• AI visibility

Guide · AI search

How do makeup brands get recommended when shoppers ask AI?

By making every shade, undertone, finish and use case easy for AI assistants to read, compare and confirm from sources they trust. Makeup shoppers already ask AI which foundation matches their skin, and the answer names a handful of products. A brand whose shade facts are missing or inconsistent is easy to leave out, and a brand that is named can win a shopper who was not loyal to anyone.

By Underneath Team Guide Updated

The short version

  1. Makeup is the largest prestige beauty category: Circana (opens in a new tab) reports US prestige beauty sales of $36 billion in 2025, with makeup up 4%, and mass-market beauty at $72.7 billion.
  2. AI is already in the beauty aisle: in Criteo’s June 2026 survey (opens in a new tab) across six countries, 38% of beauty and personal care shoppers use AI assistants, and 57% of those say the products recommended in AI answers influence what they buy.
  3. Shade is where online makeup sales break: in The Benchmarking Company’s March 2026 survey of more than 4,300 American women, reported by CosmeticOBS (opens in a new tab), 41% had trouble finding their complexion shade, rising to 56% among women with darker skin.
  4. Shoppers are open to switching: Criteo found only 34% rarely switch from their preferred beauty brands, and up to 53% of fragrance, skincare and makeup purchases came from a brand the shopper had not bought in the previous 12 months.
  5. Retailers have moved: Ulta Beauty (opens in a new tab) made its products shoppable inside Google’s AI Mode and the Gemini app in April 2026, a month after Sephora (opens in a new tab) launched its app inside ChatGPT.

How do people shop for makeup now, and why is each new customer worth fighting for?

They research hard, compare many products and switch brands easily, so every purchase is a chance to win someone new.

A makeup purchase looks small, but the decision behind it is not. Criteo’s commerce data shows US shoppers browse an average of 19 makeup products before buying. Reviews and ratings are the top purchase influence, with nearly half of shoppers relying on them, followed by friends and family, in-store testing and search engines. More than half of cosmetics and perfume purchases (57%) involve an online touchpoint.

Loyalty is thin. Because only about a third of shoppers rarely switch, and up to 53% of purchases go to a brand the shopper had not bought in a year, makeup brands win and lose customers constantly. The prize for being chosen is larger than one order: a foundation or concealer that matches becomes a repeat purchase, and a shopper who finds her shade tends to stop searching. We infer that the first match is where much of a complexion customer’s value is decided, though no public source we found puts a dollar figure on it.

The category is growing, but growth is uneven. Circana says all prestige makeup segments grew in dollars, with unit softness in face and eye makeup, and that lip was the fastest-growing segment in mass retail. Value, social media and “skinification,” hybrid products that blend color and skincare benefits, drove much of the growth. Each of those trends creates new questions that shoppers now put to AI. Jewelry shows the same effect, where the shift to lab-grown stones now has its own group of shopper questions, as our guide to jewelry brands recommended by AI shows.

Where do AI assistants already sit in a makeup purchase?

At the start, as a shopping advisor that narrows dozens of options to a few named products.

The evidence is recent and consistent. Criteo’s figures above come from a survey of 4,595 shoppers in the US, UK, France, Germany, Japan and South Korea. NielsenIQ (opens in a new tab) reports that 49% of shoppers have already received beauty product recommendations from AI tools such as ChatGPT, Claude, Gemini and Copilot, and that 84% are more likely to buy when key product attributes can be easily compared. NielsenIQ’s point about the shelf is the one executives should remember: a physical shelf might show 50 or more products, while an AI answer may show only one or two.

Retail data points the same way, though it covers all of retail, not beauty alone. Adobe’s analysis (opens in a new tab) of more than 1 trillion visits to US retail sites found AI traffic rose 393% in the first quarter of 2026 compared with a year earlier, and that AI visitors converted 42% better than other traffic in March 2026.

The biggest beauty retailers are building for this. Ulta, with more than 1,500 stores and more than 46 million loyalty members, says shoppers can now receive Ulta product recommendations, compare options and complete checkout for eligible purchases inside AI Mode and Gemini. Its chief technology officer said the company sees “a clear shift in how guests are discovering and shopping for beauty, with AI playing a much bigger role in that journey.” Sephora’s ChatGPT app launched in March 2026. Glossy’s reporters, testing the assistants, noted that ChatGPT gave long answers covering “the various finishes and the various undertones.” That is a description of one test, not a measurement, but it shows the vocabulary these answers run on.

Which questions do makeup shoppers ask AI?

Questions about shade, undertone, finish, skin type, wear time, price and dupes, usually in one sentence.

