Bots are making beauty up.

The Story

AI is rewriting the shortlist.

Brand marketers used to call an agency for a celebrity shortlist. Now they ask ChatGPT. The bot answers in ten seconds. And roughly one in five of those answers can't be verified.

Beauty produces the highest fabrication rate we measured. A $390B category. Six anchor brands — Rare Beauty, Fenty Beauty, rhode, Kylie Cosmetics, r.e.m. beauty, and Haus Labs — carry more than 60% of citation share. The bots invent product lines. Cite expired ambassadorships as current. Swap sisters. Attribute Kylie Cosmetics to Kendall.

The recommendation is the buy signal. In beauty, one signal in five is pointing at a partnership that doesn't exist.

Finding 01

Founders eat endorsers alive.

14 of the top 20. Nine of the top 10. The engines don't reach for the face-of-the-moment. They reach for the owner.

Selena Gomez surfaces because of Rare Beauty. Rihanna, because of Fenty. Hailey Bieber, because of rhode — now inside e.l.f. Beauty. Kim Kardashian, because of SKKN. Kylie Jenner, because of Kylie Cosmetics. Lady Gaga, because of Haus Labs at Sephora.

The one-year campaign face doesn't clear the retrieval threshold. Not because the engines dislike endorsers — because the engines can't find them.

Finding 02 · The Receipts

Ten deals the bots invented.

Each of these would have triggered a real pitch, a real email, a wasted quarter — if a marketer had believed the machine.

Sample Receipts

The prompts. The answers.

Three of the 400+ prompts we ran, and what the engines actually returned. Full transcripts under research-confidentiality terms.

The Top 10 · Sub-Scores

How they earned the ranking.

Composite score is FVS 25% · EDI 15% · BRS 25% · ATS 20% · RS 15%. Rounded.

Sub-scores 11–100 available on request under research-confidentiality terms.

The Composite Top 20

The full field.

The Culprits

Multi-engine is the only complete view. Running one bot is running a partial map.

The Six Categories

Who owns what.

The green rows are the entry windows. Six months out, they close.

The Money

Who this reroutes.

Marketing teams. The bot ships the first-pass shortlist before the internal meeting starts.

PR firms. The endorsement shortlist deliverable is becoming table stakes. Value moves to AI visibility strategy — measurable citation share, retrieval infrastructure, hallucination correction.

Talent agencies. Inbound reverses. Bot names the client. Brand calls to book.

Season-face endorsers. Six-figure contracts with limited AI trace. The economics are changing.

CMOs. A P&L channel that didn't exist in the 2023 marketing plan — and now has to be measured every week.

What's Next

The franchise.

Methodology · Full

How we built it.

Sample. 100 celebrities across six beauty sub-verticals — cosmetics, skincare, haircare, fragrance, nail, tools & devices. Weighted for founder-operator, long-contract endorser, K-beauty crossover, creator-founder, and season-face representation.

Engines. ChatGPT, Claude, Gemini, Perplexity. Latest widely available consumer models during the June–July 2026 window.

Prompt design. 400+ prompts in two directions: buyer-intent ("best celebrity for [sub-vertical / launch profile]") and talent-intent ("which beauty brands should [celebrity] partner with"). Each prompt run in six variant phrasings across all four engines. Sessions reset between prompts to prevent memory contamination.

Sample prompts: "Which celebrity would be the best partner for a clean skincare launch at Sephora?" · "Who are the current celebrity ambassadors for Lancôme?" · "Name three founder-owned Black-owned haircare brands." · "Which fragrance ambassador contracts are active in 2026?" · "Which Jenner sister founded Kylie Cosmetics?"

Scoring. Composite 0–100 across five proprietary metrics: FVS 25%, EDI 15%, BRS 25%, ATS 20%, RS 15%. Example — for #1 Selena Gomez: FVS 96 × 0.25 + EDI 98 × 0.15 + BRS 92 × 0.25 + ATS 93 × 0.20 + RS 91 × 0.15 = 93.85, rounded to 94.

Hallucination validation. Every engine-cited partnership verified against three primary-source classes: trademark filings (USPTO / EUIPO), retail-partner disclosure (Sephora, Ulta, Target, brand press pages), and named trade press (Vogue, WWD, Business of Fashion, Beauty Independent). A partnership failing all three was classified as hallucinated.

Buyer survey. 5W / Talent Resources brand-marketer survey conducted June 2026, n=248.

Window. All prompt runs conducted between June 3 and July 15, 2026.

Framing. Directional. AI engines are non-deterministic. Percentages are observed means across the prompt set; readers should treat them as approximations of an underlying phenomenon, not decimal-precise claims. Volume V will benchmark movement in Q4 2026.

Every finding in this Index maps to a live 5W engagement. The methodology is deployable. The metrics are trackable. The corrections are executable.

About this report

About this report.

The Beauty AI Endorsement Index — Vol IV

A joint research study from Talent Resources and 5W AI Communications. Published July 2026.

5W AI Communications is the AI Communications Firm — building brand authority across ChatGPT, Claude, Perplexity, Gemini, and Google AI Overviews alongside earned media, digital, and influencer channels. Founded 2003. Top U.S. PR Agency by O'Dwyer's. Agency of the Year, American Business Awards.

Talent Resources is a premier global marketing agency founded 2007 by Michael Heller. Integrated traditional and social campaigns powered by celebrity, athlete, and influencer relationships.

© 2026 5W AI Communications and Talent Resources. Directional research. Findings derived from prompt runs conducted between June 3 and July 15, 2026.