last-updated date changes whenever anyone edits the page, so fixing a typo can make stale content look fresh. A review date records that someone checked the whole thing and decided it should stay. Some sites show readers a \"last reviewed\" date for that reason.\n\n## Learn more\n\n- [Appendix C – Content Review Workflow](https://www.lamar.edu/web-communication/web-governance-policy/procedures-guidelines/procedure-content-review-workflow.html) – Lamar University's web policy, which requires an owner-set review date on every asset in its CMS, with static pages checked before each semester and event content unpublished once its date has passed.\n- [Content Review](https://campus.kennesaw.edu/offices-services/stratcomm/web-team/content-review.php) – Kennesaw State University's three-times-a-year review cycle, with reminders a month ahead, a checklist for page owners and follow-up for departments that don't confirm they've finished.\n- [Why a recurring content health check beats a one-off audit](https://contenthealthcheck.com/learn/beyond-the-snapshot) – Content Health Check on replacing one-off audits with regular checks that build a baseline and show trends, using a quarterly default and tighter cycles for priority pages." }, { "term": "Rollback", "alsoCalled": [], "slug": "rollback", "definition": "Restoring a selected earlier state to recover from a harmful change.", "category": "AI and tooling", "related": [ "cms", "versioning" ], "updated": "2026-10-14T21:14:00.000Z", "url": "https://contentlexicon.com/rollback", "markdown": "https://contentlexicon.com/rollback.md", "credit": "Rollback, from the Content Lexicon by Contentious (https://contentlexicon.com/rollback), CC BY 4.0", "body": "A rollback returns a page, system or set of content to a known earlier version after a change has caused harm. It reverses the change as a whole, so the team can stop the damage first and work out what went wrong afterwards, without leaving people inside the broken release.\n\nSoftware teams treat rollback as a routine move. [Content teams](https://contentlexicon.com/content-team) more often edit forward under pressure, changing a page again and again until nobody can remember what the first version said.\n\n## Why it matters for content\n\nImportant content can fail. A new instruction may omit an exception, a migration may break links or an urgent update may introduce the wrong figure. Restoring a suitable earlier version can stop continuing harm while the team investigates.\n\nThree responses are easily confused. Rolling back restores the last good version of the specific thing that changed, usually from version history. Fixing forward publishes a new correction on top of the faulty change. It suits small, well-understood errors, or cases where going back would cause more harm, such as restoring a price that's no longer valid. Restoring from backup recovers a whole system or [database](https://contentlexicon.com/database) after loss or corruption. It's slower and wider, and can discard legitimate work done since the backup was taken.\n\nA content rollback needs more than a restore button. The [CMS](https://contentlexicon.com/cms) has to keep version history for the item, and the team needs to know who's allowed to restore it. The change may also have touched things outside the page: redirects added when a URL changed, translations updated to match, navigation, [metadata](https://contentlexicon.com/metadata), linked assets and copies sent to other channels. Restoring the English page while the Welsh version and the redirect stay on the new state creates a fresh inconsistency. Check, too, that the old version still suits the current templates, [content model](https://contentlexicon.com/content-model) and service.\n\nTeams should decide which content is critical, how quickly it must be recovered and what evidence is kept for review. A rollback that's been rehearsed in advance is far more reliable than one invented during an incident.\n\n## How AI changes the picture\n\nAI can make large content changes quickly, which raises the value of reversible updates. A [prompt](https://contentlexicon.com/prompt), bulk rewrite or change to a knowledge source may affect many outputs at once.\n\n[Versioning](https://contentlexicon.com/versioning) should cover instructions, datasets and configuration as well as pages. Teams need to know which state produced a problem and which earlier state can be restored safely. Rolling back a prompt doesn't undo answers already generated, so those outputs may need finding and correcting too.\n\nAI can help compare versions and identify likely regressions, but automatic rollback rules need careful testing. Not every drop in a metric means the content is wrong.\n\n## Questions people ask\n\n**Is rollback the same as undo?**\n\nUndo reverses a recent action in an editor. Rollback restores a defined earlier state, often across several connected changes, after something has been published.\n\n**Should we fix forward or roll back?**\n\nRoll back when harm is continuing and the previous state is safer. Fix forward when reversal would create greater risk or the correction is immediate and well understood.\n\n**What needs versioning?**\n\nEverything the recovery depends on. Restoring an instruction may also need the matching content model, asset, redirect or translation, and page history won't always include them. Keep the faulty version too, so the team can investigate.\n\n**How often should rollback be tested?**\n\nOften enough that the team trusts it, and after major platform or [workflow](https://contentlexicon.com/workflow) changes. Google's reliability engineers suggest rolling back \"just because\" if a few weeks have passed without one, to find problems before an incident does.\n\n## Learn more\n\n- [Reliable releases and rollbacks](https://cloud.google.com/blog/products/gcp/reliable-releases-and-rollbacks-cre-life-lessons) – a Google Cloud blog post on the \"rollbacks are normal\" approach: roll back first, investigate second and test rollbacks regularly.\n- [Blue Green Deployment](https://martinfowler.com/bliki/BlueGreenDeployment.html) – Martin Fowler's 2010 explanation of keeping two near-identical production environments so a release can be reversed by switching back.\n- [Rollback to a previous content version](https://docs.umbraco.com/umbraco-cms/manage-and-publish-content/publishing-and-workflow/version-management/rollback-to-a-previous-version) – a concrete CMS example in which restoration creates a reviewable draft rather than erasing history." }, { "term": "ROT content", "alsoCalled": [], "slug": "rot-content", "definition": "Material that is redundant because it does not support the content strategy, out of date or too trivial to justify keeping.", "category": "Governance", "related": [ "content-audit", "content-debt", "content-decay", "content-inventory", "content-maintenance", "decentralised-governance" ], "updated": "2026-10-15T14:41:00.000Z", "url": "https://contentlexicon.com/rot-content", "markdown": "https://contentlexicon.com/rot-content.md", "credit": "ROT content, from the Content Lexicon by Contentious (https://contentlexicon.com/rot-content), CC BY 4.0", "body": "ROT content is whatever a [content audit](https://contentlexicon.com/content-audit) flags as redundant, outdated or trivial. The label marks content for investigation rather than giving permission to delete it: a page tagged ROT might be removed, merged, rewritten, archived or kept because a small audience has an important need for it.\n\nThe acronym comes from records management and information governance, where ROT, sometimes expanded as \"redundant, obsolete, trivial\", describes files and data with no ongoing value. [Content teams](https://contentlexicon.com/content-team) adopted it for web audits. Some add a fourth letter, ROTT, for content that was only ever timely, such as event pages, announcements and campaign material that has done its job.\n\nROT shows up wherever content accumulates without anyone retiring it: old news and press releases, [PDFs](https://contentlexicon.com/pdf) from past campaigns, guidance duplicated by different teams, orphan pages left out of navigation, forgotten [microsites](https://contentlexicon.com/microsite), intranet pages for teams that no longer exist and file libraries nobody has opened in years. In an audit, each item gets a ROT judgement alongside its inventory details. The patterns, such as where most of it sits and who owns it, often say more than any single page.\n\n## Why it matters for content\n\nROT makes good information harder to find and more expensive to maintain. An old contact number buried in a PDF may be more harmful than an entire obsolete microsite, because someone can still find it and trust it.\n\nThe three letters need different evidence. Redundancy is tested against the [content strategy](https://contentlexicon.com/content-strategy) and against other content: does this serve a defined audience need and organisational purpose, and does something else already do the job better? Outdated material needs factual or date checks. Triviality calls for judgement about consequence, significance and legal duties. Low traffic alone doesn't make a page trivial.\n\nROT also makes one of the quickest [business cases](https://contentlexicon.com/business-case) for governance. Every page carries a cost whether or not anyone reads it: it has to be hosted, migrated in the next redesign, checked in audits and searched past by people looking for something else. Gerry McGovern recalls a Microsoft executive telling him that four million of an estimated 15 million pages on Microsoft.com had never been accessed. Showing how much of an estate is ROT, and what it costs to carry, tends to persuade people of the need for owners, [review dates](https://contentlexicon.com/review-date) and retirement rules more readily than any principle.\n\n## How AI changes the picture\n\nROT is no longer only clutter for human visitors. Search systems and [retrieval-augmented generation](https://contentlexicon.com/rag) can surface an old page even when it has disappeared from navigation. A chatbot may then turn stale fragments into a polished answer.\n\nAudits should include the material available to AI systems: archived pages, document stores, duplicated [knowledge bases](https://contentlexicon.com/knowledge-base) and unowned files. Fixing the problem may mean changing retrieval status and canonical sources as well as [editing](https://contentlexicon.com/editing) the public [website](https://contentlexicon.com/website).\n\n## Questions people ask\n\n**Does ROT mean bad writing?**\n\nNo. Beautifully written content can still sit outside the content strategy, describe an expired service or answer a question too inconsequential to justify maintaining.\n\n**Should all ROT content be deleted?**\n\nNo. Merge useful material, redirect duplicate pages and archive records that keep their value for legal or historical reasons. Delete when removal is the right decision for that content's lifecycle.\n\n**How do you identify trivial content?**\n\nUse audience needs, evidence of use, organisational purpose and obligations. A page used by fifty people may matter greatly if all fifty need it to claim a legal right.\n\n**Who fixes ROT?**\n\nThe [content owner](https://contentlexicon.com/content-owner) should decide, with editorial support. Give each correction or retirement a responsible owner, then fix the process that let the problem build up. A review date helps only if someone carries out the review.\n\n## Learn more\n\n- [How To Deal With Redundant, Out-Of-Date And Trivial Content ROT](https://www.smashingmagazine.com/2015/06/dealing-with-redundant-out-of-date-trivial-rot-content/) – a Smashing Magazine guide to identifying and handling the three categories, including why trivial content causes the most disagreement.\n- [Content Plays](https://www.usda.gov/about-usda/policies-and-links/digital/digital-strategy/content/content-plays) – the US Department of Agriculture's content guidance, including a step-by-step ROT analysis that adds a ROT column to the audit spreadsheet and sets out when to delete, archive or revise.\n- [Digital is garbage](https://gerrymcgovern.com/digital-is-garbage/) – Gerry McGovern's polemic on digital waste, from unused data to the millions of pages on [Microsoft.com](http://Microsoft.com) that nobody had ever visited." }, { "term": "Schema", "alsoCalled": [], "slug": "schema", "definition": "A definition of what information something contains and how it is structured.", "category": "Structure and systems", "related": [ "content-model", "database", "machine-readable", "metadata", "structured-content" ], "updated": "2026-10-15T14:55:00.000Z", "url": "https://contentlexicon.com/schema", "markdown": "https://contentlexicon.com/schema.md", "credit": "Schema, from the Content Lexicon by Contentious (https://contentlexicon.com/schema), CC BY 4.0", "body": "A schema lists the fields a kind of information has, what each field holds, which are required and what form their values take. It turns an informal idea, such as \"an event\", into an explicit structure that people and software can share.\n\nAn event schema might include a title, date, location, summary, booking link and organiser, with rules that the date must be a real date and the booking link a working web address.\n\nPeople arrive at the word from different places, and the senses overlap. In a [CMS](https://contentlexicon.com/cms), a schema usually means the definition of a [content type](https://contentlexicon.com/content-type) and its fields. In a [database](https://contentlexicon.com/database), it's the full structure of tables, columns and relationships. Schema.org is a shared vocabulary of types, such as Event, Recipe and Organization, that [websites](https://contentlexicon.com/website) use in their markup so search engines can understand a page. JSON Schema is a standard for describing and validating data in JSON, the format most [APIs](https://contentlexicon.com/api) use. All four describe a shape that content or data must fit.\n\n## Why it matters for content\n\nSchemas support consistency, validation and reuse. When every event uses the same fields, systems can sort them, display them across channels and identify missing information. Authors know what's expected before drafting.\n\nGood schemas model meaning rather than one current page design. A field called \"right-hand box\" ties content to presentation; \"eligibility summary\" describes what the content does and can travel to future layouts.\n\nToo much structure burdens authors and fragments prose. Too little leaves important facts buried in rich text. Content modelling requires judgement about what repeats, what changes and what needs to be reused.\n\nEach schema needs an owner, usually a [content strategist](https://contentlexicon.com/content-strategist) or designer working with a developer, since it encodes both editorial decisions and technical constraints. Changes need managing like code changes. Removing or renaming a field can break templates, integrations and anything that reads the content through an API, and existing content may need migrating to the new shape. [Versioning](https://contentlexicon.com/versioning) the schema, announcing changes and keeping old fields until everything that uses them has moved all reduce the risk.\n\n## How AI changes the picture\n\nA schema specifies the fields and constraints a system expects. AI can be asked to return that shape, while a validator checks structural requirements. Tooling must enforce the constraints; writing them down doesn't guarantee compliance.\n\nAI can suggest schemas from examples, but it may reproduce accidental patterns rather than real requirements. Authors, users and technical teams should test the model against real variation and edge cases.\n\nA valid event record can still have the wrong date. Structural validation and factual checking answer different questions, so keep the source and responsibility for each clear.\n\n## Questions people ask\n\n**Is a schema the same as a content model?**\n\nNot quite. A [content model](https://contentlexicon.com/content-model) describes the content types an organisation needs and how they relate. Schemas express the detailed structure and rules for each type. In many CMSs the two blur, and the terms are often used interchangeably.\n\n**Should every field be required?**\n\nNo. Require only information necessary for meaning, use or governance. Excess requirements encourage filler.\n\n**Can schemas change?**\n\nYes, but changes may affect existing content and connected systems. Version and migrate them deliberately.\n\n**Can AI design the schema?**\n\nIt can propose one from examples. People must decide which distinctions support real needs and future use.\n\n## Learn more\n\n- [Content Modelling: A Master Skill](https://alistapart.com/article/content-modelling-a-master-skill/) – Rachel Lovinger's A List Apart guide to content types, attributes and relationships, including how to decide which information needs a field of its own.\n- [Schema.org](https://schema.org/) – a shared vocabulary of types and properties for describing web information; an example of agreed meaning, not a validator or guarantee of accurate data.\n- [An introduction to structured content](https://digital.gov/resources/an-introduction-to-structured-content) – practical examples and a validator for seeing schema at work on public pages." }, { "term": "Senior content editor", "alsoCalled": [ "Managing editor", "Editorial lead", "Head of editorial" ], "slug": "senior-content-editor", "definition": "A senior editorial role accountable for what gets covered, how it is commissioned and the standard it meets.", "category": "Roles", "related": [ "content-editor", "content-manager", "content-team", "copy-editing", "editing", "head-of-content", "pull-request", "style-guide" ], "updated": "2026-10-09T18:48:00.000Z", "url": "https://contentlexicon.com/senior-content-editor", "markdown": "https://contentlexicon.com/senior-content-editor.md", "credit": "Senior content editor, from the Content Lexicon by Contentious (https://contentlexicon.com/senior-content-editor), CC BY 4.0", "body": "A senior [content editor](https://contentlexicon.com/content-editor) holds editorial authority over a body of content: deciding what it should cover, commissioning and shaping the work, setting the standards it has to meet and making the call when a piece isn't ready. The role often manages [writers](https://contentlexicon.com/writer) and editors, and it usually has the last editorial word before publication, within limits set by legal, subject and organisational responsibilities.\n\nTitles vary. Managing editor, editorial lead and head of editorial can describe much the same job, and smaller organisations often combine it with [content manager](https://contentlexicon.com/content-manager). Where the two are separate, the senior editor's authority centres on editorial direction and quality, while the content manager's centres on the estate and the processes that keep it in order.\n\n## Why an organisation needs a senior content editor\n\nWithout someone holding editorial authority, decisions about quality default to whoever shouts last or ranks highest. A senior editor gives writers and subject experts a single, informed point of challenge: someone who can say that an argument doesn't hold, that two pieces cover the same ground or that the evidence isn't there yet, and whose judgement carries weight because making it is their job. They also develop the people around them, so feedback builds the team's skill rather than turning the editor into a permanent rescue service.\n\nThe need usually becomes pressing when an organisation publishes enough, or with enough at stake, that inconsistency starts to show: a report programme, a magazine or newsletter with a loyal readership, a [knowledge base](https://contentlexicon.com/knowledge-base) people rely on for decisions. The role needs authority to commission, decline and send work back, and time protected from routine correction. A senior editor can't create a strategy single-handedly or make up for briefs that arrive without a purpose; those need fixing further upstream.\n\n## How AI changes the role\n\nEditorial standards used to live in a [style guide](https://contentlexicon.com/style-guide) and in the editor's head. Once assistants draft and check material, those standards also have to exist in a form a system can follow: instructions, approved examples and criteria for judging output. Senior editors are well placed to own that translation, because they know which rules matter and which are habits. Some organisations are formalising it. At Reuters, a senior editor is responsible for regularly reviewing the newsroom's AI guidance, alongside a governance committee chaired by the editor-in-chief.\n\nThe role also takes on decisions that didn't exist a few years ago: which tasks AI may help with, what readers should be told and what never goes out without a named person checking it. A job description might now include setting and reviewing an AI policy for editorial work, judging generated output against the publication's standards and training the team to use assistants without lowering them. Commissioning gains weight too. When competent text is easy to produce, a senior editor's ability to identify what's worth saying, and who has the knowledge or reporting to say it, does most to keep a publication distinctive.\n\n## Questions people ask\n\n**How is a senior content editor different from a copy editor?**\n\nA copy editor works on individual pieces, improving accuracy, structure and consistency so they're ready to publish. A senior editor decides what gets commissioned in the first place, sets the standards copy editors work to and resolves disputes about substance. In a small team one person may do both, but the responsibilities are distinct.\n\n**Should a senior editor still edit?**\n\nYes, though selectively. [Editing](https://contentlexicon.com/editing) keeps their judgement grounded in the work and shows the team what good looks like. If they're rewriting most drafts, the problem usually lies in briefing, capability or [workflow](https://contentlexicon.com/workflow), and fixing that is a better use of their time.\n\n**Who has the final say on what's published?**\n\nWrite down editorial authority along with its limits. A senior editor might decide whether a piece is good enough and right for the publication, while legal, subject and senior organisational colleagues hold vetoes in their own areas. Left vague, the question tends to surface at the worst moment, usually the night before a launch.\n\n**What experience should we look for?**\n\nA record of commissioning and editing other people's work, not only writing their own, along with evidence of developing a team. Journalism, publishing, research communications and senior [content design](https://contentlexicon.com/content-design) all lead into it. Experience of setting standards for AI-assisted work is increasingly useful too.\n\n## Learn more\n\n- [How to Develop AI Guidelines](https://www.cjr.org/feature/how-to-develop-ai-guidelines-newsroom-leaders-policy-experts-rules-journalists.php) – Emily Goligoski's 2026 Columbia Journalism Review feature, in which newsroom leaders from Reuters, the ABC and elsewhere describe writing, testing and revising AI policies; a practical picture of the editorial leadership work AI has added.\n- [Proofreading or editing?](https://www.ciep.uk/resource/proofreading-or-editing-a-quick-guide-to-using-editorial-professionals.html) – CIEP’s clear map of editorial stages and the expertise each one requires, useful when deciding which editorial roles a team needs." }, { "term": "Sentence case", "alsoCalled": [], "slug": "sentence-case", "definition": "Capitalising only the first word and any proper nouns in a heading or label.", "category": "Editorial craft", "related": [ "copy-editing", "style-guide" ], "updated": "2026-10-12T13:51:00.000Z", "url": "https://contentlexicon.com/sentence-case", "markdown": "https://contentlexicon.com/sentence-case.md", "credit": "Sentence case, from the Content Lexicon by Contentious (https://contentlexicon.com/sentence-case), CC BY 4.0", "body": "Sentence case is a capitalisation style for headings and interface text that capitalises the first word and any proper nouns, and leaves everything else in lower case, as a sentence would. \"How to renew your passport\" is in sentence case. \"How to Renew Your Passport\" is in title case.\n\nTitle case comes from publishing tradition, where style manuals set rules for which words take capitals. It looks formal but brings decisions with it: whether \"to\", \"with\" or \"between\" count as minor words, and what to do after a hyphen or colon. Sentence case avoids most of those decisions. It's long been common in British newspaper headlines and is now the default in many digital [style guides](https://contentlexicon.com/style-guide), including Microsoft's and GOV.UK's. Apple's guidelines still use title case for many interface elements, so neither style is universal.\n\nThe same rule can apply to buttons, navigation, labels, form fields and error messages, not only headings.\n\n## Why it matters for content\n\nConsistent capitalisation helps a product or publication feel coherent. Sentence case reduces visual noise and lets wording, hierarchy and design carry emphasis. It's a down-to-earth [accessibility](https://contentlexicon.com/accessibility) practice: familiar capitalisation, without making every word announce its importance. The benefit varies by reader and context, so it belongs alongside clear wording and good heading structure, not in place of them.