Key findings
55% of consumers say AI recommendations at least sometimes influence household and grocery purchases. NielsenIQ, 31 countries.
AI referral traffic to beauty sites grew 312.5% in a year, the fastest of any industry. Similarweb.
54.7% of the sources ChatGPT cites in beauty conversations are retail and e-commerce domains. Similarweb.
How much do AI recommendations influence what people buy?
More than half of consumers say they do. NielsenIQ surveyed more than 21,000 consumers in 31 countries for its 2027 outlook, published in October 2026. 55% said AI recommendations at least sometimes influence their household and grocery purchases. In the US, NielsenIQ found that 51% had used at least one AI tool to support their shopping in the past month, a definition that includes retailers’ own recommendation features.
The numbers vary with the question asked. Capgemini found that a quarter of consumers had used generative AI shopping tools in 2025 and a further 31% planned to. Kantar’s AI barometer in the UK, as reported by Research Live, found that 15% of consumers would assume a brand was not for them if AI did not suggest it when prompted. Whichever figure you take, a brand that is missing from the answer is missing from a real share of decisions.
Which categories are moving fastest?
Beauty and fashion. Similarweb measured referral traffic from AI platforms between June 2025 and May 2026. Beauty was the fastest-growing industry, up 312.5% year over year, with fashion close behind at 278.2%. These are categories where people describe themselves and ask what suits them: skin type, fit, budget, occasion.
That is a different question from a search query. Someone who types a product name already knows the brand. Someone who asks which moisturiser is best for sensitive skin under a certain price is asking for advice, and the answer usually contains one or two names.
Where do the recommendations come from?
Mostly from pages the brand does not own. In Similarweb’s data for May 2026, 54.7% of the sources ChatGPT cited in beauty conversations were retail and e-commerce domains. Retailer listings, editorial roundups, Reddit threads and video reviews sit alongside the brand’s own site.
Retailer listings. The same product described again, often from an older data sheet.
Editorial reviews and roundups. The best-of pieces engines retrieve for recommendation questions.
Reddit, YouTube and TikTok. Customers describing what the product is like in use.
Ingredient and safety databases. Third-party records an engine can use to check a claim.
When an ingredient list, a certification or a size differs between these sources, the engine has a reason to name a brand whose facts line up.
What to do first, and sources
List the questions your customers ask by need, not by product name. Run them on ChatGPT, Gemini and Google and note which brands are named and which pages are cited. Then make the product facts identical on your site and at every retailer, and point reviews and earned coverage at the publications the engines already retrieve. How SourceNode does this for consumer brands.
The researcher
Odin Lund is co-founder and COO of SourceNode and leads its data analytics and research, including the daily tracking behind this work across ChatGPT, Gemini and Google AI Overviews. Odin founded First Query, the AI search practice that became SourceNode, and co-founded The Pixel Paper.
Meet the research team
About this research
Published by SourceNode
SourceNode is a data and research-first AI search firm. We record what AI models say about a business every day, trace each claim to its source, and use that data to get businesses named in AI answers. Every figure we publish carries its method and its sample.
Share Blog
Related research
AI search
AI search statistics for 2026: the numbers we trust, with sources
Read the research
AI search
What McKinsey’s AI search research says, and what to do about it
Read the research
