OpenAI’s retreat makes the contrast with Google sharper and more urgent, and the two could not be more different in their approach.
While OpenAI is stepping back from owning the transaction, Google is accelerating into it. At NRF 2026, Google’s CEO Sundar Pichai announced the Universal Commerce Protocol, an open standard for agentic commerce co-developed with Shopify, Etsy, Wayfair, Target, and Walmart, and endorsed by over 20 companies including Visa, Mastercard, Best Buy, and Stripe. The goal, in Google’s own words, is to take shoppers from “I’m interested” to “I’ve bought it” without ever leaving the conversation.
That is not a roadmap. It is already happening. Google says UCP will soon power shopping inside AI Mode in Search and the Gemini app, letting people buy without ever leaving the conversation. Throughout this process, the retailer remains the merchant of record, allowing them to own and shape the customer relationship, though how much data actually flows back to retailers remains an open question.
OpenAI, for now, sends the customer back to you. Google, if its approach succeeds, never does.
To understand why that contrast matters, it helps to look at the protocols behind each approach. OpenAI built its commerce infrastructure around the Agentic Commerce Protocol (ACP), co-developed with Stripe. ACP is designed as a checkout-centric transaction rail: the agent renders the checkout and relays payment credentials securely. Under the hood, it works through an Agentic Checkout API and delegated payment credentials, with Stripe as its first compatible payment implementation, meaning the merchant retains their own backend and payment flow throughout. It is narrow by design, and that narrowness is actually its strength from a retailer sovereignty perspective. The merchant stays in control of the transaction.
Google’s Universal Commerce Protocol (UCP) is built differently. It is designed to cover the full commerce journey, from discovery and consideration through checkout and post-purchase, all within Google’s own surfaces. Where ACP is a transaction rail, UCP is a platform strategy.
The practical implication of OpenAI’s pullback is that ACP, which looked like it might become the default checkout experience inside ChatGPT, now looks more like background infrastructure powering merchant apps and approved partner flows. That is not nothing, but it is a narrower role than originally anticipated. UCP, by contrast, remains tightly coupled to Google’s first-party distribution surfaces, which gives it more structural momentum right now.
These are two fundamentally different bets on the same future. For retailers, UCP is both an enormous opportunity to reach hundreds of millions of high-intent shoppers and a serious risk of losing direct customer relationships entirely. If Google’s model becomes dominant, retailers risk becoming back-end fulfilment providers for a platform they do not own. The customer relationship, the data, the loyalty mechanics, the repeat purchase logic: all of it migrates to the AI layer.
It is worth noting that the trust hurdle has not been cleared yet. According to the ChannelEngine Marketplace Shopping Behavior Report 2026, only 17% of shoppers say they would be comfortable purchasing directly through an AI assistant, while 43% say they still prefer to buy via a marketplace or brand site. But the infrastructure is being built now regardless, and that is what matters strategically
That is not a hypothetical. It is a decision retailers need to be thinking about today, while they still have a seat at the table.