\n\nIt can also puncture the pomposity of title case. A heading doesn't need a capital on every word to deserve attention. In his article [Making a case for letter case](https://medium.com/@jsaito/making-a-case-for-letter-case-19d09f653c98), Dropbox [writer](https://contentlexicon.com/writer) John Saito explains that the company deliberately uses sentence case to feel “natural and approachable”. That friendliness is a reason to choose it, distinct from any accessibility benefit.\n\nIt also reduces editorial friction. Writers don't need to decide whether short prepositions, conjunctions or infinitives deserve capitals. Translation teams avoid carrying English title conventions into languages that capitalise differently, such as German, where all nouns take a capital, or French, where titles usually don't.\n\nSentence case doesn't mean lowercasing names, brands or abbreviations that properly take capitals. \"Sign in with Google\" keeps its capital G. Sentence case also makes proper nouns more obvious in titles: their capitals stand out against the surrounding lower-case words, rather than blending into the pattern of title case.\n\nA style guide will usually explain how the rule applies to headings, buttons, job titles, product names and the first word after a colon, with examples from the organisation's own content.\n\n## How AI changes the picture\n\nAI can convert text between sentence case and title case and apply a documented convention across large content sets. It can also misread proper nouns, product names and deliberate capitalisation, so automated changes need review.\n\nClear [machine-readable](https://contentlexicon.com/machine-readable) guidance helps. \"Use sentence case for headings and buttons\" works better with examples and an approved list of names. Without that context, models often reproduce the mixed conventions in their training material.\n\nAI-generated interfaces can create headings on the fly. Documented rules and an approved name list improve the odds, but a rule in a [prompt](https://contentlexicon.com/prompt) doesn't guarantee correct capitalisation, so generated headings still need checking.\n\n## Questions people ask\n\n**Is sentence case more accessible?**\n\nSentence case is widely recommended as an accessible content practice, though it isn’t a [WCAG](https://contentlexicon.com/wcag) requirement and the benefit may vary by reader and context. Clear wording, correct heading structure and enough visual distinction matter more. Choose a consistent convention and test the real interface.\n\n**Should buttons use sentence case?**\n\nUsually, if that's the product convention. Write buttons as clear actions and keep proper nouns capitalised.\n\n**What about job titles?**\n\nMany British style guides use lower case for generic job titles and capitals only when the title begins a sentence or forms part of a formal name. Recording the organisation's choice and sticking to it is good practice, though it’s an area where people tend to be especially attached to capital letters.\n\n**Can we change an existing site automatically?**\n\nTools can help, but review names, acronyms and text embedded in images. Capitalisation is easy to change mechanically and surprisingly easy to change wrongly.\n\n## Learn more\n\n- [Making a case for letter case](https://medium.com/@jsaito/making-a-case-for-letter-case-19d09f653c98) – John Saito on Dropbox's choice of sentence case, and how capitalisation affects readability, recognition of proper nouns and the way a brand feels.\n- [Making the case for sentence case](https://www.usa.gov/blog/2023/09/making-the-case-for-sentence-case) – USAGov's account of weighing readability, grammar and consistency before changing its sites.\n- [Capitalization](https://learn.microsoft.com/en-us/style-guide/capitalization) – the Microsoft Style Guide's rules for sentence-style capitalisation, including exceptions for product names and the first word after a colon.\n- [How letter casing affects comprehension of UI content](https://uxdesign.cc/how-letter-casing-affects-comprehension-of-ui-content-34a21235cc73) – an interface-design perspective on casing and comprehension; useful reasoning to weigh against the actual context." }, { "term": "SEO", "alsoCalled": [ "Search engine optimisation" ], "slug": "seo", "definition": "Improving how content is discovered and presented in search engines.", "category": "Structure and systems", "related": [ "content-marketing", "findability", "geo", "metadata" ], "updated": "2026-10-14T12:16:00.000Z", "url": "https://contentlexicon.com/seo", "markdown": "https://contentlexicon.com/seo.md", "credit": "SEO, from the Content Lexicon by Contentious (https://contentlexicon.com/seo), CC BY 4.0", "body": "SEO is the practice of shaping pages, and the sites they sit on, so search engines can crawl them, work out what they're about and rank them for the searches they respond to. It covers technical foundations, page structure, [metadata](https://contentlexicon.com/metadata) and links, but most of it comes down to content that answers the need behind a search. The term dates from around 1997, when early search engines made ranking worth competing for.\n\nSearch engines work in stages. Crawlers follow links to discover pages. Indexing analyses each page's text, images and metadata and stores what it's about. Ranking picks the most relevant, useful results for a query from that index, weighing many signals. The results page then shows a title, a link and a snippet, often drawn from the page's meta description or its own text, and sometimes a featured answer or an AI-generated overview.\n\nAt its best, SEO overlaps with good [content design](https://contentlexicon.com/content-design). At its worst, it produces pages written for imagined algorithms, stuffed with repeated phrases and remarkably reluctant to offer anything useful.\n\n## Why it matters for content\n\nSearch is a major route into [websites](https://contentlexicon.com/website), and much of what helps is editorial: a title that says what the page is about in the words people search with, a meta description that tells people what they'll get, headings that describe each section, internal links with descriptive text and content that meets the intent behind a query, whether that's a quick fact, a set of steps or a comparison.\n\nResearch into search language can reveal how audience segments describe their needs, which may differ from organisational terminology. That evidence should improve content, not merely be inserted as keywords.\n\nSome practices do harm. Keyword stuffing makes text worse for readers and is treated as spam. Producing large numbers of thin pages, by template or by AI, to catch every variation of a query is what Google's spam policies call scaled content abuse, and it can drag down a whole site. Google's Panda update in 2011 was an early example of a search engine acting against content farms.\n\nSEO also includes maintenance: redirects, duplicate handling, structured data and preserving useful routes during migrations. Rankings aren't the only measure. The outcome that matters is whether the right people find accurate content and complete their task.\n\n## How AI changes the picture\n\nAI search and answer systems may summarise content without sending every reader to the page. [GEO](https://contentlexicon.com/geo), or generative engine optimisation, extends the same concerns to those answers, asking whether a source is selected, represented accurately and cited, not just where it ranks. Clear, original and well-structured sources become more important, while generic search filler has less reason to exist.\n\nMaking authority, dates and key claims explicit helps systems interpret the material, though dates and fields don't guarantee that a generated answer will be accurate. Nor is there much to gain from chasing every new acronym with another layer of mechanically optimised prose.\n\nAI can assist with query research, metadata and audits, but it can't manufacture experience, evidence or trustworthiness.\n\n## Questions people ask\n\n**Is SEO mainly about keywords?**\n\nNo. Search terms matter, but usefulness, structure, links, technical [accessibility](https://contentlexicon.com/accessibility) and authority all contribute.\n\n**Should we write for search engines or people?**\n\nWrite for people in language search systems can understand. If optimisation makes the content worse for readers, it has missed the point.\n\n**Does every page need to rank?**\n\nNo. Some pages support existing tasks, transactions or small groups with critical needs rather than broad discovery.\n\n**Has AI made SEO obsolete?**\n\nNo, but discovery is changing. Strong sources, [structured content](https://contentlexicon.com/structured-content) and expertise remain valuable across search and AI answers.\n\n## Learn more\n\n- [In-Depth Guide to How Google Search Works](https://developers.google.com/search/docs/fundamentals/how-search-works) – Google's explanation of crawling, indexing and serving results; one search provider's account, not a ranking formula.\n- [Creating helpful, reliable, people-first content](https://developers.google.com/search/docs/fundamentals/creating-helpful-content) – Google's self-assessment questions about original value, evidence and reader purpose; guidance from one search provider, not a ranking guarantee.\n- [20 Years of SEO: A Brief History of Search Engine Optimization](https://www.searchenginejournal.com/seo/seo-history/) – Search Engine Journal's history of the practice, from the term's first use around 1997 through PageRank to Google's 2011 Panda update against content farms." }, { "term": "Separation of concerns", "alsoCalled": [], "slug": "separation-of-concerns", "definition": "A design principle for dividing a system into parts that each deal with one kind of decision, so each can be understood and changed without disturbing the others.", "category": "AI and tooling", "related": [ "component", "content-model", "create-once-publish-everywhere", "design-system", "graph-engineering", "headless-cms", "single-source-of-truth", "structured-content", "style-guide", "system-prompt" ], "updated": "2026-10-14T22:44:00.000Z", "url": "https://contentlexicon.com/separation-of-concerns", "markdown": "https://contentlexicon.com/separation-of-concerns.md", "credit": "Separation of concerns, from the Content Lexicon by Contentious (https://contentlexicon.com/separation-of-concerns), CC BY 4.0", "body": "In software, a concern is any aspect of a system that has to be got right: what it stores, how it looks, how it behaves, who's allowed to change it. Separating concerns gives each of these its own place, so the code that fetches an event's date needn't know which font will display it, and a change to one part doesn't ripple unpredictably through the rest. On the web, the familiar version is the division of labour between HTML for structure and meaning, CSS for presentation and JavaScript for behaviour.\n\nEdsger Dijkstra coined the phrase in a 1974 essay, where it described a habit of mind as much as a technique: thinking hard about one aspect of a problem at a time while remembering that the others are still there. He conceded that the separation can never be complete. That caveat matters, because separation is a question of where the boundaries sit and which way the dependencies run. It doesn't promise that the parts never touch.\n\n## Why it matters for content\n\nContent people mostly meet the principle as the separation of content from presentation. When the words, facts and structure of a piece are stored apart from the templates that display them, a redesign doesn't mean rewriting, and the same material can appear as a web page, an app screen and a newsletter item. The CSS Zen Garden, launched by Dave Shea in 2003, made the case memorably: hundreds of designers restyled one page without touching its HTML. [Structured content](https://contentlexicon.com/structured-content), [headless CMSs](https://contentlexicon.com/headless-cms) and [create once, publish everywhere](https://contentlexicon.com/create-once-publish-everywhere) all depend on the same split. So does a piece of writing's ability to travel, since copy that leans on its layout (“see the panel on the right”, a heading that only makes sense beside its photo) breaks as soon as it moves.\n\nThe principle reaches well beyond templates. An organisation separates concerns when it keeps editorial rules in a [style guide](https://contentlexicon.com/style-guide) and visual patterns in a [design system](https://contentlexicon.com/design-system), or gives a fact one owner and one source of truth. Each boundary decides who can change what without consulting everyone else.\n\nOverdone, it creates problems of its own: [writers](https://contentlexicon.com/writer) working blind to the space their words will fill, designers drawing boxes for content nobody has written yet, and a [CMS](https://contentlexicon.com/cms) divided so finely that authors can no longer see what they're making. Adam Wathan's essay on utility CSS argued that the textbook web split often moves the coupling between markup and styles rather than removing it. Content and presentation shape each other whatever the architecture says, so good separation makes that dependency deliberate and visible.\n\n## How AI changes the picture\n\n[Large language models](https://contentlexicon.com/llm) undo one of computing's oldest separations. A conventional program keeps its instructions apart from the data it works on, whereas a model receives both as text in a single stream: the [system prompt](https://contentlexicon.com/system-prompt), the user's request, a retrieved policy document and the [email](https://contentlexicon.com/email) it's been asked to summarise. If that email contains a line telling the assistant to disregard its instructions, the model may obey it. Simon Willison named this [prompt](https://contentlexicon.com/prompt) injection in 2022, comparing it to SQL injection, where the fix was to pass the query and the user's input to the [database](https://contentlexicon.com/database) separately.\n\nLanguage models have no equivalent fix yet. Any text an AI system reads, from published pages to supplier documents and customer messages, can therefore carry instructions as well as information, and permissions, validation and approval steps have to supply the separation the model can't guarantee.\n\nSeparation also decides how well AI-assisted content work holds together. When research, drafting and checking are distinct steps with their own instructions and criteria, a poor result can be traced to the step that produced it. Keeping verified source content apart from generated versions means a correction can be made once and flow outwards, instead of being chased through every summary. Even a system prompt benefits: one that mixes [voice](https://contentlexicon.com/voice), policy, product facts and formatting rules in a single block is hard to maintain, while separate sections can each be updated by whoever owns them.\n\n## Questions people ask\n\n**Why do developers keep raising it in content projects?**\n\nContent decisions often end up buried in code: a character limit set by a card [component](https://contentlexicon.com/component), a disclaimer hard-coded into a template, a date format fixed in the [frontend](https://contentlexicon.com/frontend). Developers raise it when they want those decisions moved somewhere editors can see and change, or when a request would tie content and code back together.\n\n**Does it mean writers shouldn't think about design?**\n\nSeparation governs where decisions are stored and changed, rather than who may think about them. Writers still need to understand the contexts their content will appear in, which is why [content models](https://contentlexicon.com/content-model) include limits such as character counts. The difference is that those limits are written into the model instead of discovered by squinting at a preview.\n\n**How does it relate to modularity and structured content?**\n\nModularity means building something from parts; separation of concerns is a reason for drawing their boundaries in particular places. Structured content applies the principle to content itself, dividing a piece into fields that can be managed and reused independently. A model can have plenty of fields and still mix concerns, as when one summary field is expected to carry the search keywords and the legal caveat as well.\n\n**Can AI tools tell instructions apart from content?**\n\nPartly. Providers train models to treat the application's instructions as more authoritative than text found in a document or web page, which reduces the risk of prompt injection without removing it. The safer assumption is that external text may contain instructions, with limits on what an assistant can do without approval.\n\n## Learn more\n\n- [On the role of scientific thought](https://www.cs.utexas.edu/users/EWD/transcriptions/EWD04xx/EWD447.html) – Edsger Dijkstra's 1974 essay that introduced the phrase, presenting it as a discipline of thought and applying it to the gap between specifying what a system should do and building something that does it.\n- [CSS Utility Classes and “Separation of Concerns”](https://adamwathan.me/css-utility-classes-and-separation-of-concerns/) – Adam Wathan's 2017 argument that naming CSS after content doesn't decouple markup from styles, and that the better question is which depends on which. It set out the thinking behind Tailwind CSS.\n- [Prompt injection attacks against GPT-3](https://simonwillison.net/2022/Sep/12/prompt-injection/) – Simon Willison's 2022 post that named the problem, comparing it to SQL injection and wishing for a way to give a model instructions and data separately." }, { "term": "Sign-off", "alsoCalled": [], "slug": "sign-off", "definition": "The formal approval required before content is published, changed or released to its audience.", "category": "Governance", "related": [ "agile", "centralised-governance", "content-editor", "content-owner", "human-in-the-loop", "raci", "workflow" ], "updated": "2026-10-15T15:03:00.000Z", "url": "https://contentlexicon.com/sign-off", "markdown": "https://contentlexicon.com/sign-off.md", "credit": "Sign-off, from the Content Lexicon by Contentious (https://contentlexicon.com/sign-off), CC BY 4.0", "body": "Sign-off is the decision, made by someone with authority to accept the relevant risk, that a piece of work is ready to go out. It usually comes at the end of a defined review process and is recorded, so there's a clear answer later to who approved what and when.\n\nIt's a different job from [editing](https://contentlexicon.com/editing), fact-checking or legal review. Those checks inform the decision, and the person signing off weighs what they found. Treating every reviewer as an approver turns one decision into many, and the final step becomes a general invitation for everyone nearby to improve the adjectives.\n\n## Why it matters for content\n\nSome content needs formal approval because it commits the organisation, explains a regulated service or carries legal or reputational risk. Clear sign-off protects the audience and the people publishing.\n\nProportionate sign-off matches the approval to the risk. A common pattern uses three tiers. Routine changes, such as a corrected link or new opening hours, are approved by the [content owner](https://contentlexicon.com/content-owner) or editor. Substantive changes to guidance need the subject expert who owns the facts. Content that makes legal promises, sets out eligibility or could cause harm if it's wrong also goes to a specialist, such as a lawyer or clinician. Writing the tiers down, with examples, saves arguing each time about who needs to see what.\n\nPoorly designed sign-off protects nobody. When approvals run in series, each one waits for the last, so three approvers can hold a change for a week while each assumes someone else checked the crucial fact. Running specialist reviews in parallel, giving each reviewer a defined question and a deadline, and naming one person who makes the final decision and settles disagreements all shorten the wait without weakening the check. Heavy sign-off processes signal distrust of editors, and are often not needed. Sign-off that goes up through layers of job title hierarchy is another common but poor model.\n\nRoles are clearest when they're written down in one place: who is consulted, who reviews which aspect and who decides. A [RACI](https://contentlexicon.com/raci) chart is one common format.\n\n## How AI changes the picture\n\nMachine-written drafts raise an obvious question: who approves the output? The answer remains a person or role, but sign-off must be proportionate to volume and risk. Requiring a human click on ten thousand routine variants creates the appearance of control without meaningful review.\n\nTeams may use sampling, automated tests and tighter rules for low-risk content, with direct human approval for consequential material. The approver still needs access to sources and a clear account of what the system did. If approval is recorded automatically, make clear which checks ran and which risks the accountable owner has accepted.\n\n## Questions people ask\n\n**Who should sign off content?**\n\nThe person accountable for the relevant decision. That may be a service owner, subject lead, editor or legal specialist, depending on the risk.\n\n**How many approvers do we need?**\n\nUsually one final decision-maker, with specialist reviews added as the risk rises. A routine update may need only the content owner; a changed legal promise may also need a lawyer's review. Keep those checks separate from the final approval and name who resolves a disagreement.\n\n**Is sign-off the same as accountability?**\n\nNo. Sign-off records a decision at a point in time. Accountability includes ownership of the outcome before and after publication.\n\n**Can AI sign off content?**\n\nAI can check rules and recommend a decision. It can't hold organisational accountability. If automated publication is permitted, a named person still owns the system and its consequences.\n\n## Learn more\n\n- [Iterating on the GOV.UK content review process](https://insidegovuk.blog.gov.uk/2016/09/28/iterating-the-gov-uk-content-review-process/) – an account from Inside [GOV.UK](http://GOV.UK) of separating editing from review and clarifying the purpose of each step after delays built up.\n- [Content governance: lightweight practices your team can use](https://www.dol.gov/agencies/eta/ui-modernization/use-plain-language/content-governance) – US Department of Labor guidance on identifying reviewers, subject experts and final approvers in the content lifecycle.\n- [Everything you need to do content strategy](https://contentdesign.london/blog/everything-you-need-to-do-content-strategy) – Content Design London's example of a content strategy that replaces a six-step approval chain with an author working alongside a single expert, and drops the language of sign-off as a sign of distrust." }, { "term": "Silo", "alsoCalled": [], "slug": "silo", "definition": "A team, system or channel that keeps its content to itself, so the same content gets made twice, disagrees with itself and can’t be found.", "category": "Operations", "related": [ "centralised-governance", "content-ecosystem-map", "content-inventory", "content-strategy", "decentralised-governance", "hub-and-spoke" ], "updated": "2026-10-11T06:25:00.000Z", "url": "https://contentlexicon.com/silo", "markdown": "https://contentlexicon.com/silo.md", "credit": "Silo, from the Content Lexicon by Contentious (https://contentlexicon.com/silo), CC BY 4.0", "body": "Comms has press releases, research has [PDF](https://contentlexicon.com/pdf) reports, product has a help centre and none of them knows what the others have published. Each part may work reasonably well alone while the whole content estate becomes repetitive and contradictory.\n\nSilos are almost never created by villainy. Organisations divide work so people can specialise and manage complexity. Trouble starts when a boundary means the help centre team never hears that comms has changed the eligibility rules.\n\n## Why it matters for content\n\nContent crosses departmental lines. A service description may appear on a [website](https://contentlexicon.com/website), in [emails](https://contentlexicon.com/email), scripts, PDFs and chatbot answers. If each channel has its own source and owner, facts drift. The same change is made several times, or not made at all.\n\nSilos also hide the true scale of content. No team can govern what it can't see. A shared spreadsheet of every page and who owns it, an agreed [word list](https://contentlexicon.com/word-list) and one [style guide](https://contentlexicon.com/style-guide) give people enough connection to find related work and make joined-up decisions, without abolishing every useful boundary.\n\nBreaking silos isn't a case for centralising all writing. Teams can keep their expertise and autonomy while agreeing where canonical information lives, who owns it and how changes travel.\n\n## How AI changes the picture\n\nAI assistants can search across collections and make siloed information feel connected. That's useful, but retrieval doesn't resolve disagreement. If three teams maintain different policies, an assistant may select one, blend them or present the contradiction with unwarranted confidence.\n\nSo each policy needs a named owner and one canonical version more than before. AI can help find duplicates, map relationships and flag conflicts, but people must decide which source is authoritative and repair the process that allowed divergence.\n\nSilos also form around AI projects themselves. [Prompts](https://contentlexicon.com/prompt), evaluations and knowledge sources may sit with technical teams while content specialists remain outside. Joined-up work is essential because system behaviour depends on both technology and content.