How do beauty brands win on AI search?

On March 24, 2026, Sephora launched its app inside ChatGPT — one of the first major prestige beauty retailers to put its own shopping experience inside a large language model. A month later, Ulta Beauty made its assortment shoppable, with checkout, in Google AI Mode and the Gemini app. Ulta's loyalty program, now 47 million active members, drives approximately 95% of total sales, and the retailer uses AI to personalize it at scale. Beauty discovery is moving into AI interfaces quickly. Most brands haven't adjusted. The ones that have are pulling ahead at a measurable rate.

What changed in 2026

The single most consequential development in beauty distribution in 2026 wasn't a new retailer or a new channel. It was a structural shift in how customers begin their shopping journey — and how the major prestige retailers chose to position themselves inside that shift.

Two retailer moves in the spring drove the change. On March 24, 2026, Sephora announced the launch of its app inside ChatGPT, allowing AI conversations to surface Sephora products and route users directly into the Sephora purchase experience (Sephora Newsroom, March 24, 2026). On April 22, Ulta Beauty and Google made Ulta's assortment shoppable in Google AI Mode and the Gemini app through Google's Universal Commerce Protocol, and launched Ulta AI, a Gemini-powered shopping assistant on Ulta.com (Ulta Beauty, April 22, 2026). Earlier, at CES 2026 in January, Ulta Beauty CMO Kelly Mahoney told The Speed of Culture podcast that the retailer's loyalty program drives approximately 95% of total sales, with AI-driven motivation-based segmentation replacing demographic targeting (Adweek, May 2026).

The macro numbers explain why this happened now. ChatGPT passed 900 million weekly active users in February 2026 (TechCrunch) and was reported to be nearing 1 billion by July. Google AI Overviews appeared on 25.11% of the 21.9 million searches in Conductor's 2026 AEO/GEO Benchmarks Report — and on 48.75% of healthcare queries, though only 6.82% in consumer staples. In 2024, Gartner predicted that traditional search engine volume would drop 25% by 2026 as people shift to AI chatbots and virtual agents. Vercel CEO Guillermo Rauch said in April 2025 that ChatGPT referred 10% of new Vercel signups.

For beauty brand teams, the question is no longer whether AI will reshape beauty discovery. The question is whether your brand is optimized for the layer that has already taken hold.

Inside the Sephora × ChatGPT integration

Sephora's integration with ChatGPT, piloting initially in the United States, allows users to ask beauty-related questions in natural language inside ChatGPT — queries like "help me find a foundation for dry skin" or "what's a good vitamin C serum under $50" — and receive curated product recommendations powered by Sephora's catalog and beauty knowledge base. Users who link their Sephora Beauty Insider loyalty account get personalized recommendations based on profile preferences, purchase history, and loyalty tier (Sephora Newsroom, March 2026).

The integration includes Sephora loyalty rewards and member benefits inside ChatGPT — free shipping promotions and samples — with full payment and checkout planned for future updates. According to Sephora, the company serves more than 80 million active Beauty Insider members globally.

Sephora's framing for the launch came from Anca Marola, the retailer's global chief digital officer, who announced it at Shoptalk Spring 2026 and said Sephora is "passionate about assisting our customers with the best beauty advice and curation, wherever and whenever they want." The strategic implication is significant: Sephora is not positioning ChatGPT as a competitor to its own app and website. It is positioning ChatGPT as a new front door to the Sephora experience. Customers who arrive through AI conversation are still Sephora customers — just acquired through a different surface.

The launch came as OpenAI changed course on commerce. In March 2026 it scaled back Instant Checkout, its native in-chat payment feature, after few merchants used it, and shifted purchases to retailer apps connected to ChatGPT while keeping product discovery inside the chat (Search Engine Land, March 2026). In May it began letting retailers turn their product feeds into ChatGPT shopping ads (Digiday, May 2026). Either way, the product data a retailer sends to ChatGPT is now part of the shelf.

For beauty brands sold through Sephora, the integration creates a new performance surface to optimize for. When a ChatGPT user asks about products in a brand's category, the brand's listings on Sephora.com — including A+ content, ingredient lists, claim language, and review depth — directly influence whether the brand surfaces in the AI's recommendation.

