Sephora Inside ChatGPT: Why the Smartest Retailer in Beauty Just Made Its Most Important Tech Decision

Kristina Bokova
Ex-LVMH, Dior, Aura Blockchain Consortium · April 2026

On March 24, 2026, Sephora launched an app inside ChatGPT (Sephora Newsroom). Most of the coverage focused on the novelty: a beauty retailer living inside an AI chatbot. The more significant decision was underneath the headline. Sephora chose to bring its data, its loyalty programme, and its product expertise into the AI platform — rather than waiting for the AI platform to represent Sephora on its own terms.

That distinction — between embedding and being scraped — will likely define the next phase of retail.

What Sephora Actually Built

The app is piloted in the US. Users can prompt ChatGPT with beauty-specific questions — "Sephora, help me find a foundation for dry skin" — and receive product recommendations drawn from Sephora's catalogue. That alone would be a chatbot. What makes it different is the Beauty Insider integration.

Users who link their Beauty Insider account give the app access to their purchase history, skin type, and product preferences. The recommendations are not generic. They account for what a customer has already bought, what their profile says about their skin, and what has worked for them before. Loyalty benefits carry over: free shipping, samples, and tier-specific perks all function inside ChatGPT.

The app is built on OpenAI's Agentic Commerce Protocol (ACP), a framework that allows retailers to share product catalogues and services with ChatGPT in a structured way. Sephora and OpenAI are exploring broader collaboration across global operations (Sephora Newsroom). Payments and checkout directly within ChatGPT are planned for future releases.

Why This Is Different From a Chatbot on Your Website

Most retailers that have adopted AI in the past two years have done so by building chatbots, recommendation engines, or virtual assistants on their own websites and apps. The logic is straightforward: bring the AI to where the customer already shops.

Sephora inverted that logic. Instead of adding AI to sephora.com, it brought Sephora to where people are already asking questions.

ChatGPT has over 800 million weekly active users. A significant and growing share of product research now starts with an AI prompt rather than a search engine query. When someone asks ChatGPT "what moisturiser is best for combination skin in winter," the answer is being assembled from whatever data the model has access to. Without a structured presence, a retailer has no control over which products are recommended, how they are described, or whether the recommendation reflects current inventory.

By embedding inside the platform, Sephora ensures that when beauty questions come up in ChatGPT, the answers can draw on Sephora's actual product data, real-time availability, and the customer's own purchase history — rather than whatever the model inferred from web scraping.

The Data Play

The Beauty Insider programme has over 45 million members in North America (eMarketer, 2025). It is one of the most data-rich loyalty programmes in retail — tracking purchases, product preferences, skin profiles, shade matches, and engagement across online, in-store, and app channels.

Sephora's decision to connect this programme to ChatGPT — rather than expose raw data — is a deliberate hedge. The customer gets a personalised experience. Sephora retains data sovereignty. The AI platform gets structured access to product information and preference signals, but not the underlying customer database.

This is a meaningful architectural choice. Compare it with the alternative: a retailer that does not embed in ChatGPT has its products represented by whatever information OpenAI's crawlers collected from the open web. Pricing may be outdated. Product descriptions may be incomplete. Discontinued items may still appear. The retailer has no mechanism to correct or update any of it.

Sephora's approach trades some platform dependency for informational control. Whether that trade-off is worth it depends on the retailer. But the retailers making no trade-off at all — neither embedding nor controlling their data — are the ones most exposed.

Two Strategies for the Same Shift

Sephora is not the only retailer adapting to AI-driven commerce. Aritzia, the Canadian fashion retailer, took a different approach to the same underlying shift.

Aritzia deployed Google's AI Max campaign tool and saw online sales increase 80% (Glossy). The company posted its first billion-dollar quarter. The strategy: optimise for AI-driven discovery. Make the products findable when AI systems are assembling answers to fashion queries. Let the AI bring customers to you.

Sephora's strategy is the inverse: go to where the AI already has the audience, and bring the full retail experience with you.