\n\n## Questions people ask\n\n**Are silos always bad?**\n\nNot always. Specialisation is useful, and boundaries help teams focus and protect sensitive information. The problem is a boundary that stops teams coordinating, finding each other's work or knowing who's accountable.\n\n**Does a single CMS remove silos?**\n\nNot by itself. One platform can contain many disconnected practices and duplicate sources. Governance and relationships matter as much as storage.\n\n**Should one team own all content?**\n\nUsually not. A central content function can set standards and join things up, while the teams with subject knowledge stay responsible for the accuracy of their own pages.\n\n**Can AI unify siloed content?**\n\nIt can make information easier to retrieve, but it can't decide which conflicting version is true. That remains an organisational responsibility.\n\n## Learn more\n\n- [Connecting marketing and technical communication](https://review.content-science.com/breaking-down-silos-a-unified-approach-to-effective-content-strategy/) – Tina Kister's argument against treating marketing and technical communication as opposing disciplines.\n- [Content silos: causes, problems and solutions](https://enterprise-knowledge.com/content-silos-causes-problems-and-solutions/) – Sara Nash's account of duplication, findability and lost relationships, with options including metadata hubs and enterprise search.\n- [Breaking down silos in the workplace](https://pmc.ncbi.nlm.nih.gov/articles/PMC11419935/) – a framework drawn from a small CDC staff study, connecting collaboration to leadership support, shared goals and relationship-building." }, { "term": "Single source of truth", "alsoCalled": [], "slug": "single-source-of-truth", "definition": "The one agreed place where a piece of information is maintained, which every other copy or use defers to.", "category": "Governance", "related": [ "atomic-content", "content-debt", "content-governance", "content-owner", "create-once-publish-everywhere", "decentralised-governance", "dry", "merge-conflict", "separation-of-concerns" ], "updated": "2026-10-09T11:29:00.000Z", "url": "https://contentlexicon.com/single-source-of-truth", "markdown": "https://contentlexicon.com/single-source-of-truth.md", "credit": "Single source of truth, from the Content Lexicon by Contentious (https://contentlexicon.com/single-source-of-truth), CC BY 4.0", "body": "A single source of truth is the one place where a particular piece of information is maintained and from which every other use takes its lead. A delivery restriction might appear in a product catalogue, an order confirmation [email](https://contentlexicon.com/email), a help article and a support script, but only one approved record says what the restriction is, and the others draw on it or are checked against it. Copies can exist, as long as everyone knows which record governs and who can change it. \"Truth\" here means agreed authority rather than infallibility, so the source still needs evidence, review and a way to correct it. Nor does it mean one giant repository: prices might live in a product system and policy wording in a document library, each the authority for its own subject.\n\nThe idea comes from data management, where duplication has always been a source of error. E. F. Codd's 1970 paper on the relational [database](https://contentlexicon.com/database) treated redundancy and consistency as problems to design out, and the database normalisation that followed aims to store each fact once so an update can't leave contradictory versions behind. The Pragmatic Programmer later turned it into a general principle for software, [DRY](https://contentlexicon.com/dry), in which every piece of knowledge has a single authoritative representation. Large organisations met the same problem across whole systems, with a customer's details held in several databases at once, and master data management grew up to decide which record wins. [Content teams](https://contentlexicon.com/content-team) got there by their own route. Technical [writers](https://contentlexicon.com/writer) have long single-sourced documentation so one passage can feed several manuals, and in 2009 NPR described storing content once, separate from its display, so it could be published everywhere from its [website](https://contentlexicon.com/website) to iPods and radio displays.\n\n## Why it matters for content\n\nA single source of truth reduces contradiction, rework and confusion. Updates are easier because teams know where to change the thing that others rely on, and governance becomes workable because ownership and review can attach to one source rather than a cloud of near-duplicates. Most organisations already have the duplicates. The trouble starts when nobody knows which one wins: a policy [PDF](https://contentlexicon.com/pdf), a web page and a helpline script disagree, and staff quote whichever they found first.\n\nThe benefit is greatest where a fact is reused or several teams need to keep it consistent, such as eligibility rules, fees, deadlines, opening hours and product specifications. Naming the source is half the work. It needs an owner, a [review date](https://contentlexicon.com/review-date) and a route for people to report errors, and it needs to be easy enough to find that colleagues use it instead of keeping their own versions. A source that's hard to reach soon gets copied into spreadsheets and slide decks.\n\n## How AI changes the picture\n\nAn AI assistant is one more place where information is presented, and it can contradict the source as easily as an out-of-date page can. In 2022 Air Canada's website chatbot told a passenger he could claim a bereavement fare after booking. The airline's own policy page said otherwise, and Air Canada argued that he should have followed the chatbot's link to it. In 2024 the British Columbia Civil Resolution Tribunal disagreed and ordered the airline to pay, saying it was responsible for all the information on its website, whether it came from a static page or a chatbot.\n\nA connected assistant needs a way to select the right source, not merely access to everything. Stable identifiers, status labels and clear superseded versions help it use the approved record rather than a detached copy, provided retrieval and access rules make use of those signals. Even then, one source reduces competing originals but doesn't guarantee every downstream answer is accurate, so test whether each presentation has updated and kept its qualifications. AI can also help find the problem in the first place. A model can compare how a fee or eligibility rule is described across hundreds of pages and flag likely contradictions for someone to check against the source.\n\n## Questions people ask\n\n**Does a single source of truth mean only one copy can exist?**\n\nNo. It means one source is authoritative, even if other places display, summarise or repurpose it. Copies that are generated or synchronised from the source are less risky than copies pasted by hand, because they change when it does.\n\n**How is it different from a system of record?**\n\nIn data management, a system of record is where a particular kind of data is first created and maintained, such as an HR system for employee records. Some data teams reserve \"source of truth\" for a combined view assembled from several systems of record. In content work the terms usually blur, and a single source of truth means the record that governs a particular fact or passage.\n\n**What if two teams each think they own the truth?**\n\nSplit the subject before arguing about the system. The finance team might own the fee, the service team the eligibility rules and the content team the wording that explains both. Each element can then have one authoritative home and one owner, even when they appear together on the same page.\n\n**Can a single source of truth become a bottleneck?**\n\nIt can if ownership is weak, approval is slow or the source is poorly designed. The answer is usually better stewardship, such as delegated [editing](https://contentlexicon.com/editing) rights and clear turnaround times, rather than more competing copies.\n\n## Learn more\n\n- [System of Record vs. Source of Truth: What's the Difference?](https://www.ibm.com/think/topics/system-of-record-vs-source-of-truth) – IBM's explainer on how data teams separate the place information is created from the combined view built from it, and what the word \"single\" adds.\n- [Clean Content = Portable Content](https://www.npr.org/sections/inside/2009/02/clean_content_portable_content.html) – Daniel Jacobson's 2009 post on NPR's create once, publish everywhere approach, with content stored centrally and kept free of display-specific markup.\n- [Airline held liable for its chatbot giving passenger bad advice](https://www.bbc.com/travel/article/20240222-air-canada-chatbot-misinformation-what-travellers-should-know) – BBC Travel's account of the Air Canada ruling, including the airline's argument that its chatbot was responsible for its own actions." }, { "term": "Skill", "alsoCalled": [], "slug": "skill", "definition": "A reusable AI capability or instruction set designed to help an assistant perform a specific kind of task consistently.", "category": "AI and tooling", "related": [ "agentic-ai", "content-brief", "context-engineering", "prompt", "system-prompt" ], "updated": "2026-10-07T09:06:00.000Z", "url": "https://contentlexicon.com/skill", "markdown": "https://contentlexicon.com/skill.md", "credit": "Skill, from the Content Lexicon by Contentious (https://contentlexicon.com/skill), CC BY 4.0", "body": "A skill may contain instructions, examples, rules, templates, scripts, source material, preferred tools or a definition of what a good result looks like.\n\nA skill sits somewhere between a [prompt](https://contentlexicon.com/prompt) and a small piece of software. It can be invoked for a recurring task or selected automatically where the assistant supports that behaviour. It often combines natural-language instructions with reference material or code. The exact packaging varies between AI products, but the purpose is to capture a repeatable way of working so nobody has to reinvent it in a chat box each Tuesday.\n\n## Why it matters for content\n\n[Content teams](https://contentlexicon.com/content-team) do many jobs repeatedly: preparing briefs, checking [voice](https://contentlexicon.com/voice) and [tone](https://contentlexicon.com/tone), analysing research, turning transcripts into notes, applying [accessibility](https://contentlexicon.com/accessibility) guidance and adapting material for different channels. When each job begins with an improvised prompt, quality depends on who remembers the right context and how much patience they have that day.\n\nA well-made skill can hold or reference the organisation’s own method: its standards, source hierarchy, examples, exceptions and review steps. This turns tacit knowledge into something visible and maintainable. It can improve consistency without pretending that every piece of content should be identical, and a person still makes the interesting decisions.\n\nSkills can also expose weaknesses in the underlying process. If nobody can explain the steps, agree the source or recognise a good output, packaging the work will be difficult. That's useful information. Automation is often an unusually efficient way to discover that a supposedly established process lives in the inbox of whoever has been there longest.\n\n## How skills change the picture\n\nSkills make instructions operational. A [style guide](https://contentlexicon.com/style-guide) that once waited patiently for someone to open it can become part of a drafting or review [workflow](https://contentlexicon.com/workflow). A research method can specify which sources to search, how to distinguish evidence from inference and what must be cited. Examples and templates can travel with the task rather than being remembered separately.\n\nThis increases the value of clear, [structured content](https://contentlexicon.com/structured-content). Skills need reliable instructions, stable references and explicit boundaries. They also need maintenance. If policy changes but the skill doesn't, the assistant may apply yesterday’s rule with impressive consistency. Ownership, [versioning](https://contentlexicon.com/versioning) and testing therefore matter just as much as the first draft.\n\nA skill has limits, too. Its instructions can guide tool selection, and bundled scripts can be executed where the assistant has the necessary capabilities and permissions. The skill doesn’t grant that access by itself. It can’t repair missing evidence or make subjective judgement disappear, so it needs clear points for stopping, checking or handing work to a person.\n\n## Questions people ask\n\n**How is a skill different from a long prompt?**\n\nPrompt-like instructions are often part of it, but a skill can also contain examples, reference files, templates and executable tools. More importantly, it's reusable and selected for a class of tasks rather than written for one exchange.\n\n**Does every repeated task need a skill?**\n\nNo. Create one when the task recurs, quality matters and the method is stable enough to describe. A tiny job that changes every time may be better handled directly.\n\n**Can a skill replace a style guide or process document?**\n\nIt should usually draw from them, not become the only copy. Keep a clear canonical source so people can inspect and improve the rules outside the AI system.\n\n**How do we know whether a skill works?**\n\nTest it on representative examples, awkward edge cases and known failures. Compare its outputs with agreed criteria, then record changes rather than tuning it by mood.\n\n## Learn more\n\n- [Agent Skills overview](https://platform.claude.com/docs/en/agents-and-tools/agent-skills/overview) – the primary description of modular instructions, metadata and resources.\n- [Building effective AI agents](https://resources.anthropic.com/hubfs/Building%20Effective%20AI%20Agents-%20Architecture%20Patterns%20and%20Implementation%20Frameworks.pdf) – Anthropic’s architecture guide, including how modular skills extend single-agent and multi-agent systems; its customer results are vendor-reported.\n- [What are agent skills?](https://www.sysdig.com/learn-cloud-native/what-are-agent-skills) – a coding-agent introduction covering skill files, testing and permission risks, framed around Sysdig’s cloud-security product." }, { "term": "Slow content", "alsoCalled": [], "slug": "slow-content", "definition": "Making digital content with care: quality over quantity, depth over breadth, the citizen over the consumer.", "category": "Foundations", "related": [ "agile", "ai-slop", "attention-economy", "content-marketing", "content-principles", "evergreen-content", "longform", "website" ], "updated": "2026-10-11T06:25:00.000Z", "url": "https://contentlexicon.com/slow-content", "markdown": "https://contentlexicon.com/slow-content.md", "credit": "Slow content, from the Content Lexicon by Contentious (https://contentlexicon.com/slow-content), CC BY 4.0", "body": "Slow doesn't mean long, solemn or artisanal, and it covers maintaining content as much as making it. Sometimes it produces a short, clear answer; sometimes an essay, report, story or guide that takes the time it needs.\n\nThe idea borrows from the Slow Food movement and from web designers such as Jeffrey Zeldman, who have argued for work that gives readers time to understand rather than hurrying them along. The term itself was set out in the Slow Content manifesto.\n\n## Why it matters for content\n\nSlow content matters because so much digital content is created under pressure to publish more, faster and everywhere. That pressure rewards repetition, thinness, reactive thinking and formats chosen because they're fashionable rather than useful. The result is often a web full of material that exists only because someone had a Tuesday slot in the [content calendar](https://contentlexicon.com/content-calendar) to fill.\n\nFor [content teams](https://contentlexicon.com/content-team), slow content is a corrective. It asks whether the work is meaningful, well made, necessary and worth someone’s attention. It supports [content strategy](https://contentlexicon.com/content-strategy) because it treats content as a cultural and civic act, not only as a marketing instrument. It also supports content health: fewer, better, better-maintained pieces often serve people more effectively than sprawling estates of neglected pages.\n\n## How AI changes the picture\n\nAI makes the slow content argument more urgent. Generative tools can produce passable text at speed, which is useful when the task is bounded, source-led and properly reviewed. But they can also make it spectacularly easy to create more mediocre content than any human audience asked for. Volume was already a weak substitute for value.\n\nUsed well, AI can support slow content. It can help organise source material, test clarity, summarise evidence, check consistency, suggest structures and make maintenance easier. But it should serve the values of the work, not replace them.\n\n## Questions people ask\n\n**Does slow content mean long content?**\n\nSlow content isn't the same as [longform](https://contentlexicon.com/longform). It values depth, care and purpose, but concise content can be beautifully slow in spirit. A 50-word answer that helps someone may be more aligned with slow content than a 2,000-word article stretched like cheap chewing gum.\n\n**Is slow content anti-marketing?**\n\nNo, but it's suspicious of content that treats people only as targets, leads or consumers. Slow content can support business goals. It insists that those goals shouldn't reduce the reader to a conversion rate.\n\n**Can organisations afford to publish less?**\n\nOften, yes. What many can't afford is to keep publishing material that nobody maintains, trusts or needs. Slow content isn't an argument for idleness. It's an argument for better choices.\n\n**How does slow content relate to content strategy?**\n\nIt gives content strategy a set of values. Strategy decides what to make, for whom, why and how it will be governed. Slow content asks whether those decisions produce work that is useful, meaningful and worth putting into the world.\n\n## Learn more\n\n- [Slow Design for an Anxious World by Jeffrey Zeldman](https://adactio.com/journal/14882) – Jeremy Keith’s notes on Zeldman’s talk arguing that web design should slow readers down so they understand more.\n- [The Slow Web](https://jackcheng.com/the-slow-web/) – Jack Cheng’s 2012 essay on timely not real-time, rhythm not random and moderation not excess, with a 2016 note that user-experience practices alone can’t fix the fast web.\n- [Slow Content](https://slowcontent.org/) – the Slow Content manifesto, with its ten pillars." }, { "term": "Social media", "alsoCalled": [], "slug": "social-media", "definition": "Digital platforms where individuals and organisations publish, respond, distribute content and take part in public or semi-public conversations.", "category": "Operations", "related": [ "attention-economy", "channel-strategy", "content", "content-marketing" ], "updated": "2026-10-07T13:53:00.000Z", "url": "https://contentlexicon.com/social-media", "markdown": "https://contentlexicon.com/social-media.md", "credit": "Social media, from the Content Lexicon by Contentious (https://contentlexicon.com/social-media), CC BY 4.0", "body": "For organisations, social media can support awareness, relationships, customer service, campaigning, recruitment and the distribution of useful content.\n\nIt is not one channel. LinkedIn, Instagram, TikTok, Reddit and private community spaces have different cultures, formats and expectations. A post that technically fits everywhere may belong nowhere. Copying the same caption into six platforms is distribution, but perhaps not strategy.\n\n## Why it matters for content\n\nSocial media can bring content to people where they already spend time. It offers immediate signals – replies, questions, shares and criticism – that can reveal what an audience values or misunderstands. It also lets organisations show work in progress, respond in public and build familiarity over time rather than appearing only when they have something to sell.\n\nThe work needs a clear relationship with owned content. Social posts are often temporary, difficult to retrieve and subject to platform decisions. Important guidance, arguments and evidence should usually have a stable home on a [website](https://contentlexicon.com/website) or [knowledge base](https://contentlexicon.com/knowledge-base), with social media used to interpret, introduce and discuss them.\n\nGood social content also needs operations behind it: ownership, response expectations, moderation, [accessibility](https://contentlexicon.com/accessibility), approval routes and a sustainable rhythm. Chasing every trend is an excellent way to create a very busy calendar and a strangely indistinct body of work.\n\n## How AI changes the picture\n\nAI can help turn source material into platform-specific drafts, suggest variations and analyse recurring audience questions. Used well, it reduces repetitive adaptation and helps a team explore formats. Used badly, it produces a river of polished sameness that sounds like every organisation has employed the same enthusiastic intern.\n\nThe answer is strong source material and human selection. AI-generated posts should be grounded in actual ideas, products, evidence and language from the organisation. Each channel still needs judgement about context, timing and what the audience can reasonably do next.\n\nOrganisations could also use AI to support more attentive community care. An assistant could help a moderator find unanswered questions, retrieve relevant source material and notice where one public explanation would help several people. That would give the moderator more time for the exchange itself. It's a possible design, not a capability every platform provides, and public responses should remain accountable to people. Simulated community members or automated replies pretending to be human would undermine the relationship.\n\nFor people using social media, AI is also part of the experience: it shapes platform feeds, moderation and recommendation. Organisations do not fully control who sees a post or how it is interpreted. This makes measurement more nuanced: reach and engagement can be useful signals, but they do not prove understanding, trust or action. It also strengthens the case for keeping valuable content in an owned, findable form.\n\n## Questions people ask\n\n**Does every organisation need to be on every platform?**\n\nNo. Choose channels where a relevant audience is present and where the organisation can contribute consistently. Absence is better than a neglected account or a stream of obligation.\n\n**Should social posts always link to the website?**\n\nNo. Some posts should be complete and useful where they appear. Link when the website provides meaningful depth, a task or a stable source, not merely to improve a dashboard.\n\n**How often should we post?**\n\nOften enough to maintain a useful presence, but only at a pace the team can sustain without lowering quality. A thoughtful weekly contribution can do more than daily filler.\n\n**Can AI run our social media?**\n\nIt can assist with drafting, adaptation and analysis. People should still choose what is worth saying, check facts and [tone](https://contentlexicon.com/tone), respond to others and handle sensitive situations.\n\n## Learn more\n\n- [The Garden and the Stream: A Technopastoral](https://hapgood.us/2015/10/17/the-garden-and-the-stream-a-technopastoral/) – Mike Caulfield contrasts immediate, conversational streams with interconnected knowledge that can develop over time.\n- [Tiktok’s enshittification](https://pluralistic.net/2023/01/21/potemkin-ai/) – Cory Doctorow argues that platform incentives can redirect value away from users and creators, sharpening the question of what an organisation can depend on.\n- [Social media accessibility](https://www.section508.gov/create/social-media/) – practical accessibility guidance for images, captions and platform-specific publishing, written for US federal communicators." }, { "term": "Staging", "alsoCalled": [], "slug": "staging", "definition": "A production-like environment for reviewing and testing changes before release.", "category": "AI and tooling", "related": [ "cms", "frontend" ], "updated": "2026-10-15T15:06:00.000Z", "url": "https://contentlexicon.com/staging", "markdown": "https://contentlexicon.com/staging.md", "credit": "Staging, from the Content Lexicon by Contentious (https://contentlexicon.com/staging), CC BY 4.0", "body": "Staging is a copy of the live system, set up to match it as closely as possible, where changes are checked before release. The live system is called production. Staging sits just before it: close enough to production that passing a test there tells you something, but hidden from public view.\n\nMost teams move work through a ladder of environments. Local is a person's own computer. Development, sometimes called integration, is a shared space where everyone's changes are combined and tested together. Staging mirrors production, often with a copy of its data, settings and connected services. Production is what the public uses. Smaller teams may skip a rung, and some [CMSs](https://contentlexicon.com/cms) offer previews that cover part of what staging does.\n\n## Why it matters for content\n\nContent changes affect more than words. Staging is where [content teams](https://contentlexicon.com/content-team) can see how a change behaves in the full system before readers meet it.