A lot of skincare marketing happens on TikTok. mapou checked whether it carries over when a shopper asks an AI assistant instead.
mapou ran 20 full skincare shopping conversations, 5 each with ChatGPT, Gemini, Claude and Grok. Each simulated shopper kept asking until they decided what to buy. Then mapou tagged every reason the assistants gave for recommending a brand: 248 in all.
"Viral on TikTok" came up 2 times, in 2 of the 20 conversations. Ingredients came up 55 times.
The numbers
248 reasons across 20 conversations, 36 assistant turns, four AI engines, September 2026.
Credibility wins. Ingredients, clinical results and dermatologist recommendations together make up 39% of the reasons in conversations.
K-beauty is the one trend that shows up. It's 14% of conversation reasons, mostly because one of the five shoppers was curious about Korean skincare and asked about it directly. TikTok, celebrities and influencers barely register.
The same pattern in single answers
mapou also asked 90 skincare questions once each, 360 answers in all. The result is the same at a much larger scale: TikTok was 0.9% of 3,234 reasons, and ingredients were 28%.
3,234 reasons across 360 answers, four AI engines, September 2026.
Who gets recommended, and why
| Brand | Share of answers | Top reasons |
|---|---|---|
| SkinCeuticals | 28.3% | ingredients, clinical results, luxury feel |
| La Roche-Posay | 22.2% | ingredients, dermatologist-recommended, easy to find in stores |
| CeraVe | 18.9% | easy to find in stores, barrier-safe, ingredients |
| Paula's Choice | 11.1% | ingredients, clinical results, dermatologist-recommended |
| The Ordinary | 10.6% | ingredients, budget-friendly, easy to find in stores |
| Sunday Riley | 8.9% | ingredients, clinical results, luxury feel |
Share is the share of single answers naming the brand. Reasons are the three AI gave most often for that brand.
The brands at the top are the ones with an easy story to repeat. SkinCeuticals has its ingredients and clinical results. CeraVe is easy to find and gentle on the skin barrier. The Ordinary is ingredients at a low price.
What it means
mapou's read: an AI assistant can't feel a trend. It can repeat a claim it finds written down, like an active ingredient, a percentage, a study or a dermatologist's view. A brand that grew on TikTok still needs that story on pages AI reads, or the virality doesn't travel into the conversation.
What this doesn't show
This shows what AI says, not whether TikTok drives sales. It also doesn't mean AI never reads creator content. In September the shared search layer mapou uses ranked YouTube videos much lower for standard buying questions, and Who does AI read? covers that change.
How we measured this
- Conversations
- 20 synthetic shopper conversations, September 2026: 5 skincare shopper personas, each run once against ChatGPT, Gemini, Claude and Grok. Each persona keeps asking follow-up questions until it decides.
- Single answers
- 90 skincare questions asked once of each engine, 360 answers, as a second check.
- Engines
- ChatGPT (gpt-5.4-mini), Gemini, Claude and Grok, all through one shared search layer with web search on, so differences reflect the models rather than their search engines.
- Reasons
- Every brand recommendation was tagged with the reasons the answer gave for it (ingredients, price, TikTok and so on) by an LLM classifier, the same one used across mapou studies.
- Limits
- Small samples: 20 conversations and 248 reasons. One persona is K-beauty curious, which lifts the K-beauty share. This measures what AI says, not what drives sales.
Cite this article
mapou, “Going viral on TikTok almost never gets a skincare brand recommended by AI,” mapou.ai, October 4, 2026. https://mapou.ai/articles/tiktok-doesnt-sell-skincare-to-ai
Reporters can reach mapou through for more on the method, other breakdowns or a comment.
Why does AI recommend your brand, or not?
mapou runs full shopping conversations about your category every month and tracks the reasons each AI assistant gives for your brand and your competitors.
Monthly research newsletter
Get the next article by email.
Short findings like this one, plus the monthly panel report. Unsubscribe in one click.
We never share email addresses. One unsubscribe link in every send.