A shopper asks ChatGPT to find and buy a specific pair of running shoes in their size, under a budget, with two-day delivery. In late 2025 that request would have ended in a list of links back to Google. By mid-2026 it can end in a completed purchase, inside the chat window, without the shopper ever opening a browser tab. That shift didn't happen because AI models got smarter at reading web pages, it happened because two rival open standards, the Agentic Commerce Protocol and the Universal Commerce Protocol, gave AI agents a structured way to actually transact with a store instead of just describing one.
What is the Agentic Commerce Protocol, and how does it work?
The Agentic Commerce Protocol (ACP) is an open standard built by OpenAI and Stripe, first launched to power Instant Checkout inside ChatGPT starting with US Etsy sellers and expanding toward over a million Shopify merchants. Structurally, ACP defines a small set of composable API building blocks a merchant's backend has to expose: a product feed the agent can browse, a checkout session the agent can create, update and complete with cart and fulfillment details, and a delegated payment step that passes a payment token between the buyer, the agent and the business without the merchant ever seeing raw card data.
The full specification is maintained jointly by OpenAI and Stripe but is deliberately payment-processor agnostic, a store that doesn't use Stripe can still implement ACP against its own payment stack. That's the detail that matters for adoption: ACP isn't a Stripe lock-in play, it's an interaction contract any checkout system can speak.
What is the Universal Commerce Protocol, and how is it different from ACP?
The Universal Commerce Protocol (UCP) launched on January 11, 2026, unveiled by Google at the National Retail Federation conference and co-developed with Shopify, Etsy, Wayfair, Target and Walmart, with more than 20 endorsing partners including Visa, Mastercard and American Express. Where ACP is scoped tightly to the checkout moment, UCP is built to cover the whole shopping journey, discovery, buying and post-purchase support, and it's wired directly into Google AI Mode, the Gemini apps and Microsoft Copilot Checkout across more than 20 participating retailers at launch.
In practice, the two protocols solve overlapping problems from different starting points, ACP grew out of a single checkout partnership and generalized outward, UCP was designed from the start as a multi-platform standard with search, ads and merchant infrastructure attached. For a merchant, the practical question isn't which philosophy is correct, it's which surfaces your customers actually use, ChatGPT traffic argues for ACP, Google AI Mode and Gemini traffic argues for UCP, and most mid-size stores will end up needing to support both.
What does a store's backend actually need to expose to be reachable by either protocol?
- A structured, machine-readable product feed with accurate attributes, real-world descriptions, current pricing and live inventory, an agent can't complete a checkout on stale stock data.
- An API-driven checkout endpoint that can create and update a session and return structured delivery and fulfillment data, not a checkout flow built only for a human clicking through form fields.
- A delegated payment integration, via ACP's payment token model or a UCP-compatible processor, so the agent can pass payment authorization without the merchant handling raw card details outside its existing PCI scope.
- Clear, structured shipping, return and fulfillment policies the agent can read programmatically, since an agent won't call support to ask what the return window is.
Does agentic commerce actually convert, or is it still a novelty?
It's a fair question given how early both protocols are, ACP has been live in production only since late 2025 and UCP since January 2026. But the discovery-phase data already lines up with a real behavior shift: AI-referred shoppers complete purchases at meaningfully higher rates than shoppers arriving through traditional search, because by the time an AI agent hands off to checkout, most of the comparison and decision work is already done. That's a related but separate problem from checkout integration itself, getting an AI agent to find and recommend your product in the first place, which we cover in our guide to optimizing product pages for AI shopping agents.
The practical takeaway for a founder deciding whether to prioritize this now: agentic commerce doesn't favor the biggest brand in a category, it favors whichever store has the cleanest, most complete structured product data an agent can actually act on. That's a smaller lift than most teams assume, and it compounds with the GEO and structured-data work most businesses should already be doing to be visible to AI answer engines at all.
How AIBOOTSTRAPPER helps
AIBOOTSTRAPPER doesn't yet have a published case study specifically on agentic checkout integration, this is a genuinely new surface as of 2026 and we're honest about that rather than stretching an unrelated result to fit. What we do bring is the same discipline behind the GEO-optimized product and marketing sites we've shipped for clients like AudioBolo and Expensorr, building the structured data, clean architecture and API-first backend that both search engines and AI agents can actually read and act on.
If you're evaluating whether ACP, UCP or both are worth building against for your store, book a call and we'll help you scope what your product feed and checkout flow actually need.
Want this done for you?
Book a free strategy call and we'll show you how to build and market your business with AI.
Sources and further reading
- 1.Stripe — Agentic Commerce Protocol documentation
- 2.Stripe Newsroom — Stripe powers Instant Checkout in ChatGPT
- 3.Agentic Commerce Protocol — open specification (GitHub)
- 4.Google Developers Blog — Under the Hood: Universal Commerce Protocol (UCP)
- 5.Google Ads & Commerce Blog — New tech and tools for retailers in an agentic shopping era