\n\nRendering is the first check. A heading may wrap badly on a phone, a table may overflow or an image may lose its [alt text](https://contentlexicon.com/alt-text) in a new template. Integrations come next: a form may send the wrong confirmation, a pricing feed may not pull through or a reused [component](https://contentlexicon.com/component) may show the old wording. After a restructure, redirects need testing so old URLs and bookmarks land in the right place. Search matters too. A new page should appear in site search with the right title and summary, and filters and tags should still work.\n\nStaging is most valuable when review follows realistic routes. Arrive from a search result or a direct link, try different devices and ask someone who didn't make the change to use it. Testing only from the homepage misses the paths most people take.\n\nStaging must be protected from public discovery and handled carefully if it holds personal or sensitive data. It also has to keep pace with production. A staging site that's months out of date gives false comfort.\n\n## How AI changes the picture\n\nAI features need staging too. Teams need to test instructions, retrieval, source links, refusals, corrections and escalation in an environment that resembles real use, with the same knowledge sources the live system will draw on.\n\nSynthetic test cases can explore varied requests and edge cases, while human testing reveals expectations and trust. Log generated outputs safely and assess them against explicit criteria.\n\nStaging can't prove that every AI response will work, but it helps teams find predictable failures before the public does.\n\n## Questions people ask\n\n**Is preview the same as staging?**\n\nThey can overlap. A preview may show one draft or a whole proposed experience. Staging is an environment for testing connected behaviour before release. Check what each one reproduces rather than relying on the label.\n\n**Should staging use real data?**\n\nUse representative, protected data. A full copy of production data makes tests realistic but brings the same privacy obligations as the live system.\n\n**Who should test content in staging?**\n\nSomeone other than the author, plus relevant specialists for [accessibility](https://contentlexicon.com/accessibility), subject accuracy and technical behaviour.\n\n**Can staging replace user research?**\n\nNo. It supports testing but doesn't tell you what real users need or how they behave.\n\n## Learn more\n\n- [GOV.UK's environments (integration, staging, production)](https://docs.publishing.service.gov.uk/manual/environments.html) – GOV.UK's developer documentation on what each environment is for, including why staging is kept off the public internet.\n- [Traditional Development/Integration/Staging/Production Practice for Software Development](https://dltj.org/article/software-development-practice/index.html) – Peter Murray's outline of the four-tier ladder, including why staging should match production as closely as possible.\n- [Rethinking pixel-perfect web design](https://www.smashingmagazine.com/2026/01/rethinking-pixel-perfect-web-design) – a reminder that real content, real data and responsive states are less tidy than static mock-ups." }, { "term": "Stakeholder mapping", "alsoCalled": [], "slug": "stakeholder-mapping", "definition": "Identifying the people and groups who affect, approve, use or care about a piece of content work.", "category": "Research and evidence", "related": [ "content-governance", "content-owner", "raci", "workflow" ], "updated": "2026-10-15T15:13:00.000Z", "url": "https://contentlexicon.com/stakeholder-mapping", "markdown": "https://contentlexicon.com/stakeholder-mapping.md", "credit": "Stakeholder mapping, from the Content Lexicon by Contentious (https://contentlexicon.com/stakeholder-mapping), CC BY 4.0", "body": "A stakeholder map records who the stakeholders are, what matters to them and how they relate to the decisions ahead. It may use a simple list, a visual network or familiar categories such as levels of influence, motivations and interest.\n\nFor content work, use the map to understand whose knowledge is needed, where authority sits, who may be overlooked and how to involve people at a useful moment rather than presenting them with a nearly finished answer.\n\n## Why it matters for content\n\nContent crosses organisational boundaries. A service page may involve policy experts, operational teams, legal advisers, support staff, publishers and the people who use the service. Each group sees different risks and needs. Mapping helps a team bring those perspectives together without allowing the loudest meeting attendee to stand in for the whole system.\n\nIt also clarifies decisions. Someone who supplies evidence isn't necessarily the person who approves publication. Someone with formal authority may not understand the daily consequences. Recording roles, concerns and relationships can prevent sprawling review rounds in which everyone comments but nobody decides.\n\nGood stakeholder mapping includes people affected by the content, not only senior colleagues who can block it. Users, frontline staff, partners and underrepresented groups may have less organisational power but vital knowledge. Influence is useful to understand; it isn't a measure of whose experience matters.\n\n## How AI changes the picture\n\nAI-assisted content introduces additional stakeholders: people who own source data, govern models, manage security, review risk and handle escalation when the system fails. A feature that appears to be a writing tool may affect customer support, legal responsibility, [accessibility](https://contentlexicon.com/accessibility) and information governance.\n\nAI can help organise interview notes, identify recurring concerns and suggest missing roles. It may also reproduce the assumptions in the material it receives. A generated list can [prompt](https://contentlexicon.com/prompt) investigation, but it isn't evidence that everyone relevant has been represented. Check omissions with people who know the work, including those affected outside the project team.\n\nMapping should also consider people affected indirectly by automated content decisions. They may never use the interface themselves, yet still experience its consequences. Human oversight needs named owners with enough authority and knowledge to intervene.\n\n## Questions people ask\n\n**Who counts as a stakeholder?**\n\nAnyone who influences the work, contributes necessary knowledge, uses the result or experiences its effects. The relevant group depends on the decision you're making.\n\n**Is a power–interest grid enough?**\n\nIt can be a useful start, but it simplifies changing relationships. Add needs, knowledge, influence, likely concerns and the form of involvement required.\n\n**Should we map individual people or groups?**\n\nUse groups where their interests and roles are shared. Name individuals when a particular person holds knowledge, authority or responsibility that matters.\n\n**How often should the map be updated?**\n\nWhenever the work, organisation or decision changes substantially. A stakeholder map is a working aid, not a ceremonial photograph of the kick-off meeting.\n\n## Learn more\n\n- [A guide to stakeholder mapping and engagement](https://www.icaew.com/technical/corporate-governance/new-boardroom-agenda/guide-to-stakeholder-mapping-and-engagement) – prioritisation, limitations and keeping a map alive.\n- [Mendelow power-interest matrix](https://mutomorro.com/tools/mendelow-power-interest-matrix) – a worked example of the familiar four-quadrant tool.\n- [ISO stakeholder engagement guidance](https://www.iso.org/files/live/sites/isoorg/files/store/en/PUB100269.pdf) – Inclusive, ongoing engagement in standards development; useful questions to adapt, not content-project requirements." }, { "term": "Storytelling", "alsoCalled": [], "slug": "storytelling", "definition": "Conveying ideas or experience through an account of particular people and events unfolding over time, shaped so that an audience can follow, imagine and care about what happens.", "category": "Editorial craft", "related": [ "ai-slop", "content-marketing", "copywriting", "longform", "message-architecture", "retention-and-archiving", "voice" ], "updated": "2026-10-16T18:50:00.000Z", "url": "https://contentlexicon.com/storytelling", "markdown": "https://contentlexicon.com/storytelling.md", "credit": "Storytelling, from the Content Lexicon by Contentious (https://contentlexicon.com/storytelling), CC BY 4.0", "body": "A story follows someone through a sequence of events in which something is at stake and something changes. Storytelling is the craft of choosing and arranging that material: where to begin, which details to keep, when to slow down and what to leave the audience to work out. A list of things that happened in order isn't yet a story. It becomes one when each event seems to cause or complicate the next, and the reader starts wanting to know what happens.\n\nThe word covers a lot of ground in organisations. It can mean a fundraising appeal built around one person, a customer case study, a founder's account of why a company exists or a report that carries its data through one family's experience. It's also used loosely for key messages and positioning, which can support a story without being one. Narrative is a close relative: it often refers to the larger account an organisation or movement tells about itself, while a story is one particular telling within it.\n\n## Why it matters for content\n\nStories carry what abstractions struggle to. A report might say that a new referral process cut waiting times by six weeks; a story that follows one person through those six weeks lets readers understand what the number meant. Particulars help people remember, imagine circumstances unlike their own and see how one thing led to another. Psychologists describe the absorption a good story produces as transportation, and transported readers tend to argue back less. That makes storytelling persuasive, and it's also why a single vivid case can mislead when it stands in for a pattern it doesn't represent. Pairing a story with evidence about how typical it is lets each do its own work.\n\nMany of the stories organisations tell belong to someone else. A case study or an appeal depends on a person agreeing to share part of their life, and that agreement covers particular uses rather than every future format. The choices made in the telling matter too: whether the person is the protagonist or a backdrop for the organisation's good work, which moments of difficulty are shown and whether they'd recognise themselves in the result. Stories also age. Circumstances change, people withdraw consent and an account that was accurate three years ago can misrepresent someone now, so a story collection needs owners and [review dates](https://contentlexicon.com/review-date) like any other content.\n\n## How AI changes the picture\n\nIn seconds, a model can write a moving account of someone who never existed. For fiction and clearly labelled illustration, that's a creative tool. For content that presents stories as testimony, such as case studies, appeals, reviews and patient experiences, an invented person is a fabrication however plausible the details. Much of a told story's value rests on its having happened to someone who agreed to tell it, and neither the first-hand experience nor the permission can be generated.\n\nWith stories that did happen, AI tools can do a great deal. They can read hundreds of interview transcripts, feedback forms or support conversations and surface recurring experiences and striking moments a team would struggle to spot, with references back to the source. They can propose several structures for the same material, point out where a draft starts too early or retell an account from another point of view so a [writer](https://contentlexicon.com/writer) can compare. They can adapt a long story into a newsletter excerpt, a script for a short video or an [audio](https://contentlexicon.com/audio) version. Each adaptation needs checking against the original and against what the teller agreed to, because compression is where meaning tends to slip. Left to their defaults, models drift towards familiar arcs, tidy redemption and an uplifting final line, so they work best as collaborators whose suggestions get questioned.\n\nWhen an AI assistant summarises a page, the story is usually the first thing to go: the summary keeps the claim and drops the person. Stating the point plainly alongside a story helps it survive summary, and the story becomes one of the reasons to visit the page rather than read about it. Stories can also become something readers explore. The USC Shoah Foundation's Dimensions in Testimony lets visitors put questions to recorded Holocaust survivors and matches each one to an answer the survivor gave. Its strength is that it retrieves testimony rather than inventing it, which is the standard any conversational version of a true story needs to meet.\n\n## Questions people ask\n\n**What's the difference between a story and a narrative?**\n\nThey overlap, and people often use them interchangeably. Where a distinction is made, a story is a particular account of events, while a narrative is the broader way events are framed and connected: the account a campaign or organisation keeps returning to. Many stories can serve one narrative, and a narrative can persist without any single story being told in full.\n\n**Does all content need a story?**\n\nPlenty of content works better without one. Someone renewing a licence or checking opening times wants the answer, and wrapping it in narrative slows them down. Stories earn their place where understanding, empathy or memory matters: explaining why a service exists, showing what a change means for people or making an argument stick. A sentence or two of story is sometimes enough, without a full arc.\n\n**Can we use AI to write case studies?**\n\nIt can help draft one from a real interview, with the subject checking the result before publication. Inventing customers or beneficiaries is a different matter. A composite or illustrative character needs labelling as such, and even then readers may feel misled if the format looks like testimony.\n\n**Who should be the hero of an organisation's story?**\n\nUsually not the organisation. Marketing frameworks such as StoryBrand cast the customer as hero and the brand as guide, and many charities have moved away from narratives in which donors rescue passive recipients. A more revealing test is whose experience the story is about, and whether that person would recognise how they've been portrayed.\n\n## Learn more\n\n- [Ira Glass on Storytelling](https://www.thisamericanlife.org/extras/ira-glass-on-storytelling) – the This American Life host's much-shared Current TV interview on building a story from anecdotes and moments of reflection, and on the long gap between a beginner's taste and their ability.\n- [Who Owns the Story? Ethical Approaches to Humanitarian Narratives](https://blogs.lse.ac.uk/internationaldevelopment/2025/03/03/who-owns-the-story-ethical-approaches-to-humanitarian-narratives/) – Poppy Bullen, a content producer at the International Rescue Committee, on consent, power and expiry dates for stories in content libraries, and on moving from client-centred to client-led storytelling.\n- [Let's ditch the dangerous idea that life is a story](https://aeon.co/essays/let-s-ditch-the-dangerous-idea-that-life-is-a-story) – the philosopher Galen Strawson argues against the widely held view that everyone understands their life as a narrative, a useful counterweight to the assumption that story is how all people make sense of things." }, { "term": "Structured content", "alsoCalled": [], "slug": "structured-content", "definition": "Content broken into meaningful parts so it can be managed, reused and presented in different ways.", "category": "Structure and systems", "related": [ "atomic-content", "backend", "content", "content-model", "content-type", "create-once-publish-everywhere", "domain-model", "headless-cms", "machine-readable", "paddle-swim-dive", "pdf", "separation-of-concerns", "system-one-models", "ubiquitous-language", "website" ], "updated": "2026-10-06T20:57:00.000Z", "url": "https://contentlexicon.com/structured-content", "markdown": "https://contentlexicon.com/structured-content.md", "credit": "Structured content, from the Content Lexicon by Contentious (https://contentlexicon.com/structured-content), CC BY 4.0", "body": "The parts may be fields such as title, summary, audience, steps, quote, date or outcome, or other smaller [components](https://contentlexicon.com/component) that can be managed and reused separately.\n\nNone of this makes content formulaic; it makes the structure explicit. On an event page where the date, venue and booking link are separate fields, the same three facts can feed the newsletter, the listings page and the calendar invite, and a change of venue gets made once.\n\n## Why it matters for content\n\nStructured content makes it easier to publish the same underlying content in more than one place, change presentation without rewriting the source and manage content at scale without everything dissolving into copy and paste.\n\nIt matters most where content appears across several channels, where systems need to display it differently in different contexts, or where teams want to move beyond page-by-page publishing towards a more deliberate content architecture.\n\n## How AI changes the picture\n\nMachines work much better with explicit structure than with structure implied in prose, so structured content is far easier for AI to use well. Clear fields, types and relationships make it easier to draft, retrieve, transform, classify and check content.\n\nWeak structure gets expensive. If the content is all one blob, the system has to guess which part is the key message, which part is an action, which part is evidence and which part is [metadata](https://contentlexicon.com/metadata). Those guesses are often wrong, and the errors travel into every summary and answer built on them.\n\nStructure can also keep an account honest as it travels. A research finding linked to its method, dataset and qualifications could be explained at different depths, while always preserving a route back to the evidence. Breaking content into parts should preserve the relationships between them, not make it easier for an assistant to detach the striking claim from its conditions. That is a possible reading experience built on structure, not a promise that every model will interpret it correctly.\n\n## Questions people ask\n\n**How does structured content relate to a content model?**\n\nA [content model](https://contentlexicon.com/content-model) defines the structure; structured content is the content created within it.\n\n**Does structured content make writing more rigid?**\n\nIt can if used clumsily, but done well it provides useful constraints without flattening meaning.\n\n**Is this only for big technical systems?**\n\nNo. The benefits are most obvious there, but any team that reuses content or needs consistency can benefit from more explicit structure.\n\n**Can structured content still sound human?**\n\nYes. Structure shapes how content is organised and reused, not whether it has a [voice](https://contentlexicon.com/voice).\n\n## Learn more\n\n- [Structured Content: Your Key to Improving Customer Journey, Business Value, and Content Operations](https://review.content-science.com/structured-content-your-key-to-improving-customer-journey-business-value-and-content-operations/) – Carrie Hane’s definition, with examples such as a museum whose opening hours, stored as fields, stay consistent on its website and Google Maps.\n- [Designing Connected Content](https://www.pearson.com/en-us/subject-catalog/p/designing-connected-content-plan-and-model-digital-products-for-today-and-tomorrow/P200000009416/9780134763385) – Carrie Hane and Mike Atherton take the reader from understanding a subject domain to modelling and publishing connected content.\n- [Content maturity is AI readiness](https://contentmaturity.com/learn/content-maturity-is-ai-readiness) – why documented standards and structured, governed content are foundations for useful AI." }, { "term": "Style guide", "alsoCalled": [], "slug": "style-guide", "definition": "A shared record of how an organisation writes: voice, tone, style rules and the practical guidance people need to make consistent choices.", "category": "Governance", "related": [ "accessibility", "centralised-governance", "content-governance", "content-policy", "content-principles", "content-standards", "decentralised-governance", "editing", "linting", "machine-readable", "separation-of-concerns", "tone", "voice", "word-list" ], "updated": "2026-10-06T15:45:00.000Z", "url": "https://contentlexicon.com/style-guide", "markdown": "https://contentlexicon.com/style-guide.md", "credit": "Style guide, from the Content Lexicon by Contentious (https://contentlexicon.com/style-guide), CC BY 4.0", "body": "A style guide brings together decisions about [voice](https://contentlexicon.com/voice), [tone](https://contentlexicon.com/tone), [punctuation](https://contentlexicon.com/punctuation), [grammar](https://contentlexicon.com/grammar), terminology, capitalisation, formatting, structure and other recurring editorial choices. A good style guide isn't a cabinet of fussy preferences. It's working guidance that helps people make consistent decisions without reopening the same debate every week.\n\nSome style guides are short and tightly practical. Others are broader, covering [content principles](https://contentlexicon.com/content-principles), audience guidance and examples of how the organisation sounds in different situations. What matters is that the guidance is usable.\n\n## Why it matters for content\n\nA style guide saves time and makes an organisation sound like one organisation. [Writers](https://contentlexicon.com/writer), editors and subject experts can settle whether it's \"[email](https://contentlexicon.com/email)\" or \"e-mail\" in seconds instead of starting a thread about it, and readers stop meeting a slightly different house style on every page. New contributors get up to speed faster too, because the rules are written down rather than learned by being corrected.\n\nThe need grows with the number of people writing, the number of channels and the wish to sound recognisable without sounding mechanical. Repetition gives it away: if an editor is making the same correction for the third time this month, that correction belongs in the guide.\n\n## How AI changes the picture\n\nAI turns the style guide from a reference document into a system input. A human writer absorbs style from examples, edits and the odd raised eyebrow; a model only knows the rules it's given. A 2019 [PDF](https://contentlexicon.com/pdf) on a shared drive, or conventions nobody has written down, won't reach it at all.\n\nWhere the guide is silent, a person would ask. A model fills the gap with whatever is most common in its training data: American spellings, title-case headings, Oxford commas. Editors then spend their time correcting the same defaults in every draft, which is the work a style guide is supposed to save them. That makes the guide part of the quality infrastructure rather than a nice-to-have editorial extra.\n\n## Questions people ask\n\n**Is a style guide just about punctuation and capital letters?**\n\nThose are part of it, but a useful style guide also covers voice, tone, terminology and the practical choices that affect how content feels and functions.\n\n**How long should a style guide be?**\n\nAs long as it needs to be and no longer. The test isn’t length; it’s whether people can use it.\n\n**Do small organisations need one?**\n\nUsually, though a lighter one. If the same issues keep being corrected, the organisation already needs shared guidance.\n\n**Can AI use a style guide properly?**\n\nOften yes, if the guide is explicit and accessible enough. It works far less well when the guidance is vague or mostly cultural.\n\n## Learn more\n\n- [Mailchimp Content Style Guide](https://styleguide.mailchimp.com/) – the influential public example in full, from goals to mechanics and channels.\n- [A to Z style guide](https://guidance.publishing.service.gov.uk/writing-to-gov-uk-standards/style-guides/a-to-z-style-guide/) – the UK government's house style for its websites in one alphabetical list, from abbreviations to words to avoid.\n- [What is a content style guide?](https://voicetoneandstyle.com/learn/what-is-a-content-style-guide) – a modern definition of a living guide for voice, tone, conventions, people and AI." }, { "term": "Subject matter expert", "alsoCalled": [], "slug": "subject-matter-expert", "definition": "A person with deep topic knowledge who helps make content accurate and credible.", "category": "Governance", "related": [ "centralised-governance", "content-brief", "content-designer", "content-owner", "pair-writing", "writer" ], "updated": "2026-10-15T15:17:00.000Z", "url": "https://contentlexicon.com/subject-matter-expert", "markdown": "https://contentlexicon.com/subject-matter-expert.md", "credit": "Subject matter expert, from the Content Lexicon by Contentious (https://contentlexicon.com/subject-matter-expert), CC BY 4.0", "body": "A subject matter expert, often shortened to SME, is the person whose knowledge a piece of content depends on: the clinician behind a health page, the policy lead behind eligibility guidance, the engineer behind a technical guide. It's a part people play alongside their main job rather than a job title. The responsibility usually covers supplying knowledge, checking substance and saying when something has changed.