.jpg)
The AI Beauty Conversion Gap: What 49 Consumer Interviews Revealed
Beauty's biggest retailers are racing to embed themselves inside ChatGPT, Gemini and Google's AI agents. They are mostly going to lose money on it. Our recent qualitative study of 49 French consumers, run with Dialog and Oh My Cream, identified an AI beauty conversion gap: conversational AI now precedes most beauty decisions and closes almost none of them. The gap has a clear shape and five specific causes.
The race for the AI shopping moment
Beauty's biggest retailers spent the first half of 2026 wiring themselves into conversational AI surfaces. Sephora launched a native app inside ChatGPT, with payment and checkout planned for the next release. Ulta rolled agentic commerce into Google's AI surfaces. Fenty Beauty built an AI adviser on WhatsApp. K-beauty giant Amorepacific opened its Amore Mall inside ChatGPT. PYMNTS calls it the race to own the AI shopping moment.
The assumption underneath the race is simple. Get in front of shoppers when they ask an LLM what to buy, and conversion will follow. Our study suggests that assumption is wrong, or at least very incomplete.
What the consumers actually do
We ran 49 in-depth interviews with French beauty consumers in early 2026 using Verso's adaptive AI interviewer, which probes contradictions in real time the way a senior moderator does. The pattern was consistent across every demographic and category we covered.
Conversational AI has replaced Google as the first reflex for beauty research. Consumers describe it as faster and more synthesized. They use it for routine building, ingredient decoding (INCI lists, comedogenicity, formulation comparison), and product shortlisting. A growing use case is "cosmetic literacy": consumers now arrive at the store or product page already knowing more about the formulation than the staff.
Then they leave. Almost none of our 49 participants purchased on the strength of an AI recommendation alone. The verification step is structural and near-universal.
The post-AI verification loop
Every AI recommendation triggers the same downstream pathway.
- Read user reviews. The first reflex, almost reflexive.
- Search TikTok and Instagram for real people using the product, looking for texture, finish, application, results on real skin.
- Where possible, test in store. For products where scent or texture matter (fragrance, foundation, cleansers), this step is non-negotiable.
- Buy, or abandon.
A brand with thin reviews, no creator presence, or weak in-store distribution sees every AI recommendation evaporate between step 1 and step 3. The verification loop is not a quirk of cautious shoppers. It is the new structure of the journey.
The 5 trust barriers
What this means for CMI and brand teams
Three things follow for insights and brand leaders working in beauty right now.
One. Algorithmic visibility is necessary, but not enough. Being cited by ChatGPT, Gemini, or Perplexity is the new shelf placement. It opens the funnel. It does not move money on its own. Plan the visibility work, then plan past it.
Two. Reviews, creator content, and in-store presence are now AI-conversion infrastructure. They are the verification layer your AI recommendations have to pass through. A strong ChatGPT showing combined with thin review pages and no TikTok presence will underperform a mediocre ChatGPT showing combined with rich social proof.
Three. The biggest conversion lift is on your own product pages, not in someone else's chat surface. The verification loop happens because consumers have to leave the AI to check things. If they didn't have to leave, the loop collapses. McKinsey reports global retailers seeing conversion lifts of up to 20% from AI shopping assistants embedded at the point of purchase.
How to close the AI beauty conversion gap
Oh My Cream, the French beauty retailer that partnered on this study, has been testing this thesis with Dialog. Dialog embeds an AI assistant directly on the product page. A shopper can ask, in plain language, whether a product is comedogenic for their skin type, what real users say about its long-wear, how it compares to a similar item, what the active concentration is, whether it is in stock at their nearest store. The answers come back grounded in real product data and real reviews, on the page where the buy button lives.
The verification loop, in this configuration, does not need to happen elsewhere. Discovery, verification, and purchase converge on a single screen.
We're not arguing this is the only model. It's one of the few that maps onto what consumers actually told us they wanted.
How we ran the study (a note on method)
Two notes for fellow CMI teams. The full methodology is in the downloadable report.
We ran 49 in-depth interviews with French beauty consumers between January and March 2026 using Verso's adaptive AI interviewer. Each conversation probed the contradictions the participant produced, in real time, the way a senior qualitative researcher would. That's how the five barriers surfaced as named, distinct frictions rather than a vague "trust" theme.
We delivered the analysis in under a week from fieldwork close. Speed wasn't the point. The point was that this format made the study runnable at all. Eight weeks of traditional qual on a moving market story would have shipped findings that were already outdated.
Where to take this next
Talk to the Verso team about running a study like this for your category
Skincare, fragrance, color, dermo-cosmetic, men's grooming, premium and mass. We can ship one in days.
Methodology: 49 in‑depth qualitative interviews with French beauty consumers, conducted by Verso (AI‑moderated video interviews, hybrid synthesis) in partnership with Dialog. Full deck available on request
Download an extract from our report.
Frequently asked questions
Does AI actually drive any beauty sales today?
Yes, but rarely directly. Our study found that AI mostly shapes the consideration set. Purchases overwhelmingly happen after consumers leave the AI conversation to verify on reviews, social, and in store. The exception is when AI is integrated on the product page itself, where the verification loop can close in one place.
Which beauty brands are most cited by AI assistants?
In our French sample, Sephora was the most-referenced retailer (35% of interviews). Pharmacy and dermo brands (La Roche-Posay, CeraVe, Avene, Bioderma, SVR) dominated the brand category. External research from Yotpo shows La Roche-Posay leads US AI recommendations at 22%, with CeraVe at 20%.
What should a beauty brand do first to fix this?
Audit the verification step before the visibility step. Review density, recent creator presence, in-store availability. If those are weak, AI visibility will leak. Then layer in AI work: ingredient transparency and dermatologist validation for GEO, and a product-page AI layer that closes the loop in one place.
Is this only true in France?
Our 49 interviews were all French. The behaviors map closely to US data on Gen Z AI discovery and conversion lifts from AI shopping assistants. The verification loop appears to be structural, not local.
Will ChatGPT eventually own the entire purchase journey?
Possibly, for some categories. Beauty has a sensory and trust profile that makes the verification step hard to remove entirely. Brands that win in the meantime will be the ones that move the verification work onto their own surfaces rather than waiting for ChatGPT to absorb it.
Why did you partner with Oh My Cream and Dialog?
Oh My Cream is a French retailer working on AI-assisted product discovery in their stack. Dialog provides the AI product-page layer. The partnership grounded the study in live consumer behaviour, not hypothetical scenarios.
Can I get the raw interview data?
The downloadable report includes selected verbatim quotes and the methodology. Full raw data and interactive analysis are available to clients who run their own studies with Verso.