Beauty Just Picked Three AI Surfaces. Every Other Retailer Is Watching.

If you want to know which AI surface is going to win consumer commerce, the beauty industry just ran the experiment for you.

Sephora launched an app inside ChatGPT, now piloting in the United States, connected to its 80 million global Beauty Insider members. Fenty Beauty built Rose Amber, a WhatsApp advisor, as its first formal U.S. partnership with Meta's messaging app. Ulta Beauty rolled out agentic commerce on Google Search AI Mode and the Gemini app, running on Google's Universal Commerce Protocol, with 46 million loyalty members plugged in.

Same vertical. Same quarter. Three different surfaces.

The beauty industry just ran a live A/B/C test on the future of consumer commerce. The result of that test will reshape every other retail category.

What Each Giant Shipped

Sephora on ChatGPT. The Sephora app inside ChatGPT is personalized by Beauty Insider account data. Users get beauty advice, curated recommendations, and product comparisons. Checkout is not live yet but is coming. OpenAI's recent pivot toward merchant-controlled checkout means the transaction will land on Sephora's own payment rails when it does go live, not OpenAI's closed loop.

Fenty Beauty on WhatsApp. Rose Amber is an AI advisor built on WhatsApp. The advisor runs skin concern diagnostics, product quizzes, and surfaces recommendations across Fenty Beauty, Fenty Skin, and Fenty Hair. L'Oréal already generates 20 percent of its Brazilian direct-to-consumer online sales through WhatsApp conversational commerce and sees abandoned cart recovery rates six times higher than email. Fenty is betting that pattern travels to the US. In-app purchases are "coming," which is WhatsApp's standard framing for the next quarter.

Ulta Beauty on Google. Ulta is running two surfaces in parallel. The Ulta AI assistant lives on Ulta.com and the Ulta app. The agentic commerce layer sits on Google, specifically Search AI Mode and the Gemini app. Both are built on Gemini Enterprise for Customer Experience and the Universal Commerce Protocol, both of which Google unveiled at Cloud Next '26 a day earlier. Ulta is the first major US retailer to deploy the UCP at this scale.

Three companies. Three surfaces. Zero overlap.

Why Beauty Went First

There are four reasons it was beauty.

The products are personal and hard to choose. A foundation shade is not a commodity. Color matching, skin type, ingredient sensitivities, and fit all matter. Shoppers already spend time on the decision, which makes them willing to talk to an AI that actually helps them decide. Commerce categories where "any option works" do not have the same pull.

The loyalty programs are deep. Ulta has 46 million members. Sephora has 80 million. That is a data layer large enough to personalize in a way the AI model cannot do on its own. The agent is cold without it. With it, the agent gives better recommendations than a store associate could.

The margin is there. Beauty margins are high enough that a full conversational experience with an AI agent is cost-justified per session. Grocery and consumer electronics do not have that headroom, which is why they are moving slower.

The demo is good. The CEO of every major beauty brand can pull out their phone in a board meeting and show the AI advisor working in 30 seconds. That is not true for a mortgage product or a B2B procurement agent. Demo quality accelerates internal approval.

Beauty is not leading because it is disruptive. It is leading because the shape of the product fits the shape of the agent.

The Three Surfaces, Compared

Each brand picked a different surface for a reason.

ChatGPT has 800 million weekly active users, most of them already asking AI questions in plain text. The Sephora integration is a discovery play. The user is not going to ChatGPT to shop. She is going to ask a question. Sephora sits there when beauty comes up. Low intent, high volume.

WhatsApp is a different shape. It is the messaging app people already use to talk to businesses. In emerging markets it already processes billions of business conversations daily. Fenty is not competing for attention at the top of a search. It is sitting inside a chat thread the user already has open. High intent when it triggers, lower volume, much higher conversion.

Google is the incumbent. Search AI Mode and Gemini app are the two surfaces Google is pushing hardest into commerce. Ulta gets the benefit of Google's distribution and the UCP's agent-to-merchant integration. The trade is that Google keeps the top of funnel. Ulta owns the conversion.

Each surface has a different read of the customer relationship. ChatGPT: the AI is the audience. WhatsApp: the AI is a feature of an existing conversation. Google: the AI is a new way to do the thing people already use Google for.

If you are a retailer choosing where to deploy, the question is not which one is biggest. The question is where your customer already is when she wants what you sell.

The Infrastructure Beneath

The shipping was fast, which means the infrastructure exists. Three specific pieces.

OpenAI's Apps SDK lets brands like Sephora build inside ChatGPT without the old chatbot constraints. The catalog flows in. The conversation flows out. The payment is handed back. That is the merchant-native pattern we mapped in our agentic commerce stack piece.

Meta's WhatsApp Business Platform has been building commerce rails for five years. Fenty is not the first brand on it. Fenty is the first brand that meets the bar for the next phase, which is AI-native conversational commerce rather than the older template-driven flows.

Google's Universal Commerce Protocol is the new entrant. UCP is how an agent built on Gemini can take an action against a merchant's catalog and checkout flow without a bespoke integration. Ulta is the showcase. If UCP works for Ulta, every other retailer with a Google Cloud relationship gets pushed toward it. The Cloud Next '26 announcement gets a second life as a case study.

Three surfaces. Three protocols. All of them expecting the merchant to own the customer record and the transaction.

What Other Verticals Should Read From This

Retailers in other categories should take three things from this week.

First, the protocols are stable enough to build on. Twelve months ago, committing to a surface felt premature because the rails might change. This quarter the rails shipped. A fashion retailer, a furniture brand, a specialty grocer can now pick an agent surface and not expect to rebuild the integration inside a year.

Second, the loyalty database is the moat, not the catalog. Every beauty brand that shipped this week leaned on member data. The model is only as good as the personalization layer underneath. If your loyalty data is thin or your customer record is fragmented, the AI advisor has nothing distinctive to say. Fix the data layer before you pick the surface.

Third, pick one surface to start. All three of the beauty giants picked one primary surface and one secondary. None of them tried to ship on all three at once. That is a deliberate choice. The operational cost of running a ChatGPT app, a WhatsApp advisor, and a Google agentic integration in parallel is higher than it looks. Pick the surface that matches the customer's buying motion and ship well on it.

What To Watch

Four signals over the next two quarters.

First, the conversion data. None of these companies has published actual revenue attribution yet. When Sephora, Fenty, or Ulta share a quarterly number tied to the AI surface, the whole vertical moves. When they do not, skeptics stay skeptical.

Second, which retailer shows up on a second surface. If Sephora extends from ChatGPT to WhatsApp, or Ulta adds a ChatGPT app on top of Google, we learn whether the surfaces are complementary or substitutes. The economics change depending on the answer.

Third, whether a fashion giant moves. Beauty went first because the product shape fits. Fashion has similar economics: personalization matters, margins are there, the loyalty programs are deep. If Zara, H&M, or Uniqlo ship a surface play inside three months, the vertical is validated beyond beauty.

Fourth, how the incumbent search channels respond. Google is already inside the Ulta story through UCP. The interesting question is what happens to paid search spend. If Ulta shifts a meaningful share of its Google Ads budget into UCP-driven conversations, Google's ad business gets a new product line. If it does not, the investment signal is weaker than Cloud Next suggested.

If beauty just ran the experiment, which vertical runs it next, and which surface wins when the product shape is different?

Charlie Major is a Product Development Manager at Mastercard. The views and opinions expressed in Major Matters are his own and do not represent those of Mastercard.