Category Framing: Why the Words Customers Search With Decide Your AI Visibility

A new study challenges a core assumption of entity SEO in the AI search era: being a well-known brand does not automatically make you a brand AI recommends. Which brand a large language model surfaces depends on the exact category language customers use — and how tightly that brand is tied to that language in third-party content.

Same Brand, Opposite Results Depending on the Query

Researchers Maryanna Franco and João da Silva tested 12 sports apparel brands across ChatGPT, Gemini, Perplexity, Claude, and Google AI Overviews, changing only one variable: the word used for the category. They compared results for "athleisure" versus "athletic footwear."

The swings were dramatic. New Balance was recommended just 1% of the time under "athleisure" but jumped to 90% under "athletic footwear" — an 89-point swing. Lululemon showed the opposite pattern: dominant at 90% for "athleisure," invisible at 0% for "athletic footwear." Nike held strong visibility across both — 77% and 90% respectively — effectively owning both categories.

The Two Forces Behind Category Framing

The researchers identify two drivers. The first is the knowledge graph description — the formal category a brand is assigned, which anchors brand recognition. The second, and more decisive, is the third-party content corpus: the accumulated body of articles, reviews, and editorial coverage. Updating your knowledge graph description alone won't move recommendations if, as Franco puts it, "the corpus is still saying what it always said." A brand's self-definition matters less than whether the wider content ecosystem has linked it to a given category language.

A Practical Audit for Marketers

The researchers recommend cross-testing 5–6 different category phrasings across multiple LLMs to see which ones surface your brand. Then concentrate third-party content investment in the media covering the category language where you're currently invisible, and actively participate in the comparison and review content that defines category membership.

What This Means for Marketers

This research shifts the center of gravity for Answer Engine Optimization (AEO). Many brands have focused on polishing their own site and knowledge graph entries, but the language that actually decides AI recommendations sits outside a brand's control. Keyword research needs to expand into AI prompt research — mapping the category language customers actually use with AI, then building PR and content strategy around the outlets that dominate that language. Brand visibility in the AI era is a fight over the words the world uses to categorize you, not just who you say you are.

Building topical authority for AI search touches similar ground — see our analysis of topic authority in ChatGPT category share. If you want help auditing your own AI visibility, Best Partner's services can help.