We wrote the prompts below to illustrate the kinds of questions makeup shoppers type; they are not observed queries:

  • Shade and undertone: “Which foundation shades suit medium skin with an olive undertone, in a satin finish?”
  • Finish and skin type: “Best dewy foundation for dry, mature skin under $40.”
  • Use case: “Long-wear concealer for dark circles on deep skin that won’t crease.”
  • Wear: “Lipstick that doesn’t transfer onto a mask or a coffee cup.”
  • Values: “Cruelty-free mascara for sensitive eyes that holds a curl.”
  • Value and dupes: “Cheaper alternative to a prestige brow gel with the same hold.”

NielsenIQ lists similar real-world question shapes, such as a “good cruelty-free foundation for combination skin.” Google noted in 2022 that foundation is the most-searched category within makeup (opens in a new tab). These questions differ from skincare questions, which turn on ingredients and claims, and from retailer questions about where to buy. A makeup question is usually about how a product will look on one person’s face.

How does an AI answer turn into a makeup sale?

Through a named product and shade, a click to a retailer or brand site, a first order, then repeat purchases.

The answer. The assistant names a few products, sometimes with a suggested shade. If your shade names, undertone labels and finish descriptions are clear, a reasonable expectation is that the assistant can match them to the shopper’s words; if they are vague, it has less to work with.

The click or the checkout. For most makeup brands, the sale lands at a retailer: Sephora, Ulta, Amazon or a mass chain. With Ulta’s catalog shoppable inside Google’s AI surfaces, and OpenAI documenting that ChatGPT may show an Instant Checkout option (opens in a new tab) for some eligible products and merchants, part of that sale may never touch a website. ChatGPT’s option is in flux, though: OpenAI scaled it back in March 2026, according to FashionUnited (opens in a new tab), even as the help page still lists it. In-assistant buying makes sales harder to see in analytics, which is one reason brands should measure what AI says directly.

The shade check. This is where makeup differs from most categories. Google reported in 2022 that more than 60% of online beauty shoppers had decided not to buy a beauty item online because they did not know what color or shade to choose, and 41% had returned an item because it was the wrong shade. Perfect Corp, a try-on vendor, reports that Benefit Cosmetics saw a 113% increase in conversion rate (opens in a new tab) and a 20% increase in add-to-cart actions after adding eyebrow virtual try-on. Those are vendor-reported results, but they show that confidence about shade is what turns interest into an order. Clothing has the same problem with fit, covered in how apparel brands win customers through AI.

The repeat. A matched shade is replenished. Our inference is that AI visibility for a complexion product pays back over several purchases, not one.

What decides whether an AI assistant names your makeup product?

Platforms document product data, price and reviews; studies add that a small, visible quality edge can beat a famous name.

Documented by the platform. OpenAI says that when choosing products, ChatGPT considers structured metadata from first-party and third-party providers, such as price and product description, along with other third-party content and reviews. It can display review summaries built from public websites and labels such as “Most popular” that it generates itself. If the same foundation is sold by Sephora, Ulta and the brand itself, ChatGPT orders those sellers by availability, price, quality and whether the seller is the maker or primary seller. On Google’s side, AI Mode pulls from shopping data for billions of products (opens in a new tab), firing off several related searches at once and then merging what they return. OpenAI’s documented “Try on” button covers clothes and accessories, not makeup, so shade confidence in AI answers still depends on words and data.

Observed in a study. In a laboratory study of skincare recommendations (opens in a new tab) across three AI models, well-known brands were recommended 100% of the time when all products had the same specifications, but that dominance disappeared when a competitor had a rating advantage of less than 0.1 stars. The same study found that invented clinical claims also shifted picks, which is a reason for platforms to verify, not a tactic: we cover the risks in can you game AI shopping rankings. For makeup, we infer the practical reading is that a challenger brand needs a clear, verifiable difference, for example a wider shade range or better ratings from shoppers with a given skin tone.

Trust factors specific to makeup. Shade range and how inclusive it really is matter more here than in almost any category. In the Benchmarking Company survey, 39% of women had stopped using a brand because its range was not extensive enough, rising to 61% among women with darker skin. When a product failed, the reasons were the wrong undertone (66%), a color too light (52%) and a finish that looked different in photos than in real life (51%). Reviews that mention skin tone, undertone and wear time, retailer listings that agree with the brand site, and independent “best foundation for” coverage are the evidence an assistant can find. Our guide to which pages to target explains why third-party lists matter.

What does a makeup brand lose when AI leaves it out?

New customers at the moment they were ready to switch, and often the repeat orders that follow a shade match.

