How to Get ChatGPT to Recommend Your Beauty Brand

A customer wants a vitamin C serum. In 2019 she'd have googled it. In 2023 she'd have opened the Sephora app or watched a TikTok.

In 2026 she asks ChatGPT.

And the brands it names — usually three or four — become her shortlist before she visits a single website. Everyone else is out of the running before the race starts.

Here's the part most beauty marketers get wrong about fixing that.

The short version

96.1% of AI beauty citations come from third-party sources. Your website is 3.9% of the picture.
Editorial coverage alone accounts for 63.2% of citations. PR beats content marketing here.
Ingredient-led indies earn ~31% of citations. Legacy mass brands earn ~4%. Size doesn't help.
The Ordinary leads ChatGPT citation share — because its product names are ingredient names.
Brand mention rates are falling, which means AI is spreading recommendations wider. That's your window.

The Core FindingYour Website Barely Matters Here

This is the most lopsided data I've seen in any industry.

Avenue Z's 2026 AI Visibility Index Beauty Report examined how ChatGPT recommends beauty brands. It covered 60 brands, 565 AI citations, and 75 consumer prompts across cosmetics, haircare and skincare.

The split:

Where AI beauty citations come from	Share
Third-party sources (editorial, media, reference sites)	96.1%
— of which, editorial publications alone	63.2%
Brand-owned content (your website, your blog, your product pages)	3.9%

Read that again if you've just signed off a content marketing budget.

Avenue Z put it bluntly: AI visibility in beauty isn't a content volume game, it's a credibility game. You can publish 200 blog posts about vitamin C. If Allure, Byrdie and a few dermatologists haven't mentioned you, ChatGPT has very little reason to name you.

In beauty, what other people say about you outweighs what you say about yourself by roughly 25 to 1.

Who's WinningAnd Why It Isn't Who You'd Guess

eMarketer runs an AI Visibility Index based on more than 5,200 ChatGPT responses across nine personal care and beauty categories. Some findings from 2026:

La Roche-Posay was recommended in 81% of facial skincare queries in Q1 — the highest rate of any brand in any category tracked.
CeraVe sat just behind it, with the two trading places month to month.
In makeup, there was no clear standout. Maybelline led lip and eye, but didn't crack the top five in face makeup.
Maybelline appeared in 51% of lip makeup queries versus 36% for Dior — the widest category gap recorded.

Now the more interesting cut. 5WPR analysed citation share across 80+ beauty prompts on ChatGPT, Claude, Perplexity, Gemini and Google AI Overviews:

Sephora-house brands: ~38% of all beauty AI citations
Ingredient-led independents (The Ordinary, Drunk Elephant, Glow Recipe): ~31%
Legacy mass brands: ~4%

Scrappy independents are beating enormous legacy brands by roughly eight to one. Marketing budget isn't the variable.

The Ordinary LessonName the Ingredient, Win the Citation

5WPR found The Ordinary holds around 7.0% citation share on ChatGPT — the highest of any single beauty brand.

Their explanation is one sentence, and it's the most actionable thing in the whole dataset. It's driven by ingredient-name product transparency that AI engines parse as structured data.

Think about what The Ordinary actually calls its products:

Niacinamide 10% + Zinc 1%
Hyaluronic Acid 2% + B5
Glycolic Acid 7% Toning Solution

Now think about how most beauty products are named. "Radiance Renewal Elixir." "Midnight Recovery Concentrate." Beautiful names. Completely opaque to a machine matching a query about niacinamide.

When someone asks ChatGPT for "a niacinamide serum for oily skin," The Ordinary's product name is the answer. No inference required.

You don't have to rename your products Most brands can't rebrand a hero SKU around an INCI name, and shouldn't. But you can make the ingredient prominent everywhere else — in the H1 subtitle, the first line of the description, the structured data, the alt text, and the editorial pitch. The goal is that an AI never has to guess what's in the bottle.

Two SystemsWhy Perplexity and ChatGPT Recommend Differently

Beauty AI visibility runs on two separate layers, and they reward different things.

The training layer — what ChatGPT and Claude absorbed during training — favours editorial coverage, dermatologist endorsement, and years of accumulated Reddit and community consensus. It's slow to build and slow to lose.

The retrieval layer — what Perplexity and Google AI Overviews pull from live search — rewards retailer curation, recent press and creator cycles. It moves fast.

Practically: a new launch can break into Perplexity within weeks on the back of good press, while ChatGPT's default recommendations take considerably longer to shift. Charlotte Tilbury was noted as the only brand holding a top-five position on both Perplexity and Google AI Overviews at once — a genuinely hard thing to do.

The PlaybookSix Things That Actually Move the Needle

Step 01

Put real budget into editorial PR

Editorial publications are 63.2% of beauty AI citations. This is not a nice-to-have.