Inside the Ulta × AI personalization engine

Ulta Beauty's approach to AI is structurally different from Sephora's. Sephora moved first inside ChatGPT. Ulta anchors its AI strategy in its proprietary first-party data and personalization engine, and in April 2026 took that into Google's AI surfaces: its products are shoppable, with checkout, in AI Mode and the Gemini app, and Ulta AI, its on-site assistant, runs on Gemini Enterprise.

Under CMO Kelly Mahoney — who joined Ulta in 2015 and was named CMO in early 2025 after serving as SVP of customer and growth marketing — Ulta has invested over a decade in loyalty data infrastructure. The Ulta Beauty Rewards program had 46.7 million members generating approximately 95% of total sales, according to Mahoney's interview with Adweek published May 5, 2026; in August, Ulta reported 47 million active members.

Mahoney has publicly described Ulta's shift from demographic segmentation to motivation-based segmentation — categorizing customers not by age, gender, or income but by what drives their shopping behavior (routine optimization, ritual exploration, value seeking, prestige aspiration). The personalization engine, powered by Adobe Customer Data Platform and Adobe Firefly creative automation, delivers product recommendations, in-app offers, and creative variations at scale across approximately 80% of Ulta sales that still happen in-store and the rapidly growing digital surfaces.

The strategic difference between the two retailers' AI bets has direct implications for brands. Sephora's bet says: "we will be wherever the customer is, including inside ChatGPT." Ulta's bet says: "we own the data and the relationship, and we take it to the AI platforms we choose" — starting with Google Gemini. Both bets can work. They produce very different optimization requirements for the brands that sell through each retailer.

The retailer-mediated AI discovery layer

The most important structural insight about AI search for beauty in 2026 is this: AI does not treat brands as independent entities. When a customer asks ChatGPT, Perplexity, or Google AI Overviews for a beauty recommendation, the AI typically routes through a retailer — Sephora, Ulta, Amazon, or occasionally a direct DTC link — rather than surfacing the brand on its own.

This means a brand's AI visibility is largely a function of its presence and performance on the retailers AI platforms have integrated with. Brands cannot optimize for AI directly. They optimize for the retailers that AI optimizes for. This is the retailer-mediated discovery layer — and the brands that understand it are already pulling ahead.

Google is building the same layer. At I/O on May 19, 2026, it announced Universal Cart, a single shopping cart across Search, Gemini, YouTube and Gmail, with checkout through retailers including Sephora and Ulta Beauty, rolling out in the U.S. over the summer (Google, May 2026). Discovery, comparison and checkout are moving into the AI surface, and the retailer's product data decides what the shopper sees first.

The practical implication is significant. The investments most beauty brands have made in DTC site SEO and content marketing are not wasted — they still drive organic traffic and contribute to topical authority. But they no longer represent the highest-leverage GEO investment. The highest-leverage investments now sit in three areas: structured product data on retailer sites, ingredient and benefit content depth, and authoritative third-party coverage.

How AI engines actually read beauty products

AI shopping assistants don't "see" beauty products the way humans do. They don't process visual brand world, packaging, or in-store presentation. They process structured data, descriptive text, ingredient lists, claims, reviews, and surrounding editorial content. Brands with thin product data underperform regardless of brand strength.

Three categories of input matter most.

Structured product data

Every product detail page on every retailer where the brand sells should have complete, unambiguous data: full ingredient lists, benefit claims expressed in the language customers actually use, skin type and concern mappings, application instructions, and clean attribute taxonomy. The cleaner the data, the more often the product surfaces in AI answers to specific use-case queries — "best serum for sensitive skin with rosacea," "fragrance-free retinol alternative," "scalp care for textured hair." These are the queries that drive high-intent traffic, and they require structured data to win.

Ingredient and benefit language

AI models extract and recombine language from product descriptions, A+ content, and editorial coverage. The brands that write product descriptions in customer-language rather than marketing-language — describing what the product does, who it's for, and what makes it different — get cited more often than brands that lead with hero-product mythology.

