The data desk for cosmetics AI search. We publish what the numbers say about how ChatGPT, Gemini, Perplexity and Claude decide which beauty brands and manufacturers to name — and what to do about it. Every figure is sourced; every claim is dated. Measurement before opinion.
ChatGPT drives ~60% of AI search, cites 22+ sources per answer, reads owned sites deepest — and has a long memory for domains. Includes live data from our own factory test.
Read the breakdown →Perplexity overlaps Google's top 10 more than any AI (28.6%), leans on editorial "best of" lists, and rotates citations fast — we watched it drop a brand from 100% to 0% in 48 hours.
Read the breakdown →Gemini rewards Google SEO and crawls owned sites deepest — but it invents brand domains. In our test, 38% of its domain mentions were hallucinated. How to win it safely.
Read the breakdown →Claude leans on B2B aggregators (accio.com appeared 104 times in 30 runs), over-weights medical sources, and reuses authority domains across languages. How to earn its trust.
Read the breakdown →36.7% of all beauty AI citations flow to just ten domains — and not one is a brand's own website. A data map of who AI names, why, and what actually correlates with visibility (it isn't your ad budget).
Read the report →AI now answers "what should I buy" directly. Getting named is a discipline, not luck. What works, what's a waste of time (llms.txt, keyword stuffing), and how to measure it without fooling yourself.
Read the playbook →