OpenAI has been spotted testing editorial labels on ChatGPT product results. Not categories. Not filters. Labels like "best all-rounder", "best for beginners", and "best cushioning" - applied by the model itself to the products it surfaces. That is a meaningful shift in how AI search presents commercial information, and it has direct implications for any brand trying to earn visibility in these results.
Up to now, most AEO work has focused on getting cited at all. Appear in the result, get the mention, earn the referral. That is still valuable. But label-based results introduce a second question that is harder to answer: not just whether your product appears, but what the model thinks it is best at.
The Model Is Making Editorial Judgements
When ChatGPT appends a label to a product result, it is not pulling that label from a feed or a structured tag you submitted. It is synthesising a characterisation from whatever signals it has access to - your product pages, review content, third-party editorial coverage, forum discussions, and the broader corpus it was trained on. The label is the model's interpretation of your product's positioning, not yours.
This matters because the label shapes click behaviour. A user comparing running shoes who sees one product labelled "best cushioning" and another labelled "best for speed" will self-select based on their need. If the label ChatGPT assigns to your product does not match how your target customer thinks about their problem, you may earn a citation and still lose the sale.
The practical consequence is that AEO strategy can no longer stop at visibility. Brands now need to think about what characterisation they are likely to receive - and whether the signals they publish actually support that.
What Signals Shape the Label a Model Assigns
AI models do not assign labels arbitrarily. They draw on patterns across multiple content types. Product page copy that consistently frames a product around a specific use case gives the model a clear signal. Review aggregations that repeatedly mention the same attribute - cushioning, ease of setup, durability - reinforce that framing. Third-party editorial content that positions your product in a particular category adds weight. The model is looking for coherence across sources.
Where brands tend to go wrong is inconsistency. Marketing copy says one thing, review content says another, and editorial coverage positions the product somewhere else entirely. A model trying to synthesise a label from that noise will either assign a vague one or default to whatever the dominant signal happens to be - which may not be the one you want.
Schema markup plays a supporting role here. Product schema with well-populated fields - including description, review aggregates, and audience targeting where relevant - gives the model structured data to draw on rather than having to infer everything from prose. It does not guarantee a specific label, but it reduces ambiguity in how your product is characterised.
Positioning Clarity Is Now a Technical Requirement
The discipline that matters most here is not new - it is simply now enforceable by AI. If your product positioning is fuzzy internally, that fuzziness will surface in AI-assigned labels. Brands that have invested in clear, differentiated positioning - and published that positioning consistently across owned and earned content - are better placed to influence how AI characterises them.
For practitioners, this means auditing the signal landscape around your product. What does your product page actually emphasise? What attributes show up most frequently in your review corpus? What do external editorial sources say about your category positioning? If those three sources are pointing in different directions, you have a GEO problem - not a PR problem or a content problem in isolation.
Digital PR becomes especially relevant here. A review site article that consistently describes your product as "ideal for complete beginners" contributes to the pattern the model draws on. That is not a secondary benefit of PR - it is a primary one, in the context of AI label assignment.
What This Changes About Measuring AI Visibility
Tracking whether you appear in ChatGPT results is already non-trivial. Tracking what label you appear with is a further step again. At the moment, the tooling for this is thin. But the principle is clear: if label-based product results drive click behaviour in the way that comparable features do elsewhere, then the label you receive is a conversion variable - not just a visibility variable.
For brands running AI search measurement programmes, the question to add to your monitoring is not just "did we appear?" but "how were we characterised?". Prompt-based monitoring - where you systematically query ChatGPT with purchase-intent questions and record both presence and label - is a practical way to build that picture. It is manual and imperfect, but it gives you something to work with before more formal tooling exists.
Where this gets commercially interesting is in connecting label data to conversion outcomes. If you can see that users referred from ChatGPT convert differently depending on the query context, you can start to infer which characterisations are driving qualified traffic versus unqualified interest. That requires CRM tagging of AI-referred sessions and the discipline to annotate query intent at the point of measurement - but the data is there if you build for it.
The Broader Direction of Travel
ChatGPT is not the only platform moving in this direction. Perplexity already layers editorial framing around product recommendations. Google AI Overviews routinely characterise products and services in summarised prose before linking to sources. The pattern is consistent: AI is not presenting raw lists, it is presenting curated assessments. The label test at OpenAI is simply making that editorial function more explicit.
For brands, the implication is that AI search is increasingly behaving like a trusted advisor rather than a neutral directory. Advisors form opinions. Those opinions are shaped by the information they have access to. The brands that invest in publishing clear, consistent, and attributable signals about what they are genuinely best at will be better positioned to influence those opinions than brands that rely on broad claims and hope for the best.
The work is not glamorous - it is positioning discipline, content coherence, and signal auditing. But that is what separates brands that earn useful AI citations from brands that earn generic ones, or none at all.