The Protocol You Pick Matters Less Than Whether You Can Pick At All
The protocol debate is a distraction. The more important choice lives deeper in your architecture.
We recently looked at agentic commerce's governance problem— who's accountable when an agent transacts, who owns the customer relationship once an AI is the one shopping. That question is still unresolved even as traffic scales. But there's a second, more basic question sitting underneath it: can most retailers even act on these protocols yet, regardless of which one eventually wins?
Two standards are racing to become the rails for agentic commerce. Universal Commerce Protocol (UCP), backed by Google and Shopify and co-developed with Target, Walmart, Wayfair, and Etsy, launched in January 2026 to cover the full commerce lifecycle from discovery to checkout. Agentic Commerce Protocol (ACP), built by OpenAI and Stripe, shipped four months earlier, in September 2025, purpose-built for checkout inside ChatGPT.
Trade coverage keeps framing this as a contest—which protocol wins?—but there’s no point to that question. First of all, because for the foreseeable future, both will matter. But secondly, and more importantly, there can’t ever be a winner when most retailers can't act on either option to begin with.
UCP Adoption Rate 2026: Only 22.5% of Top Retailers Are Live
Let’s start with what's actually measured: ReFiBuy and Digital Commerce 360's Q2 2026 AI1000—a quarterly readiness benchmark covering the top 1,000 online retailers—found that only 225 of them, or 22.5%, have a live, detectable UCP endpoint. The index's average readiness score across all 1,000 retailers is 39.7 out of 100, down from 42.0 in Q1. The bar is rising faster than even the top retailers are clearing it.
That's the one hard number in this entire landscape, because UCP is the one standard someone bothered to independently track against a fixed universe of retailers.
ACP Adoption: No Verified Data, Just Vendor Pipeline
ACP has no equivalent. What it has instead is momentum measured in press releases: Etsy sellers are live today; more than a million Shopify merchants, including named brands like Glossier, SKIMS, and Vuori, are "coming soon"; PayPal has said its ACP server will bring "tens of millions" of additional small businesses onto the protocol sometime in 2026. None of those figures carries a date, and none has been independently verified the way the AI1000 verifies UCP.
Meanwhile ChatGPT—the surface ACP was built for—still accounts for the large majority of AI-referred retail traffic across the AI1000, even as Gemini, Claude, and Perplexity pick up share quarter over quarter. The protocol we can least verify at present is the one retailers arguably have the most commercial reason to move fast on.
Why Most Retailers' UCP Adoption Happened by Accident, Not Strategy
Buried in ReFiBuy's own release is the detail that actually explains the 22.5%: most of the retailers with a live UCP endpoint got it "for free." Meaning it’s not something they built, not something they had to put the time, energy, or resources into making happen. Their platform shipped it, and they inherited it.
When you flip that around, the 77.5% without it is no longer a story about retailers who deprioritized AI strategy, but rather a story about retailers whose platforms haven't shipped it yet, and who have no fast path to add it themselves. Which means the actual finding hiding inside a protocol-adoption stat is that readiness is an architecture property, not a strategy decision.
We can infer that readiness in agentic commerce isn't being decided by which protocol wins. It's being decided by whether a retailer's architecture lets them plug into whichever protocol matters on a given surface, and right now, most can't. And it's consistent with what the numbers show: adoption clusters around platform defaults, not around retailer intent.
Why UCP/ACP Adoption is a Composability Problem
Retailers on flexible, API-first stacks can add a new endpoint as soon as a protocol ships, because nothing about their architecture requires waiting on a vendor. Retailers on monolithic platforms wait on a roadmap, and the gap between ACP and UCP is the proof that roadmap cycles can't keep pace. Four months separated the two standards, and a platform on an annual or biannual release cycle will always be behind by definition, not by choice.
Composability isn't about correctly guessing UCP or ACP. It's about not needing to guess at all.
Is Your Commerce Architecture Ready for the Next Agentic Protocol? Why Retailers Need Agentic Commerce Readiness Before Holiday 2026
This isn't a 2027 problem. Salesforce is forecasting that 20% of 2026 holiday ecommerce traffic will come from AI agents and ReFiBuy put it plainly in their own release: "readiness is not a switch retailers can flip in November." The catalogs AI agents will use to research and recommend products this holiday season are being prepared right now, or they aren't being prepared at all.
Retailers spending this quarter debating UCP versus ACP are asking a question their architecture should have made irrelevant months ago. The ones who'll show up ready for not only mainstream UCP and ACP consumer use, but also whatever comes after it, are the ones who built systems that don't require betting on any single protocol in the first place.
Leigh Bryant
Editorial Director, Composable.com
Leigh Bryant is a seasoned content and brand strategist with over a decade of experience in digital storytelling. Starting in retail before shifting to the technology space, she has spent the past ten years crafting compelling narratives as a writer, editor, and strategist.