\n\nExpertise in a subject doesn't automatically include expertise in explaining it. The strongest work pairs subject authority with content craft (through [pair writing](https://contentlexicon.com/pair-writing), for example) rather than expecting one person to provide both.\n\n## Why it matters for content\n\nSubject experts understand exceptions, evidence and consequences that [writers](https://contentlexicon.com/writer) may miss. They can identify where simplification becomes inaccurate and where an apparently small wording change alters policy or technical meaning.\n\nCollaboration works best when responsibilities are clear. The expert provides subject authority, while the content specialist shapes structure, language and the experience around it. Final approval rights still need to be assigned for the task; expertise doesn't automatically make someone the publisher or sole approver. Pair writing can resolve questions faster than sending drafts through serial review.\n\nThe responsibility also needs time that's recognised in the expert's actual job. Reviews squeezed between other work tend to arrive late, or not at all, and a page waiting on one busy person is a page going out of date. Focused questions, clear deadlines and visible decisions produce better input than asking someone to “check everything”.\n\n## How AI changes the picture\n\nWith appropriate consent and data handling, AI can capture interviews, organise notes and turn expert sources into a first draft, which makes the expert's contribution quicker to gather. It can also fabricate plausible detail, so expert verification matters more for consequential material.\n\nThe responsibility is also moving upstream. When AI agents answer questions from an organisation's own sources, an expert's knowledge reaches readers through those sources as well as through reviewed pages. Keeping them accurate, writing test cases that show whether an agent handles difficult questions correctly and defining when it should hand over to a person all become part of the expert's contribution. Knowledge kept only in one person's head or inbox can't do that work.\n\n## Questions people ask\n\n**Should the expert write the first draft?**\n\nNot necessarily. A writer can draft from an interview or source material while the expert checks substance.\n\n**Who decides when clarity and precision conflict?**\n\nA subject matter expert and an editor should find language that preserves both. If a technical term is necessary, explain it rather than removing it.\n\n**How many experts should review?**\n\nFor a routine technical update, one expert in that subject may be sufficient; a change crossing safety and legal requirements may need both specialisms. Name what each reviewer checks and who resolves disagreement. The number follows the scope, not a quota.\n\n**Can AI become the subject expert?**\n\nNo. It can represent supplied knowledge, but it doesn't hold responsibility or guaranteed current expertise.\n\n## Learn more\n\n- [Working well with subject matter experts](https://defradigital.blog.gov.uk/2022/08/24/working-well-with-subject-matter-experts-tips-for-content-designers/) – six practical, humane tips from Defra content designers on trust, roles and working through resistance.\n- [Working effectively with subject matter experts](https://www.digital.govt.nz/standards-and-guidance/design-and-ux/content-design-guidance/research-plan-maintain/working-effectively-with-subject-matter-experts) – a useful guide to balancing technical authority with audience-centred content practice.\n- [Collaboration between designers and subject experts](https://stel.pubpub.org/pub/03-02-gottler-2023) – A 2023 study of instructional designers and subject experts in online learning; relational working in that setting, not a finding about every digital project." }, { "term": "System prompt", "alsoCalled": [], "slug": "system-prompt", "definition": "The standing instructions an AI system receives before the user's request, shaping how it behaves and responds.", "category": "AI and tooling", "related": [ "agentic-ai", "context-engineering", "context-window", "machine-readable", "prompt", "separation-of-concerns", "skill", "style-guide" ], "updated": "2026-10-06T21:15:00.000Z", "url": "https://contentlexicon.com/system-prompt", "markdown": "https://contentlexicon.com/system-prompt.md", "credit": "System prompt, from the Content Lexicon by Contentious (https://contentlexicon.com/system-prompt), CC BY 4.0", "body": "A system [prompt](https://contentlexicon.com/prompt) sets the initial rules for how the system should behave: what role it is playing, what constraints it should follow, what [tone](https://contentlexicon.com/tone) or style it should use, how it should approach tasks and sometimes which tools or sources it should prefer. Users may never see it, but it can strongly shape the output.\n\nIt's part brief, part policy and part behavioural setup. It isn't the whole system, but it's usually the first layer of instruction.\n\n## Why it matters for content\n\nSystem prompts are one of the main places where editorial guidance becomes system behaviour. If an organisation wants an assistant to sound a certain way, respect certain standards, follow specific rules or prefer certain sources, some of that logic often begins here. A line as small as \"use British spelling and never promise delivery dates\" can influence answers throughout the assistant's work. A vague system prompt produces vague or inconsistent output; a clear one makes expectations repeatable.\n\nStanding instructions carry more weight when an assistant is used every day, by many people or for tasks where mistakes cost something. Asking each user to remember the [style guide](https://contentlexicon.com/style-guide) before typing a question won't work, so the system prompt has to carry more of the load.\n\n## How system prompts change the picture\n\nA standing rule can shape thousands of exchanges without anyone reopening the style guide. That gives editorial decisions a reach they didn't have as reference documents.\n\nThat can make behaviour more consistent, but vague, outdated or overconfident instructions get baked in too, and they're invisible to the people reading the answers. A line written last spring naming \"our current flagship programme\" will keep being repeated long after the programme has closed, and the person reading the answer may not be able to see where the claim came from. A system prompt can't carry an entire [content strategy](https://contentlexicon.com/content-strategy) on its back. It needs an owner, a [review date](https://contentlexicon.com/review-date) and good source material behind it.\n\n## Questions people ask\n\n**How is a system prompt different from a user's prompt?**\n\nThe user's prompt is the immediate request. The system prompt supplies standing instructions, usually chosen by the application rather than the person typing. Providers use different names and roles for this layer, and the application has to supply or maintain it where the [API](https://contentlexicon.com/api) requires.\n\n**Can users see the system prompt?**\n\nSometimes, but often not. Many tools keep it hidden or only partly visible.\n\n**Is a better system prompt enough on its own?**\n\nUsually not. It can improve behaviour, but output also depends on what the model can see: the documents it retrieves, the examples it's given and the review that happens afterwards.\n\n**Do we need one system prompt for everything?**\n\nNot necessarily. A fundraising assistant and an internal policy helper need different standing instructions, but both should share the same underlying [voice](https://contentlexicon.com/voice) and standards.\n\n## Learn more\n\n- [Effective context engineering for AI agents](https://www.anthropic.com/engineering/effective-context-engineering-for-ai-agents) – Anthropic's 2025 account of the larger informational system around the prompt, including how specific a system prompt should be.\n- [Claude prompting guidance](https://platform.claude.com/docs/en/build-with-claude/prompt-engineering/claude-prompting-best-practices) – Anthropic's guidance on writing instructions, including setting a role in the system prompt.\n- [OpenAI prompt engineering guide](https://developers.openai.com/api/docs/guides/prompt-engineering) – another provider’s practical patterns for instructions and structured outputs." }, { "term": "Systems thinking", "alsoCalled": [], "slug": "systems-thinking", "definition": "Looking at how parts connect and influence one another, rather than treating each content problem in isolation.", "category": "Foundations", "related": [ "content-ecosystem-map", "content-operating-model", "workflow" ], "updated": "2026-10-06T15:28:00.000Z", "url": "https://contentlexicon.com/systems-thinking", "markdown": "https://contentlexicon.com/systems-thinking.md", "credit": "Systems thinking, from the Content Lexicon by Contentious (https://contentlexicon.com/systems-thinking), CC BY 4.0", "body": "A system is made from connected parts whose behaviour influences one another over time: people, rules, tools, incentives, information and the environment around them.\n\nIn content work, systems thinking helps explain why a weak page may be more than a writing problem. It may be the visible result of unclear ownership, a rushed [workflow](https://contentlexicon.com/workflow), an unsuitable [content model](https://contentlexicon.com/content-model), conflicting policies or a publishing tool that makes the right thing difficult. Rewriting the page may help today without changing what produces tomorrow’s version.\n\n## Why it matters for content\n\nContent rarely succeeds or fails on prose alone. Quality emerges from the whole arrangement around it: what the organisation is trying to achieve, what users need, who can make decisions, how evidence is gathered, which standards apply and what happens after publication.\n\nSystems thinking helps teams notice patterns instead of collecting isolated irritations. Repeated duplication may point to weak source management. Long approval times may reflect unclear authority rather than slow reviewers. A growing content estate may be caused by incentives to publish and no agreed route to retire anything. These connections suggest more useful interventions than another editorial tidy-up.\n\nIt also encourages humility: changing one part can produce effects elsewhere, sometimes after a delay. A stricter approval process might improve risk control but make updates so slow that information becomes inaccurate. You can't predict everything. You can look at what a change touches before you make it.\n\n## How AI changes the picture\n\nAI increases the number and speed of interactions within content systems. It can generate material, classify pages, retrieve sources and act on instructions. A weakness in one place can therefore travel quickly: outdated guidance may shape thousands of answers, while a strong canonical source can improve many experiences at once.\n\nThis makes relationships more important. Teams need to understand which sources feed which tools, where [prompts](https://contentlexicon.com/prompt) and rules live, who reviews outputs and how corrections return to the system. Treating an AI feature as a separate technical layer can conceal its dependence on content, governance and human judgement.\n\nAI can help map and analyse systems too, by finding repeated themes, dependencies and gaps across large collections. But its map is still an interpretation, not the territory, and people still need to decide where the boundary sits, whose experience counts and which consequences matter.\n\n## Questions people ask\n\n**Do we need to map the whole organisation?**\n\nNo. Draw a boundary around the question you need to answer. Include enough surrounding context to understand the pattern without attempting a magnificent diagram of everything.\n\n**Is systems thinking just looking at the bigger picture?**\n\nThat's part of it, but systems thinking is more specific. It examines relationships, feedback, delays, boundaries and the patterns these produce over time.\n\n**Does systems thinking make simple fixes impossible?**\n\nA small fix can still be sensible. The difference is knowing whether it treats a symptom, changes a cause or creates another problem elsewhere.\n\n**Who should be involved?**\n\nInclude people who experience different parts of the system, especially those who create, approve, maintain and use the content. No single viewpoint can reveal every connection.\n\n## Learn more\n\n- [Leverage Points](https://donellameadows.org/archives/leverage-points-places-to-intervene-in-a-system/) – Donella Meadows’s classic on where interventions in complex systems have more or less power.\n- [Thinking in Systems](https://www.chelseagreen.com/product/thinking-in-systems/) – Meadows’s accessible primer on feedback, stocks, flows and behaviour over time, from its publisher.\n- [Content and pace layers](https://contentious.ltd/thoughts/content-and-pace-layers) – a systems-minded look at the different speeds at which organisational content and infrastructure change." }, { "term": "Taxonomy", "alsoCalled": [], "slug": "taxonomy", "definition": "A controlled set of categories or labels, usually arranged from broad to narrow, used to classify content consistently.", "category": "Structure and systems", "related": [ "content-model", "findability", "information-architecture", "knowledge-base", "knowledge-graph", "metadata", "system-one-models" ], "updated": "2026-10-06T20:57:00.000Z", "url": "https://contentlexicon.com/taxonomy", "markdown": "https://contentlexicon.com/taxonomy.md", "credit": "Taxonomy, from the Content Lexicon by Contentious (https://contentlexicon.com/taxonomy), CC BY 4.0", "body": "A taxonomy helps people browse, filter, group and connect information, and it helps organisations apply the same labels in the same way over time. It is more deliberate than a casual tag list: the terms are chosen, defined and maintained rather than left to drift.\n\nIn content work, taxonomies often sit behind topics, service categories, audience segments, [content types](https://contentlexicon.com/content-type), product areas or other ways of grouping material. They are part of [information architecture](https://contentlexicon.com/information-architecture), but they also shape [metadata](https://contentlexicon.com/metadata), search, reporting and reuse.\n\n## Why it matters for content\n\nA good taxonomy improves [findability](https://contentlexicon.com/findability) and consistency. It helps readers discover related material, helps teams organise large content sets and makes it easier to filter, measure and govern what exists. It becomes essential once content has outgrown a set of pages one person could list from memory.\n\nA taxonomy earns its keep where content needs to be browsed, filtered, searched, analysed or reused across channels. It's also what stops several teams labelling the same material three ways: without shared categories, one team files things under \"Grants\", another under \"Funding\" and a third under \"Money for projects\".\n\n## How AI changes the picture\n\nAI makes taxonomy more operational. A taxonomy no longer just sits behind a navigation or filter menu; it can shape retrieval, classification, recommendation and the context an assistant uses to answer questions. If categories overlap, an assistant asked about funding for small charities may pull guidance filed under three different labels and blend it, or miss the one page filed under the label nobody searches for.\n\nModels can propose labels or classify content, but they can only apply the distinctions the taxonomy draws. If nobody on the team can explain the difference between \"Advice\" and \"Guidance\", a classifier won't manage it either.\n\n## Questions people ask\n\n**Is a taxonomy the same as navigation?**\n\nNo. Navigation is one visible way a taxonomy may appear, but a taxonomy can also sit behind filters, metadata, internal labels and search.\n\n**How does a taxonomy relate to metadata?**\n\nThey work together. Metadata is the set of fields that describe content; a taxonomy supplies the agreed values some of those fields use. The metadata field says \"topic\"; the taxonomy says which topics are allowed and how they nest.\n\n**Do users have to see the taxonomy?**\n\nNot always. Some taxonomies are visible to readers and some are mainly internal, but both affect how content is organised and found.\n\n**Who should own a taxonomy?**\n\nUsually someone close to [content design](https://contentlexicon.com/content-design), information architecture or governance, with input from the people who use and maintain the content. What matters is that someone owns the decisions and the upkeep.\n\n## Learn more\n\n- [Taxonomy 101](https://www.nngroup.com/articles/taxonomy-101/) – its central framing: a taxonomy is a backstage controlled vocabulary that complements visible navigation, and building one starts small.\n- [Information Architecture: Study Guide](https://www.nngroup.com/articles/ia-study-guide/) – broader context for where taxonomy sits within IA work.\n- [Taking a thematic approach](https://contentious.ltd/thoughts/taking-a-thematic-approach) – an accessible argument for organising content around meaningful themes rather than organisational furniture." }, { "term": "Technical debt", "alsoCalled": [], "slug": "technical-debt", "definition": "The future cost or risk created by shortcuts, structural weaknesses or deferred improvements.", "category": "Operations", "related": [ "cms", "content-audit", "content-costs", "content-debt", "content-decay", "content-maintenance", "content-model" ], "updated": "2026-10-15T15:38:00.000Z", "url": "https://contentlexicon.com/technical-debt", "markdown": "https://contentlexicon.com/technical-debt.md", "credit": "Technical debt, from the Content Lexicon by Contentious (https://contentlexicon.com/technical-debt), CC BY 4.0", "body": "Technical debt is the extra work a system demands later because of shortcuts, ageing structures or improvements put off today. Like financial debt, it can be a sensible choice, as long as someone knows it exists and pays it down. Left alone, it charges interest: every change takes longer and carries more risk.\n\nWard Cunningham coined the metaphor in 1992, writing that \"shipping first time code is like going into debt\", fine if repaid promptly with a rewrite, dangerous if not. He meant it as an argument for regular rework, not an excuse for sloppiness.\n\nIn software the debt takes familiar forms: a quick fix that became permanent, a library several versions out of date, tests nobody wrote, code only one person understands. None of these stops the system working today. Each makes the next change slower and riskier, and the bill tends to arrive at the worst moment, when something urgent needs doing. Martin Fowler adds a useful distinction between debt taken on deliberately, with eyes open, and debt nobody noticed until it was already expensive.\n\n## Why it matters for content\n\nContent lives in systems, and those systems carry technical debt like any other software. A [CMS](https://contentlexicon.com/cms) customised so heavily that nobody dares upgrade it, a [content model](https://contentlexicon.com/content-model) designed for one [website](https://contentlexicon.com/website) that now has to feed an app and a chatbot, templates that fix layout so firmly that changing a heading needs a developer, an integration held together by a script one person wrote years ago: each slows publishing and makes change expensive. Editors usually feel it first. They learn the workarounds and, after a while, stop mentioning them.\n\nThis is different from [content debt](https://contentlexicon.com/content-debt), which sits in the content itself: outdated pages, duplicated facts, unclear ownership. The two grow together. A rigid template encourages copying and pasting, and a content model with no field for a [branch](https://contentlexicon.com/branch)'s opening hours ends up with them typed into every branch page. Fixing one rarely clears the other. A new CMS won't rescue content nobody looks after, and well-kept content still suffers in a platform nobody can change.\n\nPlatform work is easy to postpone because the system still works, more or less. Recording technical debt in the same register as content debt, with its consequences spelled out, makes the case visible. The GOV.UK team behind the publishing platform for government departments did this openly in 2013, slowing new features for a while to pay down debt in the code behind editorial [workflow](https://contentlexicon.com/workflow) and attachments.\n\n## How AI changes the picture\n\nAI coding assistants make it quicker to write code, including code nobody on the team fully understands, which is one of the oldest sources of technical debt. They can also help pay it down. They explain unfamiliar legacy code, suggest refactors, write missing tests and flag outdated dependencies, with a developer reviewing what changes.\n\nAI features bring debt of their own. [Prompts](https://contentlexicon.com/prompt) hard-coded into an application, retrieval pipelines built on a content model that was never designed for them and workflows nobody has retested since the underlying model changed are all shortcuts that charge interest later. For [content teams](https://contentlexicon.com/content-team), a common version is an assistant connected to a CMS whose structure can't tell it which version of a page is current. These belong in the debt register too, with owners and review points.\n\n## Questions people ask\n\n**Is technical debt always a bad thing?**\n\nNo. Debt taken on knowingly can be good judgement. Launching on a simple setup to learn what people need, then rebuilding, is often wiser than designing the perfect platform up front. Problems start when the debt goes unrecorded and the rebuild never happens.\n\n**Is technical debt the same as content debt?**\n\nNo. Technical debt sits in platforms, models and code; content debt sits in pages, files and the facts they contain. They feed each other, so it helps to track both in one place.\n\n**Will a new CMS clear our debt?**\n\nA migration can clear a lot of technical debt. It moves content debt across intact unless someone audits and prunes the content first, which makes a migration one of the best moments to delete what nobody needs.\n\n**Who should own technical debt in a content platform?**\n\nWhoever owns the platform, usually a product or digital team, should own the register. Editors and [content designers](https://contentlexicon.com/content-designer) should be able to add to it, because they often see the interest being paid before developers do.\n\n## Learn more\n\n- [The WyCash Portfolio Management System](https://c2.com/doc/oopsla92.html) – Ward Cunningham's 1992 experience report, the short text where the debt metaphor first appears, framed as a case for regular rework.\n- [Technical Debt](https://martinfowler.com/bliki/TechnicalDebt.html) – Martin Fowler's 2019 account of future change costs and the limits of the financial metaphor, including why estimates and gradual improvement matter.\n- [Paying down (technical) debt in the departments and policy publishing platform](https://insidegovuk.blog.gov.uk/2013/12/10/paying-down-technical-debt-in-the-departments-and-policy-publishing-platform/) – Tekin Suleyman's 2013 account of the GOV.UK team slowing new features to repay debt built up while moving more than 100 government organisations onto one publishing platform.\n- [Technical debt in content operations](https://www.scriptorium.com/2024/07/technical-debt-in-content-operations/) – Sarah O'Keefe on recognising and managing the infrastructure version of the problem." }, { "term": "Tone", "alsoCalled": [], "slug": "tone", "definition": "How an organisation's voice adjusts to suit the audience, context and moment.", "category": "Editorial craft", "related": [ "conversational-design", "microcopy", "persona", "style-guide", "voice" ], "updated": "2026-10-11T06:17:00.000Z", "url": "https://contentlexicon.com/tone", "markdown": "https://contentlexicon.com/tone.md", "credit": "Tone, from the Content Lexicon by Contentious (https://contentlexicon.com/tone), CC BY 4.0", "body": "A welcome [email](https://contentlexicon.com/email), a service apology, a policy update and a campaign page may all come from the same organisation while sounding different in the ways they should.\n\nTone isn't decoration. It affects how content lands, how trustworthy it feels and whether it meets the reader in the right spirit.\n\n## Why it matters for content\n\nThe same facts can work or fail on tone alone. A charity might thank a new donor with a line that grins (\"You've made our week\") and, a month later, report a data breach in short, plain sentences: what happened, what it means for you, what to do now. Swap the two and both fail, however accurate the facts. (Nobody wants an exclamation mark in a breach notice.)\n\nIt matters most where stakes or emotions are high: service failure, sensitive topics, difficult instructions, moments of uncertainty, fundraising asks and public responses. It also matters wherever several teams are creating content, because tone drifts quickly if nobody has defined how it should flex.\n\n## How AI changes the picture\n\nModels are good at producing plausible-sounding language, but without examples and constraints they default to bland helpfulness or borrowed internet styles, like the chirpy \"Great question!\" in reply to a bereavement enquiry. That flattens an organisation's [voice](https://contentlexicon.com/voice) and creates jarring mismatches between message and moment.