The targets that matter are the ones AI engines already quote in your category — Allure, Byrdie, Refinery29, Cosmopolitan, Who What Wear, plus the dermatologist-led and ingredient-focused outlets specific to your niche.

Evidence it works reasonably quickly: 5WPR documented Sol de Janeiro's Cheirosa 40 mist earning coverage in Allure, Byrdie and Refinery29 that fed into Perplexity and ChatGPT citations within roughly eight weeks.

Step 02

Make your ingredients machine-readable

Every product page should state the active ingredient and its concentration in plain text, near the top, not buried in a collapsed INCI list at the bottom.

Add ingredient flags as structured attributes rather than icons on a photo: fragrance-free, non-comedogenic, vegan, cruelty-free, reef-safe, silicone-free. An AI can't read your badge graphics.

This overlaps with the wider structured data work we cover in optimizing product pages for AI recommendations.

Step 03

Structure the beauty-specific attributes

Beauty has more filterable attributes than almost any other category, and AI shopping systems use them as hard constraints.

The ones that matter: skin type compatibility (oily, dry, combination, sensitive), undertone, shade, finish, coverage level, formulation type, application method, time of day, and layering compatibility.

If "suitable for sensitive skin" only appears in your marketing copy, an AI filtering for sensitive skin cannot confirm it and will recommend a competitor who structured that field properly.

Step 04

Get dermatologist and expert association

Look at who dominates: La Roche-Posay, CeraVe. Both are heavily associated with dermatologist recommendation — that association is baked into how these models understand skincare authority.

Expert endorsement, dermatologist-bylined content, and clinical testing coverage all feed this. K18 was noted as climbing the index partly on the back of dermatologist-bylined editorial content alongside expanded Sephora placement.

Step 05

Treat reviews as infrastructure

Beauty has the highest review-per-SKU rate of any retail category, and AI systems weight review volume and sentiment heavily when deciding what's safe to recommend.

Spread matters as much as volume. Reviews on your own site, on retailer pages, and on independent platforms all contribute. A product with 40 reviews in one place is a weaker signal than one with 400 across four.

Step 06

Check you're not invisible to the crawlers

Unglamorous but binary. Open your robots.txt and look for GPTBot, OAI-SearchBot, PerplexityBot, ClaudeBot and Google-Extended.

If any are blocked — often left over from a site migration — that engine cannot use your content at all. Also check your CDN's bot rules, which can block silently while robots.txt looks clean.

If nothing is working and you're not sure why, why your website doesn't appear in ChatGPT walks through the full diagnostic.

TimingWhy the Window Is Open Right Now

One detail in the eMarketer data is easy to miss and genuinely important.

Brand mention rates declined across categories in Q1 2026. La Roche-Posay slipped from 22% to 21%. The pattern held broadly.

That's not a fall in AI usage. It means ChatGPT is distributing recommendations across more brands rather than defaulting to the same few. Concentration is loosening.

For an established leader that's mild bad news. For everyone else it's the opposite — there is more room in the answer than there was six months ago.

Worth knowing about agentic commerce Beauty is the leading vertical in AI shopping integrations. Ulta went live on Google's commerce protocol as a launch partner, Glossier ships through ChatGPT's, and SKIMS Beauty runs on Stripe's. The reason beauty went first is the same reason this article matters — beauty queries map cleanly onto structured attributes. If your product data isn't clean, you won't just miss recommendations; you'll be unable to participate in AI-completed purchases at all.

Honest CaveatsWhat This Data Can't Tell You

Most of it is vendor research. Avenue Z and 5WPR are PR firms. eMarketer sells subscriptions. Their findings are directionally consistent with each other and with independent work in other verticals, which is why they're worth using — but methodologies aren't fully disclosed and none are peer-reviewed.

Correlation, not causation. The Ordinary has ingredient-led naming and enormous cultural presence. Untangling which one drives the citations isn't possible from this data. The mechanism is plausible, and testable on your own catalogue, but don't treat it as proven.

AI answers vary run to run. Ask the same question twice and you may get different brands. Any single observation means very little. Track a fixed set of prompts over time before concluding anything.

It moves. Rankings shifted noticeably between Q4 2025 and Q1 2026. Whatever you read today about who's winning is a snapshot.

Where to StartThis Week

Run ten real customer prompts through ChatGPT, Perplexity and Gemini. "Best serum for hyperpigmentation." "Fragrance-free moisturiser for sensitive skin." Write down every brand named.
Check whether you appear at all. If you do, note whether you're named first or listed as an also-ran.
Open the editorial articles the AI cites. Check whether you're in them. Usually you won't be — that's your PR target list.
Audit one hero product page. Is the active ingredient in plain text near the top? Are skin type and ingredient flags structured data, or just graphics?
Check robots.txt for the five AI crawlers.