Both approaches respond to the same reality. An increasing share of product discovery is mediated by AI — whether through Google's AI Overviews, ChatGPT conversations, or AI-powered shopping assistants. Brands that are not structured for AI retrieval are becoming harder to find. Brands that are not present inside AI platforms are absent from a growing share of purchase journeys.

The brands doing neither — relying entirely on traditional search, social, and owned channels — are not yet invisible. But the gap is widening quarter by quarter.

What This Means for Luxury

Most luxury brands do not have a loyalty programme as data-rich as Beauty Insider. Few maisons operate a points-based system at all. But they do have clienteling data, purchase histories, product expertise, and editorial content that is often more detailed and more carefully curated than anything in mass retail.

The question is whether a maison will embed its product catalogue, its editorial voice, and its product knowledge inside AI platforms on its own terms — or wait until those platforms start recommending its products based on incomplete, uncontrolled information.

The second scenario is already happening. Ask ChatGPT to recommend a watch for a specific occasion, a fragrance for a specific profile, or a handbag at a specific price point, and it will produce an answer. That answer draws on whatever data the model has. It may be accurate. It may not. The brand has no input and no recourse.

Clienteling data — the detailed knowledge a sales associate has about a VIC's preferences, occasions, and purchase rhythm — is arguably more valuable than a loyalty programme's transaction log. A maison that could connect even a fraction of that intelligence to an AI platform would offer something no scraper can replicate: genuine product expertise applied to individual context.

Whether luxury brands should embed inside ChatGPT specifically is a separate question. The principle is broader: in a world where AI mediates product discovery, the brands that structure their data for AI consumption will shape how they are represented. The brands that do not will be represented anyway — just less accurately.

The Speed Signal

Boll & Branch, a $200M-revenue bedding and home goods company, offers a different data point on the pace of this shift. CEO Scott Tannen built an AI agent called "Tess" on OpenClaw, running operations across marketing, customer service, and merchandising (Glossy). The implementation took weeks, not quarters.

Sephora built an app inside ChatGPT. Aritzia optimised its entire media strategy for AI discovery. Boll & Branch deployed an AI agent across core business functions. Three different scales, three different approaches, one shared pattern: AI adoption in retail is moving faster than AI strategy.

The companies acting now are setting the terms of how their brands appear in AI-generated recommendations, how their data integrates with AI platforms, and how their customers interact with AI as a shopping channel. The companies waiting for a comprehensive AI strategy before taking any action are, in practice, letting AI platforms define how they are represented.

There is no neutral position. A brand that is absent from AI platforms is still present in AI-generated answers — it just has no say in how it appears there.

Frequently Asked Questions

What is Sephora's ChatGPT app?

Sephora launched an app inside ChatGPT on March 24, 2026, initially piloted in the US. Users can prompt beauty advice directly within ChatGPT — for example, asking for a foundation recommendation for dry skin. The app is built on OpenAI's Agentic Commerce Protocol (ACP), which allows retailers to share product data and services with ChatGPT.

How does Beauty Insider work inside ChatGPT?

Users can link their Beauty Insider account to the Sephora app inside ChatGPT. Once connected, ChatGPT can access purchase history, skin type preferences, and loyalty status to personalise product recommendations. Loyalty benefits — including free shipping and samples — also carry over into the ChatGPT experience.

Should luxury brands embed in AI platforms like ChatGPT?

Luxury brands face a strategic choice: embed their product knowledge and client data into AI platforms on their own terms, or wait for those platforms to recommend products based on incomplete, uncontrolled information. Brands with strong clienteling data, purchase history, and product expertise are well-positioned to shape how AI represents them — but only if they act before the defaults are set by others.

What is the Agentic Commerce Protocol (ACP)?

ACP is OpenAI's protocol that enables retailers to share product catalogues, inventory data, and services with ChatGPT. Retailers build apps that sit inside ChatGPT, allowing users to browse, get recommendations, and — in future — complete purchases without leaving the platform. Sephora's ChatGPT app was built on ACP.