New holiday season data tracking ads across Google AI Overviews, AI Mode and ChatGPT has surfaced a detail that matters well beyond the paid search team. In beauty and travel, most ChatGPT ads ran on research prompts - not transactional ones. That single finding reframes where organic AI visibility actually needs to work hardest.
The report covered five shopping categories and mapped ad placements to prompt types. The pattern in beauty and travel is notable precisely because those two sectors attract high-consideration buyers who browse, compare and research before they convert. The AI platforms know this. And now the ad formats are being designed around it.
Research Prompts Are Where AI Search Operates at Scale
When advertisers choose to show ads on research-stage prompts inside ChatGPT, they are signalling something about user behaviour: this is where the audience is spending time and attention. Paid teams do not willingly burn budget on prompt types that do not perform. The fact that research prompts dominate ad placement in these categories tells you the audience is active and engaged at that stage.
For organic AI visibility, this is the strategic implication. If paid is competing on research prompts, organic content needs to be there too - and it needs to be structured so AI systems can extract and cite it. A brand that only has strong transactional content, good product pages and a clean checkout flow, but thin informational content, will miss the stage where AI search is doing its heaviest lifting.
Beauty and travel are categories where research intent is long and layered. A user asking ChatGPT which SPF moisturiser suits combination skin, or what to pack for a city break in October, is not ready to buy - but they are forming preferences. That is the moment a cited brand gets embedded in the decision. Earning that citation organically is cheaper than buying it, and it compounds over time.
What Research-Stage Content Needs to Look Like for AI Extraction
AI systems pulling citations from the web at research prompts are looking for content that directly answers a question, attributes a clear position and is structurally easy to extract. That means well-defined headings, concise answer paragraphs, and content that does not bury the useful information inside marketing copy. A 2,000-word brand story with the product recommendation in paragraph fourteen is not what gets cited.
For beauty brands specifically, this means ingredient explainers, skin type guides and routine recommendations written for clarity rather than conversion. For travel, it means destination guides, packing lists and practical comparison content - the kind of material that answers the exact prompts users are typing into ChatGPT. Schema helps signal the topic and structure, but the underlying content still has to be genuinely useful.
One structural point worth building in: FAQ sections on category and product pages, written to match the actual research questions users raise, give AI systems clean, attributable answer units. These are easier to extract than flowing prose. If you want to test whether your content is extraction-ready, paste the page into a prompt and ask the AI to summarise the brand's position on a specific topic. If it cannot do it cleanly, your content is not ready.
The Prompt-Type Gap Most Brands Are Not Measuring
Most brands measuring AI search visibility are doing so at the brand name level - they check whether their name appears in a response to a branded query. That is a reasonable starting point, but it misses the more valuable question: does the brand appear in category and research prompts, where AI search is influencing the most users?
The data from this report frames that gap clearly. If paid ads are appearing on research prompts in beauty and travel, there is a volume and engagement argument being made by advertisers. Organic programmes should be auditing against equivalent prompt types - category research questions, product comparison prompts, occasion-based queries - and checking whether they earn citations there. Tools like Perplexity allow you to run structured prompt tests manually. AI Mode in Google can be sampled similarly. Neither gives you scale, but they give you directional signals.
The more rigorous approach is to build a prompt set that maps to your buyer journey, run it regularly, and track citation presence over time. That is not a one-off exercise - it is a programme, the same way rank tracking was for organic search. The difference is that prompt-level visibility tells you something about where in the decision journey your brand is appearing, not just whether it appears.
What Running Ads on Research Prompts Changes for Attribution
If paid ads are appearing on research prompts, the conversion path from that ad to a sale is longer. A user who clicks an ad while researching SPF moisturisers is probably not buying that session. This creates attribution complexity that mirrors what organic AI visibility teams already face: the AI surface influences the decision, but the eventual conversion happens elsewhere, later, often through a different channel.
For teams trying to measure AI-referred revenue in a CRM, this reinforces the case for UTM tagging on AI referral traffic, session stitching across visits, and lead source attribution that captures the first meaningful touchpoint. If a prospect first encounters your brand in a ChatGPT research response, clicks through, and converts three days later via a branded search, the AI touchpoint needs to be in the record. Without it, the AI visibility programme looks like it has no commercial impact - even when it has driven the initial preference.
The paid and organic AI visibility programmes also need to talk to each other here. If paid is buying research prompts and organic is earning citations on research prompts, they may be reaching the same user at the same stage. Understanding where each appears - and whether they reinforce or overlap - requires shared prompt-level reporting, not just channel-level dashboards.
The Broader Signal Across Categories
The report covered five shopping categories, with beauty and travel showing the strongest concentration of ads on research prompts. That variation across categories matters. It suggests that prompt-type distribution is not uniform - it reflects category-specific buyer behaviour. Apparel and electronics may show different patterns, potentially skewing more towards comparison or transactional prompts.
For brands operating across multiple categories, this means prompt-type strategy cannot be applied uniformly. A beauty brand's AEO content programme should prioritise research and education content. An electronics brand may need to weight comparison and specification content more heavily. The underlying discipline is the same - structured, extractable content that earns citations at the right prompt type - but the content shape varies by category.
The holiday season timing of this data also signals something worth tracking forward. Seasonal uplift in AI search activity, particularly in research-heavy categories, may create cyclical windows where AI visibility matters more acutely. Brands that build their content and citation programmes outside of peak periods will be better positioned when query volumes spike. That is not a new lesson from search, but it is one that applies directly to how AI search programmes should be planned and resourced.