The evidence on lost sales is indirect, so we are careful here. What is documented: shoppers are open to new brands, more than a third use AI while shopping for beauty in Criteo’s data, and those who do mostly say the answers influence what they buy. Adobe found about 34% of retail product pages cannot be properly accessed by AI, across all retail, which suggests many brands start with a gap they have not measured. What we infer: when an answer names two or three foundations for “olive undertone, satin finish,” the brands not named lose a shopper who might have stayed for years. And when an answer names your product but suggests a shade from outdated or wrong data, the cost shows up as a return or a bad review instead of a missed sale. What happens if you skip GEO covers the broader effect on search traffic.

How does GEO work for a cosmetics brand?

It makes your shade, finish and use-case facts easy for AI to find, trust and repeat.

For a makeup brand, generative engine optimization (GEO) means getting AI answers to name your products and quote their shades and finishes correctly. It cannot buy a recommendation, and no one can promise a placement. For makeup, it usually covers seven things:

  1. Shade-level product facts. Every shade with its depth, undertone, finish, coverage, skin type and wear claims, written the same way on your site, in product feeds and in retailer listings. NielsenIQ argues that AI systems rely on “structured, complete, and easily interpretable product information”; our article on what product content AI shopping assistants prefer covers what tests show.
  2. Answers to shade questions. Plain pages that answer the questions shoppers ask: how your undertone labels work, which shade suits which skin, how the finish wears on oily or dry skin, which shades replace a discontinued one.
  3. Reviews with context. Encouraging honest reviews on retailer sites that mention skin tone, undertone and wear, since reviews are the top purchase influence and assistants summarize them. Never fake ones: fake reviews and AI recommendations explains the risk.
  4. Independent coverage. Beauty editors, creators’ written reviews and “best of” roundups that test your products across skin tones. Digital PR that earns this coverage gives assistants a source other than your own claims. Our study of what makes assistants recommend a brand tests how much outside coverage matters.
  5. Consistent facts across retailers. Shade names, prices and availability that match on Sephora, Ulta, Amazon and your site, so an assistant does not see conflicting information.
  6. Readable pages. Product pages that AI systems can actually access and read, not shade data hidden in images or scripts.
  7. Measurement across AI surfaces. Tracking a fixed set of shade, finish and dupe questions across ChatGPT, Google AI Mode, Gemini, Perplexity and retailer assistants, repeated over time, because answers vary from run to run.

Which makeup questions does the research leave open?

How often AI answers name the right shade, and how much makeup revenue comes through AI today.

We found no public data on beauty sales completed inside AI assistants, on how accurate AI shade suggestions are, or on whether AI-referred makeup shoppers return products less often. The most specific survey figures come from Criteo and NielsenIQ, both companies that sell services to brands, and Criteo’s spans six countries rather than the US alone. Adobe’s conversion and traffic figures cover all US retail. The skincare study ran in a laboratory, with invented products, so it shows how models can respond, not what they do on live shopping pages. Google’s shade figures date from 2022. And the retailer launches at Ulta and Sephora are months old, with no published results yet.

Where should a makeup brand start?

With your hero complexion products: check what AI assistants say about their shades, finishes and where to buy them.

Pick the 20 to 30 questions your shoppers really ask about your foundations, concealers and lip products, by shade and undertone. Run them across the main AI assistants and retailer assistants, and record which brands are named, which shades are suggested, which retailer gets the click and where the facts are wrong. That shows whether you have a visibility problem, an accuracy problem or both. Should you want a hand, our team can run it with you: we will map how AI assistants describe your range today and set out the product-data, review and coverage work most likely to win more first purchases and fewer wrong-shade returns. Shade-level product facts, reviews with skin-tone context and matching retailer listings are each part of our generative engine optimization service, which explains how the work runs.

Frequently asked questions

Do AI assistants offer makeup virtual try-on?

Not in ChatGPT’s documented shopping features. OpenAI’s help page describes a “Try on” button for clothes and accessories. Makeup try-on lives on brand and retailer sites and in tools from vendors such as Perfect Corp, so in AI answers your shade descriptions do the work.

Should a makeup brand focus on its own site or on retailers?

Both. Most makeup sales happen at retailers, and assistants read retailer listings and reviews as well as brand sites. The facts should match everywhere.

Is AI visibility different for skincare and makeup?

Yes. Skincare questions turn on ingredients and claims, which carry regulatory limits. Makeup questions turn on shade, undertone, finish and wear, which are matters of accurate product description.

Can a smaller makeup brand compete with the big names in AI answers?

In tests, yes, when it has a clear and checkable advantage. The evidence is gathered in how small brands get recommended by AI and whether AI assistants favor big brands.

Sources

Free strategy call

Some questions are easier to answer about your own business.

Bring the one that matters most. On a free 30-minute call we’ll take a first look at it and send you a short written read afterward.

Ask about my businessHow we work