Short answer: if you sell on Shopify, ChatGPT Instant Checkout is likely already available to you or close to it, and turning it on costs little more than approving the integration. If you sell anywhere else, setting it up yourself right now means building a product feed to OpenAI’s Agentic Commerce Protocol and wiring in Stripe’s shared payment token flow, which is real engineering work for a payoff that’s still mostly speculative. Most small merchants outside Shopify should watch this closely and prepare the groundwork, but shouldn’t drop everything to build it this month.
Here’s what Instant Checkout actually is, what it requires under the hood, and how we’re advising clients to prioritize it against everything else competing for their attention.
What Instant Checkout Actually Is
OpenAI launched Instant Checkout inside ChatGPT in late 2025, letting a shopper complete a purchase without leaving the chat window. Instead of ChatGPT linking out to a retailer’s site, the product, price, and payment happen inside the conversation, and the order gets placed through the merchant’s existing systems on the back end. Etsy and Shopify were the first major integrations, and the market noticed immediately: Etsy’s stock jumped double digits the day the feature was announced, which tells you how seriously investors think this could reroute retail traffic.
The mechanism behind it is called the Agentic Commerce Protocol, or ACP, which OpenAI built together with Stripe. It’s worth understanding the protocol separately from the ChatGPT feature, because ACP is open and other AI assistants and retailers can adopt it independently of OpenAI.
How the Agentic Commerce Protocol Actually Works
Strip away the branding and ACP does three jobs:
- It standardizes the product feed. Merchants expose their catalog, pricing, availability, and shipping details in a format an AI agent can read and reason about reliably, similar in spirit to a Google Shopping feed but built for conversational, not search-grid, presentation.
- It hands off payment through a shared token, not a redirect. Stripe generates a payment credential that the AI application passes to the merchant at the moment of purchase, so the merchant processes the transaction through their own existing payment setup rather than the AI platform holding funds or acting as a reseller.
- It keeps the merchant as the seller of record. This matters more than it sounds: you keep your customer relationship, your fulfillment obligations, and your existing tax and returns processes. ChatGPT is the storefront interface for that one transaction, not a new marketplace you’re listing on.
Google is building a comparable protocol of its own for agentic checkout across its properties, which is a strong signal this isn’t a single-platform experiment. Expect more than one AI assistant to want a version of your product feed within the next year or two.
Two Paths In: Shopify’s Native Support vs. a Custom ACP Feed
For most small businesses, the practical decision comes down to which of these two paths you’re on.
| Shopify merchants | Everyone else | |
|---|---|---|
| What’s required | Enable the integration through Shopify’s existing tools; the platform handles feed formatting and the payment handoff | Build or source a product feed that meets the ACP specification, and integrate Stripe’s shared payment token flow yourself or through a developer |
| Rough effort | Low; mostly a configuration and approval step | Meaningful development work, closer to a payment integration project than a marketing task |
| Who should move now | Any Shopify store with decent product data hygiene already in place | Larger catalogs with in-house development resources, or businesses for whom being an early, citable example matters strategically |
| Who should wait | Stores whose product data is messy; fix that first | Small catalogs on WooCommerce, Squarespace, or custom builds without dev capacity to spare right now |
What an AI Checkout Feed Needs That Your Google Shopping Feed Doesn’t
If you already run a Google Shopping or Meta catalog feed, you have a head start but not a finished asset. The ACP product feed specification asks for the same core fields, product ID, title, price, availability, images, but adds requirements around real-time inventory accuracy and structured return and fulfillment policy data, because the AI agent is representing your policies directly to the shopper during the conversation rather than the shopper reading your returns page themselves. A feed with stale stock levels or vague shipping terms will produce bad or contradicted answers inside the chat, which is a worse outcome than simply not being listed.
Is It Worth Setting Up Now?
This is genuinely early. Adoption on the shopper side is still small relative to normal search and social traffic, and the ecosystem is moving fast enough that specifications built today may need rework in a year. That argues for patience. Three things argue against waiting too long:
- Product data cleanup pays off regardless of which protocol wins. Accurate, structured inventory and policy data helps your existing Shopping feed, your GEO visibility, and any future agentic commerce standard. This is not a sunk cost even if ACP specifically doesn’t become the dominant standard.
- Category concentration matters. If you sell the kind of product people research conversationally, gifts, home goods, anything with a “help me choose” quality, you’re more exposed to this shift than a business selling commodity items people buy by exact SKU.
- Being an early, well-structured example can matter for citation. The businesses that get held up as clean implementations tend to get referenced and recommended more, in the same way early adopters of structured data benefited disproportionately when AI Overviews first rolled out.
Checkout Is Downstream of Being Found
It’s worth saying plainly: Instant Checkout only matters if the AI agent recommends your product in the first place. Enabling checkout without also earning a mention inside ChatGPT’s answers is building a door nobody walks through. We’ve written separately about how to get found when AI agents do the shopping, and that visibility work, clean product pages, structured data, genuine reviews, comes before checkout infrastructure on almost every client’s priority list.
A Practical First Step
If you want to be ready without overbuilding: audit your product data for the same things ACP cares about, accurate real-time stock, clear return windows, complete shipping information, before you touch any integration work. That’s useful today regardless of what happens with agentic checkout, and it’s the actual bottleneck we run into with clients more often than the technical integration itself.
Do I need Instant Checkout if I’m not on Shopify?
Not urgently. Prioritize product data quality first; the integration work can follow once the ecosystem and your bandwidth both settle.
Is this the same as being cited in ChatGPT’s answers?
No. Citation and recommendation happen through the same fundamentals as any GEO work, structured content, clear answers to real questions, and credible sourcing. See our notes on optimising product pages for AI shopping answers for that side of it specifically. Checkout is what happens after an agent has already decided to recommend you.
What about Google’s own version?
Google has signaled it’s building a parallel agentic checkout capability for its own ecosystem. Expect more convergence than divergence over time, and treat clean, structured product data as the investment that pays off across whichever protocols end up mattering. For the browser-agent side of that shift more broadly, we covered how tools like Claude, Comet, and Atlas are already researching your business before a shopper ever reaches checkout.