Agentic Commerce Protocol, One Year Later: What ACP Actually Means for Headless Merchants

A year is long enough to tell whether a new commerce standard is shipping real transaction volume or just press releases. The Agentic Commerce Protocol has a year of history now, and the honest read is that the protocol itself is more durable than the specific product it was built to power. We covered how ACP compares to Shopify and Google's competing Universal Commerce Protocol in our earlier breakdown of the two standards; this piece is about what actually happened to ACP's own flagship use case since then, because the gap between the pitch and the rollout is the part merchants actually need to plan around.

What ACP was supposed to deliver

The original pitch was specific: a shopper would describe what they wanted inside ChatGPT, the agent would search a merchant's catalog, build a cart, and complete the purchase using a delegated payment token, never handing raw card details to the model. OpenAI launched this as Instant Checkout in October 2025 with Etsy and a subset of Walmart's catalog, then expanded toward broader Shopify merchant access. The promise for merchants was a new, large, low friction sales channel that lived entirely inside a conversation they did not control the interface for.

What actually shipped

Instant Checkout did ship, and it did process real transactions through Etsy and Walmart first. But by March 2026, OpenAI told reporters it was shifting its approach: rather than general purpose shopping inside the main ChatGPT chat surface, checkout would move into OpenAI's app ecosystem, where a shopper either pays through a retailer's own app embedded in ChatGPT or gets handed off to the retailer's website to finish the purchase. An OpenAI spokesperson framed it as evolution rather than retreat, saying the company was "evolving how we approach commerce in ChatGPT to better meet merchants and users where they are," and that checkout was "moving to Apps, where purchases can happen more seamlessly." Reporting at the time also noted OpenAI had not yet built systems to collect and remit state sales tax for the transactions Instant Checkout was generating, which is the kind of unglamorous operational gap that explains a pullback better than any strategic narrative does.

What has not pulled back is the standard underneath it. Checkout.com announced on November 25, 2025 that it would support ACP directly, positioning itself specifically toward enterprise merchants rather than the small sellers Etsy represented at launch. Checkout.com's pitch adds the pieces Instant Checkout's first version was thin on: verified merchant onboarding, identity management, and fraud prevention built for a world where the party initiating a purchase is an agent acting for a shopper rather than the shopper directly at a keyboard. Checkout.com also said it is working with Visa, Mastercard, and Google on broader payment standards, which matters because it signals ACP is being treated as infrastructure to build on top of, not a single vendor's walled product.

Why the gap between the pitch and the rollout matters for merchants

The lesson for a merchant evaluating whether to invest engineering time in ACP integration is not that agentic checkout failed. Transaction volume through Instant Checkout was real, if smaller than the initial coverage implied; one contemporaneous analysis pegged Amazon and Shopify's combined share of US ecommerce at roughly half the market, against which a chat native checkout experience with a narrow catalog was never going to look dominant in year one. The lesson is that the layer worth building against is the protocol, cart, feed, orders, authentication, and now identity and fraud tooling through partners like Checkout.com, rather than any single surface implementation of it. OpenAI's own commerce UX inside ChatGPT changed shape twice in a year. ACP as a spec, maintained in the open and now with a second major payments company building enterprise tooling on top of it, did not.

For a merchant on Shopify, this is lower stakes, since Shopify's own Storefront MCP and the competing Universal Commerce Protocol already give a path into agent driven discovery and checkout that does not depend on OpenAI's roadmap; we laid out that architecture in our piece on Shopify's Storefront MCP. For a merchant on Salesforce Commerce Cloud, a headless build on commercetools, or any platform outside the Shopify and Google alliance, ACP through a payments partner like Checkout.com is currently the more direct path to agent initiated checkout, and the enterprise focused tooling Checkout.com is building matters more than the consumer facing chat experience that got most of the press coverage last year.

What to actually build toward

If you are integrating ACP today, build against the cart, feed, orders, and authentication resources directly rather than assuming a specific chat surface, since that surface has already changed once and is likely to change again as OpenAI iterates on where checkout lives inside its app ecosystem. Prioritize the identity and fraud tooling a payments partner provides over the raw protocol spec alone; an agent completing a purchase on a shopper's behalf changes your fraud model in ways a standard checkout flow does not, and that is precisely the gap Checkout.com's enterprise pitch is built to close. And treat ACP and UCP as parallel investments rather than a choice, since your customers are not consolidating onto one assistant; they are spread across ChatGPT, Gemini, and whatever launches next, covered in more detail in our comparison of ACP and UCP.

When this applies to your stack

If your storefront runs outside the Shopify and Google alliance and you have been waiting to see whether agentic checkout was a durable channel before committing engineering time, the signal from this year is to build against the protocol and its enterprise payments tooling, not against any single vendor's consumer facing product, which is the part most likely to keep changing. Contra Collective's headless commerce team has implemented ACP and UCP integrations across Salesforce Commerce Cloud and commerce platform migrations; if your team is deciding where to invest first, that is a scoping conversation worth having before the next protocol revision ships.

FAQ

Is ACP still active after OpenAI scaled back Instant Checkout? Yes. The product surface changed, moving checkout into OpenAI's app ecosystem, but the underlying protocol, including its cart, feed, orders, and authentication resources, remains active and is now supported by payments partners beyond Stripe, including Checkout.com.

Do I need to support both ACP and UCP? If your customers use more than one AI assistant to shop, which is the common case, supporting both protects you from depending on a single vendor's roadmap. Shopify merchants get UCP support by default through Storefront MCP; merchants outside that alliance need to integrate ACP directly or through a payments partner.

Does Checkout.com's ACP support replace the need for Stripe? No. Checkout.com joining as an ACP payments partner adds competition and enterprise focused tooling, such as identity and fraud prevention, rather than replacing Stripe's original role in the protocol.

What should a merchant build first if resources are limited? Start with the identity and fraud tooling a payments partner provides rather than the raw protocol spec, since an agent completing a purchase on a shopper's behalf changes fraud exposure in ways a standard checkout does not.