Review and content depth

AI models weight authority and recency. A product with 200 reviews from the last six months outperforms a product with 2,000 reviews from three years ago. Brands that maintain review velocity through sampling programs, loyalty member outreach, and creator partnerships build the kind of evergreen authority signal that AI models reward. The same applies to editorial coverage: a brand mentioned recently in WWD, Glossy, Allure, or Beauty Independent has a citation footprint AI can read. Adjacent examples of platform-native beauty commerce — like the Roblox Pride campaigns Glossy covered — show how the category is testing new commerce surfaces beyond retailer storefronts.

The 5 GEO techniques Princeton proved work

The empirical foundation for AI search optimization is a 2023 research paper led by Princeton researchers with IIT Delhi — "GEO: Generative Engine Optimization" by Aggarwal et al., presented at ACM KDD 2024. The team tested nine optimization tactics on GEO-bench, a 10,000-query benchmark, using a generative engine modeled on Bing Chat, then validated findings on Perplexity. The strongest tactics, adding quotations, statistics and source citations, lifted a source's visibility in AI answers by roughly 30 to 40 percent; fluency edits added 15 to 30 percent. Keyword stuffing performed worse than doing nothing.

Figures are the paper's relative lift in position-adjusted word count, its main visibility measure, against unoptimized content. On Perplexity, quotation addition lifted visibility 22%.

The paper's most consequential finding for emerging beauty brands is that lower-ranked websites benefit far more from GEO than top-ranked sites. A brand that is already dominant in traditional search has less to gain from GEO optimization than a brand fighting for visibility — meaning challenger brands and indie beauty have a window to outperform much larger competitors with disciplined GEO investment.

The Princeton team found that Statistics Addition paired with Fluency Optimization produced the largest combined effect. The implication: it isn't enough to add data; the data must be embedded in well-structured prose that AI engines can extract cleanly.

Which AI engines matter most for beauty

Not all AI engines behave the same. Beauty brands should monitor visibility across five major engines, each with distinct retrieval and citation patterns.

The mix is shifting fast. Similarweb data show ChatGPT's share of generative-AI web traffic falling from 76% to about 53% in the twelve months to May 2026, while Gemini's rose from 9% to 27% and Claude's from under 2% to about 9% (PPC Land, June 2026). A brand that only tests its visibility in ChatGPT is measuring half the market.

The new KPIs: measuring AI search performance

The KPIs that defined SEO from 2010 to 2020 — organic rank, click-through rate, total impressions — remain relevant but no longer capture the full picture. AI search has its own metrics, and most brands aren't measuring them.

The four emerging KPIs that matter for beauty AI search:

• Share of voice in AI-generated answers. For category queries ("best vitamin C serum") and use-case queries ("what should I use for dry skin in winter"). Requires running 15-25 priority prompts daily across ChatGPT, Perplexity, Gemini, Claude, and Google AI Overviews, tracking which brands appear. GenOptima recommends a 10%+ answer-level mention rate across 8 engines as a strong commercial baseline.
• Brand citation rate. How often the brand is named when AI generates answers in your category. In an Ahrefs study of 75,000 brands, web mentions correlated 0.664 with brand visibility in Google AI Overviews, while backlinks correlated only 0.218 — a fundamental shift from traditional SEO authority signals.
• AI referral traffic in GA4. ChatGPT, Perplexity, Gemini, and Claude now drive measurable referral traffic. Tagging and segmenting these sessions reveals which AI engines convert best for the brand's category.
• Retailer site visibility for AI-routed sessions. Most retailers don't yet break out AI-sourced traffic in shared dashboards, but the brands that ask their retail partners for this data are getting ahead of the measurement curve.

Most beauty brands are not yet measuring any of these. According to industry benchmarks, fewer than 12% of marketing teams have a documented strategy for appearing in AI-generated answers (GenOptima, April 2026). The brands that build measurement infrastructure now will know what's working before competitors realize there's something to measure.

The 60-day GEO playbook for beauty brands

For a beauty brand serious about the AI discovery layer, the following 60-day playbook produces measurable results in the 4-8 week window most GEO infrastructure investments require.