\n\nIf teams want assistants to sound right in different contexts, they need tone guidance a machine can act on: the same message written in two or three registers, notes on when each applies, and someone reviewing what comes out. A generic instruction to sound warm or professional gets you warm or professional in every situation, including the ones that call for neither.\n\nTone can also flex within a single exchange. When someone asks for the practical next step, an assistant can become more direct; when they ask for the reasoning, it can offer a fuller explanation. The adjustment responds to what the person says and to the circumstances of the service, not to a claimed insight into how anyone feels. Clear language, accurate sources and an appropriate route to human support still matter.\n\n## Questions people ask\n\n**Can one organisation have several tones?**\n\nIt usually should, as long as they still feel recognisably part of the same voice.\n\n**Is tone the same as voice?**\n\nNo. Voice is the underlying personality. Tone is how that personality adapts.\n\n**Is tone just about sounding friendly?**\n\nFriendly is one setting among many. The right tone may be warm, firm, calm, celebratory, direct or restrained depending on the context.\n\n**Can AI match tone reliably?**\n\nSometimes, but it works much better when the organisation has explicit guidance and examples rather than broad adjectives alone.\n\n## Learn more\n\n- [Monzo tone of voice](https://monzo.com/tone-of-voice) – Monzo's guidance in use, including how tone changes for difficult news.\n- [Voice and Tone](https://styleguide.mailchimp.com/voice-and-tone) – Mailchimp's distinction between a stable voice and a tone that changes with circumstances.\n- [The Four Dimensions of Tone of Voice](https://www.nngroup.com/articles/tone-of-voice-dimensions/) – Kate Moran's Nielsen Norman Group framework describing tone on four scales: funny or serious, formal or casual, respectful or irreverent, enthusiastic or matter-of-fact." }, { "term": "Typography", "alsoCalled": [], "slug": "typography", "definition": "The design of text on a page: typeface, size, spacing, hierarchy and the other choices that shape how readable content feels.", "category": "Editorial craft", "related": [ "accessibility", "design-system", "frontend", "readability", "sentence-case", "wireframe" ], "updated": "2026-10-06T20:57:00.000Z", "url": "https://contentlexicon.com/typography", "markdown": "https://contentlexicon.com/typography.md", "credit": "Typography, from the Content Lexicon by Contentious (https://contentlexicon.com/typography), CC BY 4.0", "body": "Typography isn't a finish applied once the writing is done; it changes how quickly people can read, understand and trust what's in front of them.\n\nFor content work, typography matters because [readability](https://contentlexicon.com/readability) is partly written and partly designed. Clear sentences can still be hard work if the text is cramped. Good typography lets the [writer](https://contentlexicon.com/writer)'s structure show.\n\n## Why it matters for content\n\nTypography signals hierarchy, separates ideas and tells the reader where to look first. Line length is a good example of how much the details matter: a common guideline is roughly 45 to 75 characters a line, because much longer lines make the eye lose its place on the way back. It also affects [accessibility](https://contentlexicon.com/accessibility). Pale grey 12px text on a white benefits page may have looked elegant in the brand deck; for someone with low vision it's close to unreadable.\n\nContent-heavy pages feel it most: articles, guidance, service pages, [longform](https://contentlexicon.com/longform) landing pages, reports and [knowledge bases](https://contentlexicon.com/knowledge-base). So do moments of stress or complexity, where a reader is trying to complete a task, compare options or understand difficult information.\n\n## How AI changes the picture\n\nMore content is now drafted, templated and published through systems, often without a designer looking at the finished page. That puts the weight on defaults. A template with feeble heading styles or a pale body colour applies them to every page that passes through it, and nobody notices until a reader complains.\n\nThe same shift strengthens the case for separating structure from presentation. When one piece of content appears in an app, on a web page and in a chat window, typography has to work from the structure underneath (real headings, real lists) rather than from a manual layout pass at the end. Content that's only bold text pretending to be a heading loses its hierarchy the moment it leaves the page it was made on.\n\n## Questions people ask\n\n**How much does typography affect readability?**\n\nMore than most people expect. [WCAG](https://contentlexicon.com/wcag)'s text spacing criterion asks that content still works when readers increase line height to 1.5 times the font size and paragraph spacing to twice the font size, because many people need that room to read comfortably. Content that's technically clear can still feel difficult if the typography is doing it no favours.\n\n**Is typography just choosing a font?**\n\nTypeface choice is part of it, but typography also includes spacing, hierarchy, rhythm, contrast and how text behaves across devices and screen sizes.\n\n**Isn't this the designer's job rather than the writer's?**\n\nIt's mainly a design responsibility, but content people need enough understanding to notice when presentation is undermining meaning. Structure and wording alone can't fix poor text design.\n\n**Is typography part of accessibility?**\n\nYes. Accessible content depends on whether the words can be comfortably perceived and navigated, not only on which words were chosen.\n\n## Learn more\n\n- [Typography](https://web.dev/learn/accessibility/typography) – web.dev guidance on readable, accessible text on the web.\n- [Designing for Web Accessibility](https://www.w3.org/WAI/tips/designing/) – W3C guidance on accessible visual design, including text.\n- [WCAG text spacing](https://www.w3.org/WAI/WCAG21/Understanding/text-spacing) – W3C's Success Criterion 1.4.12 on keeping content readable and operable when users override line, paragraph, letter and word spacing." }, { "term": "Ubiquitous language", "alsoCalled": [], "slug": "ubiquitous-language", "definition": "A shared, precise vocabulary used across conversations, documentation and systems within a defined area of work.", "category": "Structure and systems", "related": [ "content-model", "content-operating-model", "context-engineering", "domain-model", "structured-content" ], "updated": "2026-10-20T15:37:00.000Z", "url": "https://contentlexicon.com/ubiquitous-language", "markdown": "https://contentlexicon.com/ubiquitous-language.md", "credit": "Ubiquitous language, from the Content Lexicon by Contentious (https://contentlexicon.com/ubiquitous-language), CC BY 4.0", "body": "Ubiquitous language is a shared, precise vocabulary for a particular area of work, used by the people who understand it and reflected in the models and systems that support it. The term comes from domain-driven design, where software developers and domain experts build the language together and use it to test the [Domain model](https://contentlexicon.com/domain-model). If a phrase sounds natural in a meeting but names two different rules in a system, the gap is useful evidence: the model or conversation needs attention.\n\nIt is ubiquitous within a bounded context: a part of a domain where terms have agreed meanings. “Customer” might have a precise meaning in sales and another in support. The aim is consistency within each context, with clear relationships where contexts meet, not one corporate dictionary pretending words never shift.\n\n## Why it matters for content\n\nShared terms help people connect a service’s concepts to its [content model](https://contentlexicon.com/content-model), [metadata](https://contentlexicon.com/metadata), editorial guidance and public explanations. For example, a team may need to distinguish someone making an application from someone opening a claim if those stages carry different rules. Agreeing names with domain experts helps [writers](https://contentlexicon.com/writer), product teams and developers make those distinctions in briefs, navigation, forms, help content and structured records.\n\nThat doesn’t mean exposing every internal label to readers. A service may need one precise operational term while using clearer wording in a form or help article. The useful work is knowing which meaning that public phrase maps to and keeping the map current. A [glossary](https://contentlexicon.com/glossary) can record agreed usage, but it can’t create it by itself; the language is tested through decisions and work.\n\nUnresolved differences should [prompt](https://contentlexicon.com/prompt) a question, not a forced compromise. One label can conceal different responsibilities or rules in different parts of an organisation. Making those differences visible gives teams a better basis for content, services and systems than choosing a word everyone can tolerate but nobody means in quite the same way.\n\n## How AI changes the picture\n\nWhen definitions, examples and distinctions appear in content models, [taxonomies](https://contentlexicon.com/taxonomy) and other source material, an AI assistant can use them to draft consistently, map synonyms or flag places where wording has drifted. That gives [content teams](https://contentlexicon.com/content-team) a practical way to check terminology across a large body of material, provided the assistant can identify which context and source apply.\n\nA model can also smooth over a disagreement by choosing the most plausible definition and presenting it with confidence. Terms therefore need scope and authoritative examples, and generated work needs review where a mistaken distinction could change what someone understands or does. A tidy vocabulary helps a system follow the team’s model; it can’t decide whether that model is right or whether its language makes sense to the public.\n\n## Questions people ask\n\n**Is ubiquitous language just a glossary?**\n\nNo. A glossary can record agreed meanings, but the practice is people using and revising the same model in conversation, documentation and systems.\n\n**Should everyone in an organisation use the same terms?**\n\nYes. People should use the same terms for the same concepts and distinguish concepts that mean different things. If “customer” means an organisation in one department and an individual in another, more precise terms such as “customer organisation” and “customer contact” can prevent confusion. Developing a shared language means finding and resolving these ambiguities, rather than expecting everyone to remember which departmental meaning applies.\n\n**Should public content use the organisation’s internal terms?**\n\nInternal and public language should agree on what things mean. Public content doesn’t need to expose the [domain model](https://contentlexicon.com/domain-model) or every specialist term, but a word shouldn’t mean one thing to staff and another to the people using the service. Where an internal term needs a plainer alternative, both should refer unambiguously to the same concept. Developing the shared vocabulary with readers’ needs in mind reduces the need to translate it later.\n\n**How does a team develop its language?**\n\nThrough repeated work with the people who understand the domain: naming concepts, testing those names against concrete cases and revising them when the model changes. The agreed terms then need to appear in the working material that depends on them, from content models and guidance to software.\n\n## Learn more\n\n- [Ubiquitous Language](https://martinfowler.com/bliki/UbiquitousLanguage.html) – Martin Fowler explains Eric Evans’s concept and the way language, domain models and conversations with experts develop together.\n- [Bounded Context](https://martinfowler.com/bliki/BoundedContext.html) – Fowler explains why a large organisation may need distinct models and explicit relationships between them.\n- [Ubiquitous Language in Domain-Driven Design](https://milanjovanovic.tech/blog/ubiquitous-language-ddd) – Milan Jovanović works through examples across conversations, code and tests, including why a term can change meaning between contexts." }, { "term": "User needs", "alsoCalled": [], "slug": "user-needs", "definition": "What people need to know, do or understand, established through evidence and used to decide what content should exist.", "category": "Research and evidence", "related": [ "audience-segmentation", "card-sorting", "content-brief", "content-design", "content-strategy", "design-thinking", "gap-analysis", "persona", "user-testing", "ux" ], "updated": "2026-10-14T21:27:00.000Z", "url": "https://contentlexicon.com/user-needs", "markdown": "https://contentlexicon.com/user-needs.md", "credit": "User needs, from the Content Lexicon by Contentious (https://contentlexicon.com/user-needs), CC BY 4.0", "body": "User needs are what people need to know, do or understand when they come to an organisation's content or services, described from their point of view and grounded in evidence. A useful need is specific enough to guide decisions but broad enough to allow different solutions. “As a carer, I need to know whether I can claim this allowance so I can plan my finances” is a user need. “Raise awareness of our programme” is an organisational aim; it doesn't say what anyone needs from the content.\n\nThe idea grew from user-centred design, which Don Norman and Stephen Draper set out in their 1986 book User Centered System Design. The familiar “As a… I want… so that…” template came from [agile software development](https://contentlexicon.com/agile), where the coach Rachel Davies devised it for user stories at the UK company Connextra in the early 2000s. The UK Government Digital Service put “Start with user needs” first in its 2012 design principles, and [content design](https://contentlexicon.com/content-design) made the user need its starting point.\n\n## Why it matters for content\n\nUser needs give content a reason to exist. They help teams decide what to include, how to structure it and when not to publish anything at all. The need might be to understand a change, compare options, explore an idea or take part in a discussion, not only to complete a transaction. Internal announcements meet staff needs too, though they are often treated as less important than public content.\n\nNeeds also help with prioritisation and with difficult conversations. A request to publish something becomes easier to assess once someone asks which need it meets and what evidence shows that need exists. Primary needs should be easy to find and complete, while less common ones need appropriate routes without overwhelming everyone else. Acceptance criteria can describe what someone should understand or be able to do when the need is met, which gives a draft something to be tested against.\n\nContent has authors, and their needs matter too. A publishing system that makes accurate updates difficult will eventually harm readers, however elegant the public template.\n\n## How AI changes the picture\n\nAI can analyse search terms, support queries, feedback and research notes to suggest patterns in user needs, which helps when the evidence runs to thousands of rows. It can organise and summarise that evidence, but it shouldn't manufacture needs from generic [personas](https://contentlexicon.com/persona) or plausible assumptions. A generated need that sounds right remains an invention until someone checks it against what people do and say.\n\nConversational interfaces may reveal needs in people's own words, including questions that traditional navigation never anticipated. Chat logs can be rich evidence, but they need privacy safeguards and careful interpretation, since people tend to phrase requests around what they think a system can do.\n\nAI products should be evaluated against needs rather than fluency. Ask whether the system helps the person understand, explore or act in the relevant context. A fluent answer can still miss the need, and a task-completion measure may not capture an exploratory one.\n\n## Questions people ask\n\n**Is a user need the same as a business goal?**\n\nNo. Both matter, but a user need describes what a person needs to achieve. Good strategy explains the relationship between them without disguising the organisational aim as a user need.\n\n**Must every need use “As a… I need… so that…”?**\n\nThe format helps by making a team name who has the need and why, but evidence and specificity matter more than the wording. Some teams write needs as the questions people ask, such as “Can I claim this if I work part time?”\n\n**Where do needs come from?**\n\nResearch, search data, support questions, analytics, observation and policy or service evidence. The strongest needs draw on more than one source.\n\n**Can people state their needs directly?**\n\nSometimes, but stated requests and underlying needs may differ. Someone asking for a downloadable [PDF](https://contentlexicon.com/pdf) may need to share information with a relative who isn't online. Combine conversation with observation and behavioural evidence.\n\n## Learn more\n\n- [User Need Statements](https://www.nngroup.com/articles/user-need-statements/) – Sarah Gibbons's NN/g guide to writing needs as a user, a need and a goal, and to keeping solutions out of the statement.\n- [User Story Template: What It Is and Why It Works So Well](https://www.mountaingoatsoftware.com/agile/why-the-three-part-user-story-template-works-so-well) – Mike Cohn on the who, what and why of the agile user story and its origins at Connextra.\n- [Why Content Design Matters: An Interview with Sarah Richards](https://medium.com/ux-and-front-end-interviews/why-content-design-matters-an-interview-with-sarah-richards-a9ad0de3c37a) – an interview on designing around people’s needs rather than an organisation’s desire to publish; one practitioner’s perspective." }, { "term": "User testing", "alsoCalled": [ "Usability testing" ], "slug": "user-testing", "definition": "Watching or listening to real users try to use content or a service, so the team can learn what works and what fails.", "category": "Research and evidence", "related": [ "agile", "card-sorting", "design-thinking", "readability", "user-needs", "ux" ], "updated": "2026-10-06T15:16:00.000Z", "url": "https://contentlexicon.com/user-testing", "markdown": "https://contentlexicon.com/user-testing.md", "credit": "User testing, from the Content Lexicon by Contentious (https://contentlexicon.com/user-testing), CC BY 4.0", "body": "For [content teams](https://contentlexicon.com/content-team), user testing shows whether the words, structure, labels and journeys work in practice rather than only in the meeting where everyone already knew what they meant.\n\nIt can be formal or lightweight, remote or in person, moderated or unmoderated. Many practitioners call it usability testing, which is more precise: the content is on trial, not the person. Either way, people outside the project team try the content and show where it helps and where it confuses.\n\n## Why it matters for content\n\nContent teams make decisions on behalf of users all the time: what to call things, what to explain first, how much detail to provide, what action to offer and how a journey should unfold. User testing gives those decisions some evidence. It replaces “I think people will understand this” with “we watched five people try, and three of them got stuck on the same sentence”.\n\nIt's especially useful for high-stakes journeys, complex information and pages where internal language has crept in unnoticed: the team writes \"eligibility criteria\" while the reader is looking for \"can I apply?\". Testing can reveal problems in headings, labels, instructions, error messages, forms, search terms and content hierarchy. Analytics can tell you what people did. User testing can help you understand why.\n\nDifferent methods answer different content questions. A highlighter test shows which parts feel clear or confusing; a cloze test checks whether people can understand a passage well enough to supply missing words; first-click and tree tests show whether labels and structure send people in the right direction. None replaces observing a full journey, but each can isolate a problem that a broad usability session might only hint at.\n\nThe task you set can distort the result. “Find the eligibility criteria” gives away the language the participant is meant to discover. A better task starts with their situation and asks what they would do next. Good facilitation leaves room for hesitation and wrong turns instead of nudging people towards the answer.\n\n## How AI changes the picture\n\nAI can help prepare and analyse user testing. It can suggest test tasks, summarise notes, cluster observations, compare findings across sessions and help turn evidence into clearer recommendations. That helps most when teams have more research material than time.\n\nWhat it can't do is observe. The boundary worth holding is between analysing real sessions, generating hypotheses and simulating participants. AI can help researchers inspect recordings and organise observations, but a generated reaction is not an observation of anyone using the content. In AI-assisted content work, user testing may also need to include AI-mediated journeys: can someone locate the source behind an answer, recognise a qualification and recover when the assistant misunderstands? Test that with people rather than treating a convincing transcript as evidence of success.\n\n## Questions people ask\n\n**How many people do you need for user testing?**\n\nIt depends on the question, but small tests can be valuable. Even five participants can reveal repeated usability problems. The aim isn't always statistical certainty. Often it's to find clear, fixable obstacles before they become expensive public problems.\n\n**Can we test content before it's designed?**\n\nYes. You can test draft copy, [information architecture](https://contentlexicon.com/information-architecture), labels, prototypes, forms, search terms and task flows before a final interface exists. That's often wise, because fixing content early is less painful than prising it out of a finished design.\n\n**Is user testing only for websites?**\n\nFar from it. You can test [emails](https://contentlexicon.com/email), letters, guidance, forms, app content, support scripts, onboarding material, intranet pages and service journeys. If someone has to understand or use it, it can usually be tested.\n\n**What should we do with the findings?**\n\nTurn them into specific content decisions. “Users were confused” isn't enough. Which word, step, assumption or missing explanation caused the confusion? What will change, who will change it and how will you know whether the change helped?\n\n## Learn more\n\n- [Usability (User) Testing 101](https://www.nngroup.com/articles/usability-testing-101/) – a clear introduction to participants, tasks and observation.\n- [Usability Test Demo by Steve Krug](https://www.youtube.com/watch?v=1UCDUOB_aS8) – a complete session showing the method rather than merely describing it.\n- [Evaluating AI-Simulated Behavior: Insights from Three Studies on Digital Twins and Synthetic Users](https://www.nngroup.com/articles/ai-simulations-studies/) – Raluca Budiu examines where simulations can predict responses or extend survey data, and where bias, limited context and reduced variability restrict what can be inferred." }, { "term": "UX", "alsoCalled": [ "User experience" ], "slug": "ux", "definition": "User experience: how someone experiences using a service, product, website or piece of content.", "category": "Foundations", "related": [ "accessibility", "content-design", "frontend", "user-needs", "user-testing", "ux-writer" ], "updated": "2026-10-11T06:25:00.000Z", "url": "https://contentlexicon.com/ux", "markdown": "https://contentlexicon.com/ux.md", "credit": "UX, from the Content Lexicon by Contentious (https://contentlexicon.com/ux), CC BY 4.0", "body": "Someone can complete a task successfully and still come away frustrated, uncertain or reluctant to return. UX takes that wider experience seriously, including how well something meets people's needs and how accessible it is. It applies to physical products and services as well as [websites](https://contentlexicon.com/website) and apps.\n\nBooking a train journey involves finding a route, comparing fares, paying and retrieving your ticket when you need it. A clear screen helps, but so do sensible ticket options, a payment process that doesn't lose your booking and a ticket you can open without a mobile signal. An attractive app can still leave you stranded at the barrier.\n\n## Why it matters for content\n\nFor content work, words are part of the experience, not decoration added once the interface is finished. Labels, headings, instructions, error messages, help text, navigation, search results and [tone](https://contentlexicon.com/tone) all shape whether something works for the person using it.\n\nContent is often where user experience succeeds or collapses. A well-designed form or a beautiful navigation system can still fail if the questions are unclear or the labels use internal language. However good the flow, a service will feel hostile if the error messages are cold or vague.\n\nUX gives [content teams](https://contentlexicon.com/content-team) a way to think beyond individual pages and into journeys. What does the user know at this point? What are they trying to do, and in whose words? Which reassurance, instruction or evidence will help them? These are content questions as much as design questions.