That's about two hours and it will tell you more than any tool subscription. If you'd rather we ran it, the AI visibility checker is the fastest starting point, and our AI SEO for beauty and lifestyle page covers how we approach the category. For the broader commerce angle, see AI SEO for ecommerce and our piece on how ChatGPT and Perplexity are changing online shopping.

Find out which beauty brands ChatGPT names instead of you

We'll run real customer prompts across ChatGPT, Perplexity and Gemini, show you who gets cited, and identify the editorial gaps behind it.

Get Your Free AI Visibility Audit

Questions first? Talk to us. Or see how we optimize websites for ChatGPT.

Frequently asked questions

How does ChatGPT decide which beauty brands to recommend? +

Mostly from what other people write about you, not what you write about yourself. Avenue Z's 2026 AI Visibility Index Beauty Report analysed 565 AI citations across 60 beauty brands and found 96.1% came from third-party sources such as editorial coverage and trusted reference sites. Brand-owned content accounted for just 3.9%, with editorial publications alone making up 63.2% of all citations.

Which beauty brands does ChatGPT recommend most? +

La Roche-Posay has led most indices through 2026. eMarketer, analysing more than 5,200 ChatGPT responses, found it recommended in 81% of facial skincare queries in Q1 2026. CeraVe and Olaplex also rank consistently high. In makeup the picture is far less concentrated — Maybelline led lip and eye categories, but no single brand dominated across all makeup types.

Can a small indie beauty brand get recommended by ChatGPT? +

Yes, and the data suggests indie brands are outperforming legacy ones. 5WPR analysis found ingredient-led independents such as The Ordinary, Drunk Elephant and Glow Recipe earned around 31% of beauty AI citations, while legacy mass brands earned roughly 4%. Brand mention rates have also been declining across categories, which indicates ChatGPT is spreading recommendations across a wider set of brands rather than concentrating them.

Why does The Ordinary do so well in AI recommendations? +

Ingredient naming. 5WPR found The Ordinary held roughly 7.0% citation share on ChatGPT — the highest of any single beauty brand — and attributed it to ingredient-name product transparency that AI engines can parse as structured data. When a product is literally called Niacinamide 10% + Zinc 1%, an AI matching a query about niacinamide has an unambiguous match. Most beauty naming conventions hide the ingredient behind a marketing name.

Does editorial PR help beauty brands appear in AI search? +

It is arguably the single highest-leverage activity. Editorial publications accounted for 63.2% of AI beauty citations in the Avenue Z analysis. There is also evidence it works relatively quickly — 5WPR documented Sol de Janeiro's Cheirosa 40 mist earning coverage in Allure, Byrdie and Refinery29 that fed into Perplexity and ChatGPT citations within roughly eight weeks.

What product attributes should beauty brands structure for AI? +

Beauty has unusually specific filters. The attributes AI shopping systems weight most include skin type compatibility, undertone, shade, finish, coverage level, ingredient flags such as fragrance-free or vegan, formulation type, application method, time of day, and layering compatibility. Each needs to exist as structured machine-readable data rather than only appearing in marketing copy or product photography.

How long does it take for AI to start recommending a beauty brand? +

Faster than most categories, but not instant. Editorial coverage has been observed feeding into retrieval-based engines such as Perplexity and ChatGPT Search within around eight weeks. Deeper training-layer visibility, which drives recommendations when the AI answers from memory rather than searching, accumulates over much longer periods and depends on sustained coverage and community discussion.

Sources & references

Avenue Z — 2026 AI Visibility Index Beauty Report (60 brands, 565 citations, 75 prompts; 96.1% third-party, 63.2% editorial), as reported by Personal Care Insights
eMarketer — AI Visibility Index: Personal Care & Beauty Q1 2026 (5,200+ ChatGPT responses; La Roche-Posay 81% of facial skincare queries; declining mention rates)
5WPR — Beauty AI Visibility Index 2026 (80+ prompts across five engines; The Ordinary 7.0% citation share; category splits; Sol de Janeiro and K18 case notes)
BeautyMatter — The Beauty Brands That Are Ruling AI (47% prefer AI-recommended products)
AI Beauty Authority Index — training layer vs retrieval layer framework
Paz.ai — AI Shopping for Beauty Brands 2026 (beauty-specific attributes; agentic commerce integrations; review-per-SKU rates)
OpenAI — crawler documentation (GPTBot and OAI-SearchBot)

Sourcing note: several of these indices are published by PR and research firms that sell AI visibility services, and their full methodologies are not public. We cite them because multiple independent analyses report consistent directional findings, but the specific percentages should be treated as estimates rather than precise measurements. Rankings also shift quarter to quarter.