Days 1-14: Audit and baseline

Pull every product detail page on Sephora.com, Ulta.com, Amazon, the brand's DTC site, Target.com if applicable, and any other retailer. Verify ingredient completeness, claim consistency, skin-type and concern tagging, and attribute taxonomy. Run 15-25 priority queries across ChatGPT, Perplexity, Claude, Gemini, and Google AI Overviews. Document which brands surface and where you appear. This becomes your baseline.

Days 15-30: Content optimization

Apply the five Princeton techniques across owned content. Target one hyperlinked statistic per 150-200 words. Add one named-expert quotation per major section. Replace qualitative claims with quantitative ones wherever possible. Audit the brand's About, FAQ, Ingredient, and Education pages for fluency, varied sentence structure, and clear topic sentences.

Days 31-45: Schema and structured data

Implement Article schema, FAQPage schema, HowTo schema, Product schema with full ingredient and benefit attributes, BreadcrumbList schema, and Organization schema with sameAs links to LinkedIn, Wikipedia (if applicable), and industry directories. Schema markup may not be the dominant factor in AI citation, but it's a baseline signal that AI engines use to disambiguate entities.

Days 46-60: Third-party authority

Build the editorial footprint that AI engines weight. Founder commentary in industry publications, product reviews in WWD/Glossy/Allure/Beauty Independent, bylined contributions on category questions, podcast appearances. Recency matters: AI engines weight recent mentions more heavily than older ones. A monthly cadence of authoritative third-party mentions sustains the signal.

The opportunity

Beauty AI search in 2026 is the moment Google search was in 2002 — an established player, an obvious shift in customer behavior, and a window during which the brands that move quickly can establish positions very hard to dislodge later. The brands that built SEO capabilities early earned authority that compounded for two decades. The same dynamic is now playing out for AI search.

The retailer-mediated structure means this isn't a winner-take-all game. Brands can win in AI search without owning their own AI assistant, without building proprietary technology, and without abandoning the channels they already operate. They win by being more deliberate about the inputs AI models read — and by treating AI search as the discovery layer it actually is, not a future curiosity.

Sources

1. Sephora Newsroom — Sephora App in ChatGPT Brings a New Personalized Beauty Experience (March 24, 2026)
2. Ulta Beauty — Ulta Beauty and Google Introduce Gemini-Enabled Shopping Experiences (April 22, 2026)
3. Ulta Beauty Q2 fiscal 2026 earnings call transcript (August 27, 2026)
4. Adweek — Ulta Beauty says AI is supercharging its rewards program (May 5, 2026)
5. Brand Innovators — CMO of the Week: Ulta Beauty's Kelly Mahoney (January 5, 2026)
6. Suzy — Intelligent Beauty: how Ulta Beauty uses AI personalization (The Speed of Culture, February 2026)
7. Retail Dive — Ulta Beauty names Kelly Mahoney CMO (February 2025)
8. Search Engine Land — OpenAI's ChatGPT Instant Checkout plan changes (March 2026)
9. Digiday — OpenAI makes it easier to run shopping ads in ChatGPT (May 12, 2026)
10. Google — Universal Cart and agentic shopping (May 19, 2026)
11. TechCrunch — ChatGPT reaches 900M weekly active users (February 27, 2026)
12. The Next Web — ChatGPT nears 1 billion weekly active users (July 29, 2026)
13. PPC Land — Similarweb generative-AI traffic share, May 2026 (June 2026)
14. Conductor — The 2026 AEO/GEO Benchmarks Report
15. Gartner — Search engine volume will drop 25% by 2026 (February 19, 2024)
16. Aggarwal et al. — GEO: Generative Engine Optimization (KDD 2024)
17. Ahrefs — AI Overview brand visibility factors (75K brands studied)
18. GenOptima — Generative Engine Optimization Best Practices 2026
19. Guillermo Rauch (Vercel) — ChatGPT refers 10% of new Vercel signups (April 2025)
20. FashionNetwork — Sephora launches an app within ChatGPT (March 2026)