\n\n## How AI changes the picture\n\nAI introduces new user experiences and new content responsibilities. People may now interact with organisations through chatbots, generated answers, search summaries, voice interfaces and agentic tools. The content behind those experiences affects what the system says, how helpful it feels and whether it can be trusted. But the experience also depends on how the interaction works. Can someone check an answer, correct a misunderstanding or reach a person? If an AI tool takes action on their behalf, can they see what it plans to do and stop or undo it? Fluent answers are only one part of a usable service.\n\nAI can support UX work too, by summarising research, drafting interface copy, analysing feedback and testing variants. But it can easily generate plausible [microcopy](https://contentlexicon.com/microcopy) that sounds fine while failing the journey. The machine can help make options. It can't feel the small panic of a user who doesn't know which button will cancel their application.\n\n## Questions people ask\n\n**Where does content fit into UX?**\n\nContent helps people understand, decide and act. It appears in navigation, labels, instructions, forms, messages, help pages, onboarding, product screens and support journeys.\n\n**What's the difference between UX and UI?**\n\nUI stands for user interface: the controls and displays through which someone interacts with a product. Buttons, menus, input fields and screen layouts are parts of a digital interface. UX includes the experience of using that interface, but also whether the product does something useful and how it fits into the person's wider task. A beautifully designed booking form is little comfort if there are no suitable appointments.\n\n**Is UX only a designer's responsibility?**\n\nUX specialists help teams understand people's needs and test design decisions. But developers, content practitioners, product managers and the people running the service all influence the experience. A designer can't resolve a restrictive returns policy by rearranging the screen. The team needs the authority and willingness to change what causes the problem.\n\n**Can good content fix bad UX?**\n\nSometimes it can reduce harm, but it can't rescue everything. Content can clarify, guide and reassure. It can't make an impossible process sensible by sprinkling friendly words over it. There are limits to even the most heroic tooltip.\n\n## Learn more\n\n- [The definition of user experience](https://www.nngroup.com/articles/definition-user-experience/) – Don Norman and Jakob Nielsen's 1998 account of UX beyond the interface and beyond usability.\n- [What Is User Experience (and What Is It Not)?](https://www.nngroup.com/articles/what-is-user-experience) – Kate Kaplan's correction to the habit of treating UX as interface polish.\n- [How to be human](https://contentious.ltd/thoughts/how-to-be-human) – a reminder that good content experiences are relationships built through respect, learning and communication." }, { "term": "UX writer", "alsoCalled": [], "slug": "ux-writer", "definition": "A writer specialising in clear interface words that help people complete tasks.", "category": "Roles", "related": [ "content-design", "content-designer", "design-system", "microcopy", "ux", "writer" ], "updated": "2026-10-09T18:28:00.000Z", "url": "https://contentlexicon.com/ux-writer", "markdown": "https://contentlexicon.com/ux-writer.md", "credit": "UX writer, from the Content Lexicon by Contentious (https://contentlexicon.com/ux-writer), CC BY 4.0", "body": "A [UX](https://contentlexicon.com/ux) [writer](https://contentlexicon.com/writer) writes and looks after the language inside a digital product: labels, buttons, instructions, error messages, confirmations and onboarding. They usually sit within a product design team, working alongside designers, researchers and developers, and they're accountable for whether people understand what the product is doing and what it needs from them.\n\nThe role overlaps heavily with [content designer](https://contentlexicon.com/content-designer), and many organisations now use that title for the same job. Where both exist, a content designer usually takes a broader view of the whole journey and whether content should exist at all, while a UX writer concentrates on interface language.\n\n## Why an organisation needs a UX writer\n\nWithout one, interface language gets written by whoever is nearest: a designer filling a placeholder, a developer naming a button after the function behind it, a product manager pasting in legal text. Each choice is reasonable on its own, and together they produce a product that calls the same thing three names and explains errors in terms only its builders understand. A UX writer gives that language an owner who can see across screens and teams, keep terminology consistent and test whether people understand it.\n\nThe need tends to appear as a product grows: more features, more teams shipping changes, a [design system](https://contentlexicon.com/design-system), translation into other languages or a rise in support calls about things the interface should have explained. The role works best when it joins projects early enough to influence the flow, rather than being handed finished screens to fill. Better words can't rescue a confusing process on their own, but a UX writer is often well placed to show where the process needs changing.\n\n## How AI changes the role\n\nWhen a product includes an AI assistant, much of the interface stops being fixed text. The words people see are generated in the moment, so the UX writer's work moves into what shapes them: the system instructions that set an assistant's [voice](https://contentlexicon.com/voice) and boundaries, example responses, the wording of fallbacks when it can't help and the points where it should hand over to a person. At Microsoft, content designers write the [system prompts](https://contentlexicon.com/system-prompt) behind Copilot features alongside engineers and product managers. Nobody can approve every sentence a user will see, so the job becomes setting the conditions for good sentences and testing whether they hold.\n\nThat changes what to hire for. Alongside interface writing, a job description might now ask for experience of writing and testing instructions for language models, evaluating generated responses at scale and working with engineers on grounding and safety. The established craft still matters: fixed interface text hasn't disappeared, and an assistant that's eloquent about a booking it hasn't confirmed is worse than a plain error message.\n\n## Questions people ask\n\n**Should we hire a UX writer or a content designer?**\n\nIt depends on the problem. If the main issue is inconsistent or confusing interface language, a UX writer fits. If the questions run deeper – what content a service needs, in which format and where – a content designer's broader remit helps. Many job ads use the titles interchangeably, so read the responsibilities rather than the title.\n\n**Where should a UX writer sit?**\n\nUsually in product design, close to designers and researchers. Placing the role in marketing tends to pull it towards persuasion and away from the product decisions that shape the words.\n\n**When should a UX writer join a project?**\n\nBefore layouts harden. Language exposes structural problems, such as a step that needs explaining because it shouldn't exist, and those are cheap to fix in a sketch and expensive in code.\n\n**What background suits the role?**\n\n[Copywriting](https://contentlexicon.com/copywriting), journalism, technical writing and [content design](https://contentlexicon.com/content-design) all lead into it. More telling than the route are comfort with research and testing, the ability to argue for a word choice with evidence and enough technical curiosity to work with developers and, increasingly, with language models.\n\n## Learn more\n\n- [Content Designers: The unsung heroes of design](https://microsoft.design/articles/content-designers-the-unsung-heroes-of-design) – Microsoft Design's 2024 account of content designers writing the system prompts behind Copilot, and of interface language becoming probabilistic rather than fixed.\n- [What I wish I knew before becoming a UX content designer](https://indeed.design/article/what-i-wish-i-knew-before-i-became-a-ux-content-designer/) – thoughtful reflections on synthesis, logic, research and the skills beyond prose.\n- [What is UX writing?](https://uxcontent.com/what-is-ux-writing) – UX Content Collective on interface writing as a product and design discipline, not a supply of tiny sentences." }, { "term": "Versioning", "alsoCalled": [], "slug": "versioning", "definition": "Keeping track of different versions so changes can be compared, reviewed or restored.", "category": "Structure and systems", "related": [ "branch", "cms", "git", "retention-and-archiving", "rollback", "single-source-of-truth" ], "updated": "2026-10-14T21:33:00.000Z", "url": "https://contentlexicon.com/versioning", "markdown": "https://contentlexicon.com/versioning.md", "credit": "Versioning, from the Content Lexicon by Contentious (https://contentlexicon.com/versioning), CC BY 4.0", "body": "Versioning is the practice of recording distinct states of something over time, with enough information to tell them apart, compare them and return to one if needed. A good version record links each state to a date, an author and a reason for the change.\n\nThree kinds of version are often confused. Version history is the automatic record a developer, [CMS](https://contentlexicon.com/cms) or editor keeps of every save. It's useful for recovery but noisy. Published versions are the states users or readers saw, each with the date it went live and, where it matters, who approved it. Named releases are deliberate, labelled states, sometimes maintained side by side, such as documentation for version 2 and version 3 of a product, or a guide marked 2.1 while work continues on 2.2. Software often numbers releases using semantic versioning, where the number signals how significant the change is.\n\n## Why it matters for content\n\nVersioning supports collaboration and accountability. Editors can review differences rather than reread everything, and teams can trace when a claim changed and who approved it. An editor checking a changed eligibility rule needs to know which text was published, which is under review and what changed between them.\n\nA retained earlier state can support recovery after an error. Check that it's still suitable and compatible with current systems, because version history alone doesn't make restoration safe.\n\nSome old versions should stay public. Customers on an older product need documentation that matches what they're using, so many documentation sites keep several releases available with a version switcher. Terms and conditions, privacy notices and policies may need earlier versions available because people agreed to them or decisions were made under them. Archived versions should be clearly labelled, kept out of search results where possible and linked to the current one. Each maintained version adds work, so retire versions once nobody relies on them.\n\nVersioning needs conventions: what creates a new version, which one is current, how versions are named and how long history is kept. Record enough context to explain a consequential change, not only its timestamp. A changelog does this for readers, summarising notable changes in [plain language](https://contentlexicon.com/plain-language) rather than listing every edit.\n\n## How AI changes the picture\n\nAI can compare versions, summarise changes and flag possible regressions. It can help reviewers focus on altered meaning and dependencies.\n\nVersion the instructions, retrieval sources and configuration that an evaluation depends on. Record the exact state tested, including a model version where the provider exposes it. Some services change without offering a reproducible snapshot.\n\nGenerated content also needs provenance, so teams know which source, instruction and review produced the published state.\n\n## Questions people ask\n\n**Does a CMS revision history count as versioning?**\n\nPartly. It records every save, which helps with recovery, but it rarely shows which state readers saw, who approved it or why it changed. That takes labels, approval records and release notes on top.\n\n**What is the current version?**\n\nThe published or approved state, and it should be marked as such. A document labelled \"current\" in an [editing](https://contentlexicon.com/editing) folder may be the next draft, not the version readers are using.\n\n**Should every tiny edit create a version?**\n\nKeep automatic edit history where useful, but label releases that readers or maintainers need to tell apart. A typo correction may not need a new release number. A changed policy may need a clearly identified state.\n\n**Can versions be deleted?**\n\nRetention depends on legal, operational and historical needs. Whatever is kept, people using the content shouldn't mistake an archived version for the current one.\n\n## Learn more\n\n- [Documentation versioning in Docusaurus](https://docusaurus.io/docs/versioning) – a concrete guide to released documentation snapshots and an evolving current version, including the added complexity and reasons to limit maintained versions.\n- [Semantic Versioning 2.0.0](https://semver.org/) – the widely used convention for numbering releases so the number signals whether a change is a fix, an addition or a breaking change.\n- [Keep a Changelog](https://keepachangelog.com/en/1.1.0/) – a short convention for writing changelogs for people rather than machines, with an entry for every version and the latest first." }, { "term": "Voice", "alsoCalled": [], "slug": "voice", "definition": "An organisation's recognisable personality in words.", "category": "Editorial craft", "related": [ "content-principles", "editing", "machine-readable", "message-architecture", "storytelling", "style-guide", "tone" ], "updated": "2026-10-07T06:41:00.000Z", "url": "https://contentlexicon.com/voice", "markdown": "https://contentlexicon.com/voice.md", "credit": "Voice, from the Content Lexicon by Contentious (https://contentlexicon.com/voice), CC BY 4.0", "body": "Voice is the underlying character people hear across channels, topics and formats: formal or conversational, warm or reserved, plain-spoken or technical, brisk or expansive. Voice should remain relatively stable even when the situation changes.\n\n[Tone](https://contentlexicon.com/tone) is a different thing. It flexes according to audience, context and moment; voice is the through-line that makes the content still sound like the same organisation.\n\n## Why it matters for content\n\nTake a message confirming that someone's application has arrived. One organisation writes \"Your application has been received. A response will be issued within five working days.\" Another writes \"Got it. We'll be in touch within five working days.\" The information is the same; the voice tells you who's talking, and whether you'd want to talk back. Without a defined voice, organisations sound like a different person on every page, depending on who wrote the copy, which team owns it or whether the last draft came from legal, marketing or an AI assistant.\n\nVoice needs pinning down once many people create content, once it spreads across several channels, or when the organisation wants to balance clarity with a distinctive point of view. It works best as a shared editorial reference that people check drafts against. Contentious describes its own as sassy, warm and wise, each with a \"never\" attached: sassy, never disrespectful; warm, never overbearing; wise, never superior. (The \"nevers\" do a lot of the work.)\n\n## How AI changes the picture\n\nA model can infer stylistic patterns from examples, but examples alone can leave it copying surface habits or missing the boundaries. Written voice guidance, paired with examples of what works and what doesn't, gives [writers](https://contentlexicon.com/writer) and systems a shared reference for producing and checking drafts.\n\nThen there's the temptation to accept fluency as a substitute. Generated drafts can be smooth, agreeable and slightly breathless, and a draft can be grammatically flawless while sounding like nobody in particular. Voice guidance gives editors a basis for rejecting polished copy that doesn't sound like the organisation.\n\n## Questions people ask\n\n**Can an organisation have one voice?**\n\nNot a rigid one. A charity's letter from its chief executive and its cookie banner will never sound identical, but they can share recognisable characteristics: the same directness, the same refusal to hide behind [jargon](https://contentlexicon.com/jargon). Sub-brands and campaigns can stray further from the centre as long as the family resemblance holds.\n\n**Is voice the same as tone?**\n\nThey're related but different. Voice is the underlying personality; tone is how that personality flexes for different situations. A warm voice stays warm in a complaint response, but the tone becomes calmer and plainer.\n\n**Is voice just a branding extra?**\n\nNo. Voice affects whether people trust what they read and how they interpret it. The same reassurance written in a stiff, legalistic register can read like a warning.\n\n**Can AI write in our voice?**\n\nUp to a point. It works far better when the voice is defined with examples and usable guidance than when it's left as a vague instruction such as \"sound warm\". A few paragraphs you're proud of, and a few you'd never publish, give it something to aim at.\n\n## Learn more\n\n- [Mailchimp voice and tone](https://styleguide.mailchimp.com/voice-and-tone/) – a well-known example of defining character and then adapting it to context.\n- [Monzo writing principles](https://monzo.com/tone-of-voice) – three principles, dialled up or down by channel, from operational messages to social posts.\n- [The three tenets of a content style guide](https://voicetoneandstyle.com/learn/three-tenets-of-a-content-style-guide) – how voice sits alongside culture, editorial decisions and technical structure in a guide that people can use." }, { "term": "WCAG", "alsoCalled": [ "Web Content Accessibility Guidelines" ], "slug": "wcag", "definition": "The Web Content Accessibility Guidelines: the main international standard for accessible digital content.", "category": "Editorial craft", "related": [ "accessibility", "alt-text", "automated-testing", "content-governance" ], "updated": "2026-10-11T17:46:00.000Z", "url": "https://contentlexicon.com/wcag", "markdown": "https://contentlexicon.com/wcag.md", "credit": "WCAG, from the Content Lexicon by Contentious (https://contentlexicon.com/wcag), CC BY 4.0", "body": "WCAG is the W3C's standard for making web content accessible to disabled people, and the reference point most [accessibility](https://contentlexicon.com/accessibility) laws and policies use. It sets out what content and interfaces need to achieve rather than how to build them, so it applies to text, images, video, forms and documents as well as code.\n\nIt's built in layers. Four principles say content must be perceivable, operable, understandable and robust. Under them sit guidelines, 13 in WCAG 2.2, and under those, testable success criteria, which are what an audit checks. Each criterion has a level: A for the most basic requirements, AA for the usual target and AAA for the most demanding. Most laws and procurement rules point to level AA of WCAG 2.1 or 2.2, so name the version and level when setting a target.\n\nWCAG 1.0 was published in 1999. WCAG 2.0 followed in 2008 with the technology-neutral, testable approach still used today, and versions 2.1 (2018) and 2.2 (2023) added criteria for mobile, low vision and cognitive needs. WCAG 3, a broader rewrite with a different structure, is still a working draft and isn't expected to replace WCAG 2 for some years.\n\n## Why it matters for content\n\nWCAG gives teams a shared standard rather than leaving accessibility to taste or intention. It supplies testable criteria, but conformance doesn't establish that every disabled person can use the experience, so involve people and look at needs beyond the checklist too.\n\nMany criteria are met or failed by editorial choices before any developer touches the code. [Writers](https://contentlexicon.com/writer) and editors own text alternatives for images (1.1.1), headings and structure that reflect the content (1.3.1 and 2.4.6), page titles (2.4.2), link text that makes sense in context (2.4.4), labels and instructions on forms (3.3.2), error messages that say what went wrong and how to fix it (3.3.1 and 3.3.3), not relying on colour alone (1.4.1) and captions and transcripts for [audio](https://contentlexicon.com/audio) and video. At AAA, reading level (3.1.5) asks for supplementary content when text needs more than lower secondary education to understand.\n\nConformance covers the content and service as a whole, including editorial choices, so a site can fall out of conformance because of what's published on it long after launch.\n\n## How AI changes the picture\n\nAI makes the standard more important, because content is now generated, transformed and summarised in more places. WCAG gives teams a way to judge whether those outputs are usable.\n\nAI can help with some checks and first-pass fixes, such as drafting [alt text](https://contentlexicon.com/alt-text) or flagging vague links. It can also create failures at scale: bland link text, alt text that describes an image without conveying its purpose, summaries that drop key steps or reformatted content that loses its heading structure. The existence of a machine doesn't lower the standard.\n\n[Automated testing](https://contentlexicon.com/automated-testing) catches only some WCAG failures. Criteria that depend on meaning, such as whether alt text is appropriate or a heading describes its section, still need a person to judge.\n\n## Questions people ask\n\n**Is WCAG only for developers?**\n\nNo. Developers are crucial, but writers, designers and editors affect many of the outcomes WCAG is trying to secure.\n\n**Do we need AAA?**\n\nAA is the common target, but required versions and levels depend on the applicable policy, contract or law. Some AAA criteria can improve a particular experience; don't assume a blanket AAA target suits every kind of content.\n\n**Is WCAG the whole of accessibility?**\n\nNo. It's a standard and a very important one, but accessibility also includes culture, testing with disabled people, judgement and ongoing maintenance.\n\n**Can AI check WCAG compliance for us?**\n\nOnly partly. Some checks can be automated, and AI can extend them, but many criteria still need human judgement and testing with assistive technology.\n\n## Learn more\n\n- [WCAG 2 Overview](https://www.w3.org/WAI/standards-guidelines/wcag/) – W3C's explanation of published versions, success criteria and conformance, including WCAG 2.2; the reference point for specifying a target.\n- [Easy Checks](https://www.w3.org/WAI/test-evaluate/preliminary/) – W3C's preliminary checks for common accessibility issues; passing them isn't a complete conformance assessment.\n- [Writing Clearly and Simply](https://webaim.org/techniques/writing/) – WebAIM's guidance for writers on organising ideas, using headings, avoiding jargon and writing for people with reading or cognitive disabilities." }, { "term": "Website", "alsoCalled": [], "slug": "website", "definition": "The part of the web an organisation owns outright, where its content can inform, educate, persuade and support, and where anything said about it elsewhere can be checked.", "category": "Operations", "related": [ "accessibility", "channel-strategy", "cms", "content", "geo", "microsite", "paddle-swim-dive", "slow-content", "structured-content" ], "updated": "2026-10-07T17:50:00.000Z", "url": "https://contentlexicon.com/website", "markdown": "https://contentlexicon.com/website.md", "credit": "Website, from the Content Lexicon by Contentious (https://contentlexicon.com/website), CC BY 4.0", "body": "A website is one or more pages and services published under a domain an organisation controls. It may include search, forms, tools, transactions, documentation and downloadable data.\n\nA [microsite](https://contentlexicon.com/microsite) is a website with a narrow purpose. An intranet is a website for an organisation's own staff. A profile on a social platform is not a website: the platform controls how it works, who sees it and whether it remains available.\n\nThat control makes the website the natural home for the definitive version of things: the organisation decides what goes where, how deep it goes, how long it stays and which version is current.\n\nSocial feeds, newsletters and press coverage carry messages towards people, and the website is where they arrive to find out what is true and use what is on offer.\n\n## Why it matters for content\n\nA website is better understood as the [frontend](https://contentlexicon.com/frontend) of a connected content system rather than as a technical solution or a pile of documents with the same header.\n\nContent can inform, educate, persuade, entertain or support. Many websites stop at the first: they describe the organisation's services rather than providing them. A page explaining that a helpline exists is information about support; a page that helps someone through the first hour after a diagnosis _is_ support. Software documentation is half the product. An evidence library is how a think tank gets quoted.\n\nA website is also where many buying decisions happen, as well as actual purchases. People use a website’s content to discover products and services, understand what they will get, compare their options and judge whether what they will receive will deliver what it promises.