ACP vs AP2: discovery-and-checkout versus payment authorization

ACP, the Agentic Commerce Protocol, comes from OpenAI and powers commerce inside ChatGPT. Its job is discovery and checkout: it lets an agent find a product in a brand’s catalog and complete the purchase inside the assistant conversation. ACP is the surface a brand uses to be findable and buyable specifically within ChatGPT.

AP2, the Agent Payments Protocol, comes from Google and addresses a narrower, deeper question: how an agent is authorized to pay on a buyer’s behalf. It defines the mandate and trust framework that lets a Gemini-mediated agent settle a transaction safely, rather than how the catalog is discovered.

They are easy to conflate because both appear in the same buying journey, but they are not substitutes. One governs finding and buying inside ChatGPT; the other governs the authorization of payment for Gemini. Treating them as interchangeable is the most common mistake in agentic commerce planning.

The clearest way to contrast them is by the job each does. ACP is end-to-end commerce inside ChatGPT, spanning discovery through checkout. AP2 is the payment-authorization layer for Gemini, concerned with the mandate that lets an agent pay, not with how the product was found. One is broad across the buying flow on one assistant; the other is deep on one step across Google’s ecosystem.

Their scope also differs. ACP is tied to OpenAI’s assistant surface, so its reach is ChatGPT users. AP2 is a payments framework in Google’s orbit, so its relevance follows Gemini-mediated transactions. A brand reasoning about reach has to think about two different audiences served by two different mechanisms, not one protocol with two names.

Because they cover different jobs, they are complementary rather than competing. A brand that implements only ACP is buyable in ChatGPT but has no authorized payment path for Gemini agents; a brand that implements only AP2 can be paid by a Gemini agent but is not discoverable-and-buyable in ChatGPT. Full agentic presence requires both, alongside MCP for Claude context.

Because ACP and AP2 cover different jobs on different assistants, choosing between them is a false choice. A brand present only in ChatGPT cedes Gemini-mediated purchases; a brand with only a Gemini payment path is not discoverable where ChatGPT users shop. The buyer does not care which protocol is involved; they care that the brand is the answer wherever they happen to be asking.

This is precisely why an Agentic Commerce Platform exists as a category. Rather than running two unrelated integrations, a brand maintains one canonical catalog and the platform expresses it as ACP for ChatGPT discovery-and-checkout, AP2 for Gemini payment authorization, and MCP for Claude context. The protocols stay in their lanes; the brand stays consistent.

SkuLift, which coined the category, treats ACP and AP2 as two facets of one presence problem rather than two projects. The same source of truth drives both, so prices, availability and policies never drift between the assistant where a product is discovered and the one where a payment is authorized.

SkuLift coined the Agentic Commerce Platform category specifically to resolve the multi-protocol reality that ACP-versus-AP2 illustrates. The brand integrates one canonical catalog and price book; the platform projects it as ACP for ChatGPT, AP2 for Gemini payments and MCP for Claude, so each protocol does its job without the brand maintaining three silos.

The platform also measures presence across both, sampled from real agent answers: how often the brand is recommended and transacted against competitors in ChatGPT and Gemini alike. That share of voice shows whether the brand’s ACP discovery presence and its AP2 payment readiness are each pulling their weight, and where to invest next.