A card sort shows how people think about information: what they expect to find together and what language they naturally use for it.
In content and user experience (UX) work, card sorting is often used to shape navigation, information architecture, 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.
Why it matters for content
Card 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.
The 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.
Card sorting is also useful in voice and 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.
How AI changes the picture
AI can make card sorting faster to prepare and easier to analyse. It can turn a 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.
But 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.
The outputs travel further now, too. The groupings and labels a card sort produces can feed content models, metadata, search and AI retrieval as well as navigation, so a good sort can help people and machines find the same things.
Questions people ask
- Is card sorting only for navigation?
Navigation 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.
- What’s the difference between open and closed card sorting?
In 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.
- How many people do you need?
Fewer 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.
- Does card sorting give you the final structure?
It 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”.
Learn more
- Card Sorting: Uncover Users’ Mental Models – Nielsen Norman Group guidance on running and interpreting card sorts.
- Card Sorting: A Definitive Guide – Donna Spencer and Todd Warfel’s guide to preparing, running and analysing a sort.
- Tree Testing: Evaluate Menu Labels and Categories – a useful companion method for testing whether a proposed structure works.