\n\nThe cycle of a full website redesign every four or five years often rebuilds the container and keeps the [content model](https://contentlexicon.com/content-model), so the new site is a slightly better version of the old one. What changes the result are solid content foundations and the continuous, unglamorous work: an owner for each section, [review dates](https://contentlexicon.com/review-date), pages that work for disabled people and on a weak phone signal, and a reason for every page to exist. Without it, a website gradually becomes an archive of organisational indecision presented as current information.\n\n## How AI changes the picture\n\nAI summaries and assistants can answer some questions without a visit to the website. That changes what a visit is for: people may arrive to verify something, explore it in depth or use a service that an answer alone can't provide.\n\nThe website remains an important source of information, but its direct value also lies in experiences the organisation can offer. An interactive exhibition guide, for example, could let a visitor compare objects, listen to a curator and choose a route around their interests. A summary can introduce the exhibition; the website can help someone explore it.\n\nAs AI opens up cheaper, faster software development, the boundary between website and software will become increasingly blurred. This will speed up a trend that was already happening: websites will become places people and agents do things (apply, join, create), as much as opportunities to read and absorb. Expect more interaction and less broadcasting.\n\n## Questions people ask\n\n**Do we still need a website if people get their answers from AI?**\n\nMore than before, in one respect: the answers are made from somewhere, and your own site is the version you control. What changes is the job. Fewer people browse, and the ones who come are checking a claim, going deeper or doing something, so those pages deserve care.\n\n**Our traffic is falling. Is the website failing?**\n\nIt may be doing better. Traffic was always a proxy. Look at completions instead (applications, donations, bookings, downloads of the full report), at questions the support team gets that the site should have answered, and at what AI assistants say about you when asked. A site with a third fewer visits and the same number of applications is doing more with less.\n\n**How often should we redesign?**\n\nThe container rarely, the content continuously. A sound site can go five years or more between rebuilds if someone is reviewing, merging and retiring pages every month. Rebuilding every few years without changing how content is owned and maintained mostly moves the same problems to a new address.\n\n**Should every piece of content be a web page?**\n\nNo. Some material belongs in a service, a tool, a dataset, an [email](https://contentlexicon.com/email) or an accessible document. Choose the format around what people need and who'll maintain it, not habit.\n\n## Learn more\n\n- [Google users are less likely to click on links when an AI summary appears in the results](https://www.pewresearch.org/short-reads/2025/07/22/google-users-are-less-likely-to-click-on-links-when-an-ai-summary-appears-in-the-results/) – Pew Research Center's analysis of real browsing data, showing how much less often people click through when an AI summary appears.\n- [POSSE](https://indieweb.org/POSSE) – the IndieWeb practice of publishing on your own site first and sharing copies elsewhere, with links back to the original.\n- [A Brief History & Ethos of the Digital Garden](https://maggieappleton.com/garden-history) – Maggie Appleton explores personal websites as evolving, interconnected spaces for ideas, offering an alternative to chronological feeds and finished publications." }, { "term": "Wireframe", "alsoCalled": [], "slug": "wireframe", "definition": "A sketch of the structure of a page or screen before the final design is made.", "category": "Structure and systems", "related": [ "component", "content-design", "information-architecture", "typography" ], "updated": "2026-10-06T15:45:00.000Z", "url": "https://contentlexicon.com/wireframe", "markdown": "https://contentlexicon.com/wireframe.md", "credit": "Wireframe, from the Content Lexicon by Contentious (https://contentlexicon.com/wireframe), CC BY 4.0", "body": "A wireframe usually shows layout, hierarchy, content areas, interactions and priority without settling every detail of colour, [typography](https://contentlexicon.com/typography) or imagery.\n\nWireframes range from quick pencil sketches to more detailed digital layouts. Their value is not visual polish. It is that they make an emerging idea concrete enough to discuss, test and change while change is still cheap. A wireframe covered in useful questions is often doing a better job than one wearing a smart blazer.\n\n## Why it matters for content\n\nWireframing reveals the relationship between content and interface early. It shows whether the important information appears in the right order, whether a heading carries enough meaning, whether an action has suitable context and whether the page is trying to fit six priorities into a space designed for two.\n\nContent people should be involved before the boxes harden into decisions. If wireframes use only generic labels and [lorem ipsum](https://contentlexicon.com/lorem-ipsum), teams can approve a layout that collapses as soon as real content arrives. Realistic headings, summaries, errors and edge cases help expose structural problems. They also allow the team to design around [user needs](https://contentlexicon.com/user-needs) rather than around whatever quantity of text looks tidy.\n\nWireframes are especially useful for complex journeys, forms and services. They let content, design and development examine the same proposition from different angles: what the person needs to know, what the system needs from them and what should happen next.\n\n## How AI changes the picture\n\nAI can produce rough layouts, suggest content [components](https://contentlexicon.com/component) and generate placeholder copy quickly. That can make exploration faster, but speed is not the same as understanding. A model can draw a conventional dashboard or landing page without knowing whether that pattern suits the task, audience or evidence.\n\nAI also makes realistic content easier to bring into early prototypes. Teams can create varied examples, long names, missing data, error states and alternative user journeys instead of testing only the neat happy path. These generated examples should be checked, but they can help reveal where a structure is brittle.\n\nConversely, wireframes now need to account for AI-shaped interactions: uncertain answers, source links, user corrections, consent, escalation and the possibility that generated content changes length dramatically. Designing the behaviour and content states matters more than decorating the rectangle.\n\n## Questions people ask\n\n**Should wireframes contain real content?**\n\nThey should contain content realistic enough to test the structure. Final prose is unnecessary, but meaningful headings, representative lengths and awkward cases are invaluable.\n\n**Is a wireframe the same as a prototype?**\n\nNot usually. A wireframe mainly explores structure and hierarchy. A prototype simulates behaviour or a journey and may be interactive. The two can overlap, particularly in wireflows.\n\n**Who should create the wireframe?**\n\nOften a designer leads, but content, research and development should contribute. Collaborative sketching is useful because the artefact belongs to the problem, not to a professional territory.\n\n**How polished should a wireframe be?**\n\nOnly as polished as the decision requires. Low fidelity encourages change. High fidelity can help test detail, but it may also make weak ideas look prematurely settled.\n\n## Learn more\n\n- [How to Draw a Wireframe (Even if You Can't Draw)](https://www.nngroup.com/articles/draw-wireframe-even-if-you-cant-draw/) – a step-by-step guide to sketching conventions for common wireframe components, from navigation and headers to buttons, dialogs and progress bars.\n- [Wireflows: A UX Deliverable for Workflows and Apps](https://www.nngroup.com/articles/wireflows/) – an explanation of combining wireframes and flowcharts for dynamic interactions.\n- [Writing for Web Accessibility](https://www.w3.org/WAI/tips/writing/) – useful checks for the real content placed into layouts." }, { "term": "Word list", "alsoCalled": [], "slug": "word-list", "definition": "A record of an organisation's decisions on how to spell, capitalise, hyphenate and use particular words and terms.", "category": "Editorial craft", "related": [ "copy-editing", "linting", "style-guide", "taxonomy" ], "updated": "2026-10-15T15:47:00.000Z", "url": "https://contentlexicon.com/word-list", "markdown": "https://contentlexicon.com/word-list.md", "credit": "Word list, from the Content Lexicon by Contentious (https://contentlexicon.com/word-list), CC BY 4.0", "body": "A word list records how an organisation writes the words that cause hesitation: whether it's “[email](https://contentlexicon.com/email)” or “e-mail”, “log in” or “login”, which product and team names take capitals and which terms it avoids altogether. It usually sits inside a [style guide](https://contentlexicon.com/style-guide), often arranged A to Z, and each entry gives the preferred form with enough context to apply it. A good list is local. It doesn't repeat what a dictionary already says; it settles the choices where the dictionary allows more than one answer or where the organisation has reasons of its own.\n\nPrinters have kept lists like this for well over a century. Horace Hart of the Oxford University Press first issued his Rules for Compositors and Readers in 1893 as a small booklet for staff, and it grew over many editions into Hart's Rules. Newspapers developed house style books for the same reason, and the Guardian's online A–Z still settles questions such as whether to write “an” or “a” before “historian”.\n\n## Why it matters for content\n\nSmall terminology decisions recur constantly. Without a word list, [writers](https://contentlexicon.com/writer) look up or debate the same words again and again, editors correct them inconsistently and readers meet variation that suggests carelessness or, worse, that two words mean two different things. If a help article says “account number”, the form asks for a “customer reference” and the letter mentions a “membership ID”, people reasonably wonder whether they need three numbers. A word list turns scattered preferences into one place to check.\n\nGood entries are short and practical. They give the preferred form, any discouraged alternatives, an example where usage is subtle and a link to related guidance. Many of the most useful distinctions are grammatical rather than a matter of taste: “log in” as a verb, “login” as a noun. High-use and high-risk terms deserve priority, since an exhaustive catalogue isn't useful if people can't find the decision they need.\n\nWord lists matter most where the same concept appears in many places: interfaces, support content, legal text, translation and product documentation. Consistency there helps people recognise actions and reduces rework for translators. Lists also grow by accretion, so they need occasional pruning. An entry whose reason nobody remembers is either missing an explanation or no longer needed.\n\n## How AI changes the picture\n\nA word list is one of the easiest parts of a style guide to turn into something a machine can use. Structured as preferred terms, alternatives to replace and the context for each, it can feed AI drafting instructions, automated checks and [editing](https://contentlexicon.com/editing) tools that flag “e-mail” or a retired product name as someone types. That moves the list from final review to the point of writing.\n\nAutomation also exposes weak entries. “Avoid where possible” may make sense to an experienced editor, but a system needs to know when it's possible. Entries need context, exceptions and examples, or a checker will either nag constantly or replace words where the distinction mattered.\n\nAI can help maintain the list as well, by spotting recurring corrections, inconsistent terms across a site or new words appearing in search queries and support tickets. People should decide which variations matter and whether a suggested replacement preserves meaning and inclusion.\n\n## Questions people ask\n\n**What belongs in a word list?**\n\nWords people hesitate over or use inconsistently: compounds and hyphenation, capitalisation of product and team names, abbreviations, spellings with accepted variants, terms the organisation avoids and [inclusive-language](https://contentlexicon.com/inclusive-language) choices. If writers keep asking about a word, it belongs on the list.\n\n**How is it different from a glossary?**\n\nA [glossary](https://contentlexicon.com/glossary) explains what terms mean, usually for readers. A word list tells writers which form to use. The two often link, since a product term may need a definition in the glossary as well as an entry on its capitalisation.\n\n**Should every entry ban alternatives?**\n\nOnly where the alternative is wrong for the organisation. Many words depend on context, so explain the distinction rather than inventing a universal rule.\n\n**Who maintains it?**\n\nUsually whoever owns the style guide, with input from subject experts, product teams, translators and the questions writers ask. Record why each decision was made, so the next person can tell a considered choice from an accident.\n\n## Learn more\n\n- [Rules for compositors and readers at the University Press, Oxford](http://www.gutenberg.org/ebooks/71188) – the 22nd edition of Horace Hart's rules, free on Project Gutenberg; a house word list in its original setting, written for the people setting type.\n- [Guardian style guide: A](https://www.theguardian.com/guardian-style-guide-a) – the first letter of the Guardian's A–Z, from “a or an before H?” to how to write abbreviations and acronyms.\n- [Mailchimp word list](https://styleguide.mailchimp.com/word-list) – a brisk example sorted into standardised spellings, words to use carefully and words to avoid, deferring to the AP Stylebook for everything else.\n- [Microsoft Writing Style Guide](https://learn.microsoft.com/en-us/style-guide/welcome/) – the online successor to the Microsoft Manual of Style, with an A–Z word list and term collections for writing about technology." }, { "term": "Workflow", "alsoCalled": [], "slug": "workflow", "definition": "The route content takes from idea through drafting, review, publication and ongoing maintenance.", "category": "Operations", "related": [ "agile", "author-experience", "backend", "channel-strategy", "content-calendar", "content-governance", "content-lifecycle", "content-operating-model", "content-operations", "content-plan", "graph-engineering", "human-in-the-loop", "pull-request", "sign-off", "stakeholder-mapping", "systems-thinking" ], "updated": "2026-10-07T13:53:00.000Z", "url": "https://contentlexicon.com/workflow", "markdown": "https://contentlexicon.com/workflow.md", "credit": "Workflow, from the Content Lexicon by Contentious (https://contentlexicon.com/workflow), CC BY 4.0", "body": "Workflow includes the stages, people, decisions, checks and handovers involved. A workflow can be formal and documented, or informal and mostly unwritten, but either way it shapes how content gets made.\n\nIn content work, workflow is rarely just about producing a draft. It also covers review, [sign-off](https://contentlexicon.com/sign-off), publishing, updating, retiring and the moments where work stalls because nobody is quite sure whose turn it is.\n\n## Why it matters for content\n\nWith a workflow, content stops depending on someone staying late to rescue it. It helps teams make work visible, reduce bottlenecks, clarify who does what and build checks into the process instead of catching problems the night before launch. It also makes it easier to improve quality, because recurring problems can be traced to a stage or handover rather than blamed on individual people.\n\nThe need is greatest where several people touch the same content, where approval or compliance matters, or where content needs ongoing care after it goes live. Good workflow is one of the main differences between content that's maintained and content that's merely launched.\n\n## How AI changes the picture\n\nAI can change how work is organised. An approved source update could [prompt](https://contentlexicon.com/prompt) a draft revision, identify other affected pages and request the appropriate specialist review. A change to eligibility rules might need a different route from correcting a broken link. This is a possible [content-maintenance](https://contentlexicon.com/content-maintenance) design: the team defines the routes and authority, rather than leaving an agent to decide what it can publish.\n\nAI adds new stages and new decisions to workflow. Someone has to decide when AI is used, what source material it can draw from, what gets reviewed by a person, how changes are checked and where automated steps sit in the sequence. Skip those decisions and the usual failure is a missing step rather than a bad one: the draft that used to go past a subject expert now goes straight from the model to the editor, and nobody checks the eligibility figures before they're published.\n\nAI also makes it easier to expose workflow that was previously implicit. A vague process is hard to automate well, which is often another way of saying it was never clear enough for humans either.\n\n## Questions people ask\n\n**Does every team need a formal workflow?**\n\nNot formal in a heavy way, but most teams benefit from making the main stages and responsibilities explicit. If the work regularly gets stuck or rushed, workflow is usually part of the reason.\n\n**Is workflow just a project plan?**\n\nThey overlap. A project plan says what needs to happen once, and by when. Workflow describes the repeating process content follows every time.\n\n**Who should own the workflow?**\n\nSomeone with the authority to change it, often a [head of content](https://contentlexicon.com/head-of-content) or [content operations](https://contentlexicon.com/content-operations) lead, with input from the people who work inside it. Without an owner, it drifts back to whatever the most senior reviewer prefers.\n\n**Can workflow make content too rigid?**\n\nIt can if it's overdesigned. A good workflow provides enough structure to support quality and speed without forcing every piece of content through the same unnecessary ceremony.\n\n## Learn more\n\n- [Content workflow tips](https://dfedigital.blog.gov.uk/2024/11/27/content-workflow-tips) – the Department for Education on agreeing a workflow and allowing realistic time for each kind of content, from a page update to a training module.\n- [Building Effective AI Agents](https://www.anthropic.com/engineering/building-effective-agents) – Anthropic distinguishes predefined workflows from agents and shows patterns for routing, parallel work and iterative evaluation: technical reading that helps explain how a content process could change.\n- [Making it stick](https://voicetoneandstyle.com/learn/making-it-stick) – how to embed a style guide in real workflows, build support and keep it alive after launch." }, { "term": "Writer", "alsoCalled": [], "slug": "writer", "definition": "A role that researches and writes content for an organisation, shaped by its audience, voice and purpose.", "category": "Roles", "related": [ "content-brief", "content-design", "content-designer", "content-editor", "content-team", "copywriting", "editing", "subject-matter-expert", "ux-writer" ], "updated": "2026-10-11T06:25:00.000Z", "url": "https://contentlexicon.com/writer", "markdown": "https://contentlexicon.com/writer.md", "credit": "Writer, from the Content Lexicon by Contentious (https://contentlexicon.com/writer), CC BY 4.0", "body": "A writer in an organisation researches and writes material: features, guides, reports, newsletters, web pages and the many smaller pieces in between. Literary and other creative roles have different remits. A content writer usually works to a brief and within a house [voice](https://contentlexicon.com/voice), alongside editors who shape and check the work and subject experts who supply knowledge.\n\nThe job title covers a wide range of roles, from a generalist producing a charity's supporter communications to a specialist writing technical documentation or health information. Copywriter, [content designer](https://contentlexicon.com/content-designer), [UX writer](https://contentlexicon.com/ux-writer) and, increasingly, storyteller are neighbouring roles with different emphases, so describe the work you need rather than relying on the title.\n\n## Why an organisation needs a writer\n\nMost organisations are full of knowledge that nobody outside can use: research findings, policy detail, practitioner experience, stories from the people they serve. A writer turns that into material someone will read and act on. Asking experts to write it themselves rarely works at scale; they're busy, close to the subject and often unsure what a reader doesn't know. A good writer brings distance, asks the questions a reader would and gives an organisation a consistent voice across many pieces.\n\nThe need is clearest when volume grows beyond what staff can write alongside their main jobs, when a subject is complex enough to need translating or when an organisation wants original material – interviews, reporting, case studies – rather than rearranged information. A writer needs access to experts and sources, a clear brief and an editor. Without those, even an excellent one will produce polished guesswork.\n\n## How AI changes the role\n\nDrafting is no longer the scarce part of the job. When fluent text is cheap, a writer's value shifts towards what a model can't supply: interviews, original observation, judgement about what to leave out and responsibility for every claim. Ted Chiang argues that writing is the product of a great many small choices, and that a model left to make them for you settles for the most predictable ones. Writers who use AI well keep those choices for themselves, while handing assistants research tasks, transcription and variations.\n\nThe role is also spreading into writing for machines. Organisations need someone to write the examples, voice guidance and instructions that help assistants produce acceptable drafts, and writers who know an organisation's voice are often best at it. Job descriptions are starting to reflect this, asking for reporting, [editing](https://contentlexicon.com/editing) generated drafts and work across formats alongside writing itself, and for writers who can say where each sentence came from.\n\nSome organisations are going further. As generated text crowds every channel and newsrooms shrink, employers from Ford to the Mozilla Foundation have recruited former journalists into roles titled storyteller, editorial director or [managing editor](https://contentlexicon.com/senior-content-editor). What they want is reporting: finding stories in an organisation's own people, data and expertise, which a model can't reach and a competitor can't copy. The title is new, but the work is what good in-house writers have always done.\n\n## Questions people ask\n\n**How is a writer different from a copywriter?**\n\nContent writing often informs, explains or sustains a relationship with an audience. [Copywriting](https://contentlexicon.com/copywriting) more often seeks a specific persuasive action, such as a sign-up or a purchase, though plenty of people do both.\n\n**Does a writer need subject expertise?**\n\nEnough to ask good questions and represent the topic accurately, supported by reliable experts and sources. Specialist fields such as health, law and finance often justify hiring writers with a relevant background.\n\n**Should we hire a writer or use freelancers?**\n\nFreelancers suit variable volume and specialist one-off pieces. An in-house writer builds knowledge of the organisation, its experts and its voice over time, which pays off when the work is continuous or the subject is complex. Many teams combine a core of staff writers with trusted freelancers.\n\n**Where should a writer sit in the organisation?**\n\nClose to the people and information they write about, with an editor to work alongside. Writers placed inside a press or crisis communications team tend to get pulled towards reactive messaging, while those near subject experts and data find the stories that make the work distinctive.\n\n**What should a writer's portfolio show?**\n\nPieces written for real readers and purposes, ideally with a note on the brief and the writer's part in it. Evidence of research and interviewing tells you more than polish, now that polish is easy to generate.\n\n## Learn more\n\n- [Why A.I. Isn’t Going to Make Art](https://www.newyorker.com/culture/the-weekend-essay/why-ai-isnt-going-to-make-art) – Ted Chiang's 2024 New Yorker essay arguing that writing consists of countless choices and that language needs an intention to communicate; written about art, but pointed for anyone deciding what a writer is for.\n- [Storytellers wanted](https://www.quillmag.com/2026/04/27/storytellers-wanted) – Cory Schouten's Quill feature on organisations hiring former journalists to report and tell stories in-house, with sensible advice on what the role needs to succeed and why it shouldn't sit inside PR.\n- [Clean floors and great cooks](https://contentious.ltd/thoughts/clean-floors-and-great-cooks) – why removing the obvious AI tells is only hygiene, while craft and a distinctive identity are the harder jobs." } ] }