Sephora launches on ChatGPT: the Agentic Commerce shift is here

Announced on March 24 at Shoptalk Spring in Las Vegas, Sephora’s app inside ChatGPT is not just another retail integration. It is the first major proof of concept for what comes after the checkout pivot – a model where AI handles the entire discovery and recommendation layer while the retailer keeps the transaction, the data, and the customer relationship.

With 880 million monthly ChatGPT users on one side and 80 million Beauty Insider members on the other, the scale of this bet is hard to dismiss.

The experience is conversational, and surprisingly specific. A US-based ChatGPT user can type “Sephora, help me find a foundation for dry skin” and receive curated product recommendations drawn from their Beauty Insider profile (purchase history, skin type, past preferences). Loyalty rewards and member benefits are already live. In-app checkout is not, but it’s coming.

Three things make this different from the generic product recommendations ChatGPT already surfaces for Shopify merchants.

The first is first-party data depth. Beauty Insider profiles give the AI access to real purchase patterns and biometric preferences, not just product metadata. This pushes recommendations from broad trend-matching toward individual predictive analysis, a level of personalisation that no search engine or affiliate link can replicate.

The second is multimodal skin diagnosis. Using GPT-5.4’s vision capabilities, the tool can analyse skin tone and condition from a selfie in real time. A service that used to require a visit to a flagship store counter is now instant, free, and frictionless.

The third is the business case behind it. By steering customers toward products that genuinely match their skin, Sephora anticipates a 30% reduction in return rates. That turns precision advice into a direct profitability lever, not just a nicer customer experience.

Sephora is the headline, but the underlying infrastructure shift is broader. Shopify merchants are now discoverable inside ChatGPT by default – no setup, no opt-in required (more on Agentic Storefronts). Products surface with live pricing and inventory whenever a user asks a relevant shopping question. When they tap “buy,” the merchant’s storefront opens in-app. The merchant keeps the checkout, the data, and the customer relationship.

A few details are worth paying attention to. Rankings are fully organic, not paid. Clean, descriptive product titles win – “32 oz Insulated Stainless Steel Water Bottle – Keeps Drinks Cold 24 Hours” gets recommended, “Water Bottle” gets ignored. AI-referred shoppers convert at 9× the rate of social traffic and 31% higher than other referral sources, though those figures predate OpenAI’s recent checkout pivot.

On the economics side, the split is revealing: 4% to OpenAI, 0% to Google AI Mode. Both pull from the same Shopify catalog feed. For merchants, that makes feed quality – rich titles, real-time pricing, complete catalogue coverage – the new competitive advantage in a channel they didn’t know existed months ago.

When OpenAI confirmed it was walking back native checkout, the question was what would replace it. Sephora’s app provides the clearest answer yet. AI handles discovery, intent, and personalisation. The transaction closes in the retailer’s own environment. OpenAI is no longer competing with Stripe. It is competing with Google Shopping and affiliate networks, discovery as the core value proposition.

That repositioning is commercially meaningful, but it’s also incomplete. No public statistics exist yet on engagement rates for apps inside ChatGPT. The channel is simply too new. And for Sephora specifically, in-app purchase remains a “future update” — the current experience is recommendation-first, not transaction-ready.

The only about twelve Shopify merchants who ever integrated with Instant Checkout are a useful reminder: building infrastructure and changing consumer habits are two entirely different problems. Sephora’s Beauty Insider linkage – which brings an existing, loyal audience into the AI layer rather than trying to build one from scratch – may turn out to be the smarter model.

The strategic implications are clear even if the technology is still maturing. Agentic presence is becoming the new SEO – if ChatGPT doesn’t recommend your product, you simply don’t exist in that channel.

Enable Agentic Storefronts in your Shopify admin if you haven’t already, and audit your product titles for AI readability.

Treat feed quality as a ranking factor, not an afterthought – the fundamentals of good feed management now determine whether AI surfaces your product or your competitor’s.

Protect your first-party data: brands racing onto external AI platforms risk ceding the customer relationship to the intermediary. Sephora’s approach – connecting an existing loyalty programme rather than handing over raw data – is a hedge worth studying.

And watch the Agentic Commerce Protocol: open-source, Stripe-backed, and likely to mature regardless of OpenAI’s own commercial trajectory, it may become the infrastructure layer that outlasts any single platform’s strategy.

Two weeks ago, OpenAI’s vision of becoming the world’s biggest shop ran into the oldest problem in commerce: people browse, they don’t buy. Sephora’s response is pragmatic and potentially more durable. Don’t fight the habit – meet customers where they already convert, and use AI to make sure they arrive with the right product already in mind. The most valuable moment in commerce may not be the payment. It may be the recommendation that made the payment inevitable.