Who owns the customer when an AI agent shops? Brand control in agentic commerce

When an AI agent shops, control of the customer relationship is split: the AI platform owns the conversation and the recommendation, while the retailer typically remains merchant of record and keeps fulfilment, returns and service. For fashion brands, the risk is losing visibility, data and presentation at the moment of choice; the levers to keep control are product data, owned relationships and deliberate decisions about which agents and checkout models to support.

Who is the merchant of record in agentic commerce?

The main protocols are explicit on this point. The Agentic Commerce Protocol, developed by OpenAI and Stripe, states that businesses remain merchant of record and control what they sell, how products are presented and how orders are fulfilled. Google's Universal Commerce Protocol states that businesses remain merchant of record and keep ownership of customer relationships, and Google says retailers remain seller of record for checkout in AI Mode and the Gemini app.

Being merchant of record, however, is not the same as owning the relationship. The platform decides which products are shown, in what order and with which words, and it holds the conversational history that reveals the shopper's intent.

What exactly can brands lose when agents shop?

McKinsey's October 2025 report warns that businesses that hesitate risk losing ground as agents become the new gatekeepers of commerce, and that retail media networks could suffer as shoppers bypass advertising channels. It also notes that traditional search optimisation may matter less when agents mediate discovery.

Read also

How Levi Strauss & Co. uses AI: the data programme and its lessons

Can brands block AI shopping agents?

The legal answer is still forming. Amazon sued Perplexity in November 2025 over its Comet browser agent. In March 2026, US District Judge Maxine Chesney granted a preliminary injunction barring Comet from accessing password-protected parts of Amazon's site to shop for users, finding that it acted with the user's permission but without Amazon's authorisation. In August 2026, the Ninth Circuit vacated that injunction, finding Amazon unlikely to succeed on its computer fraud claims on the current record because the user, not Perplexity, was accessing Amazon's systems. The court stressed that its decision was limited to the facts at hand, and the case returned to the district court.

For fashion retailers, the lesson is that blocking agents outright is legally uncertain and commercially risky, since legitimate agents may represent real customers. A more practical route is to distinguish legitimate agents acting for real customers from abusive automation, and to set clear terms for how agents may use the site.

Are platforms giving retailers more control?

There are signs of a negotiation. OpenAI launched Instant Checkout inside ChatGPT in September 2025. In March 2026, CNBC reported that OpenAI was moving away from it towards retailer-run apps inside ChatGPT, with shoppers sent to the retailer's own site to complete purchases. Walmart, which planned to bring its own Sparky assistant into ChatGPT, called Instant Checkout a very temporary moment in time. Google, for its part, launched a branded Business Agent on Search in January 2026 that answers product questions in a brand's own voice, with launch partners including Reebok and Poshmark.

Which levers do fashion brands still control?

1. Product data quality: the agent can only describe what it can read. Complete attributes, fit notes and accurate stock decide whether and how products are recommended.
2. Distinctive content: original editorial, styling advice and verified reviews give agents material that competitors cannot copy.
3. Owned relationships: loyalty programmes, member pricing and account benefits give shoppers a reason to complete purchases with the brand directly.
4. Checkout model choices: brands can decide per platform whether to allow in-agent checkout, require hand-off to their site, or build an app inside an assistant.
5. Brand-voiced agents: offering your own agent, on your site or in external platforms, keeps some of the conversation in your tone.
6. Contract terms: read platform terms on data sharing, fees and presentation before enabling features.

Read also

How does LVMH organise AI? The AI Factory, maison agents and the Innovation Award

What should marketing and e-commerce leaders do now?

Assign ownership for agent visibility, just as companies once assigned ownership for search. Measure how often your products appear in assistants for core queries, how accurately they are described and how AI referral traffic converts. Review brand guidelines so they include rules for how product facts and claims should be stated in structured data. And involve legal early: as merchant of record, the brand carries responsibility for what is sold, even when an agent made the recommendation.

There is also an organisational question. Agent visibility cuts across marketing, e-commerce, product data, legal and IT, and without a clear owner it tends to fall between teams. Some retailers are creating small cross-functional groups that review agent channels quarterly: which platforms send traffic and sales, how products are described, which terms have changed, and which experiments to run next. That cadence matches the pace at which platform models have been changing and keeps decisions about control deliberate rather than reactive. It also gives leadership a regular, evidence-based view instead of relying on headlines about agentic commerce.