What Is Instant Checkout?
Instant Checkout is an in-chat purchase flow that lets a customer move from product discovery to payment confirmation without leaving the ChatGPT conversation. OpenAI introduced it in its "Buy it in ChatGPT" launch on September 29, 2025, describing it as the first step toward ChatGPT helping people not only find products, but also buy them directly inside the chat interface.
Why Instant Checkout exists
Classic e-commerce splits the journey across multiple surfaces: search results, product pages, cart, checkout, email confirmations. In agent-led shopping, the user's "interface" becomes the conversation, so the commerce stack needs a standard way to:
- Show accurate products (price, availability, variants, policies)
- Create and update a checkout as the user makes decisions
- Complete payment securely, while keeping merchant systems authoritative
Instant Checkout is built to reduce friction and improve conversion by keeping the user in one place while the agent coordinates the transaction steps in the background.
The plumbing behind Instant Checkout: ACP
Instant Checkout is powered by the Agentic Commerce Protocol (ACP), an open standard co-developed by Stripe and OpenAI. Stripe announced ACP the same day as Instant Checkout, positioning it as a shared protocol that businesses can implement once to support agent-driven commerce experiences.
At a practical level, ACP standardizes a checkout lifecycle through a small set of endpoints. Stripe's specification focuses on methods and data structures for creating, updating, and completing checkout flows, with cancellation and state updates included as part of the protocol's integration patterns.
OpenAI publishes a complementary Agentic Checkout spec that emphasizes the same operational requirements: implement create, update, and complete endpoints, and return a rich, authoritative cart state on every response (items, totals, taxes, shipping, discounts, status).
Payments: Delegated Payments and Stripe's Shared Payment Token
Instant Checkout uses a delegated payments model. OpenAI describes a Delegated Payment spec where OpenAI securely shares payment details with the merchant or the merchant's designated payment service provider (PSP), and the merchant processes the transaction in the same manner as their other orders.
OpenAI's commerce "key concepts" explain the flow more concretely: ChatGPT prepares a one-time delegated payment request with a maximum amount and expiry; the payload goes to a trusted PSP; the PSP returns a payment token; the token is passed to the merchant to complete the payment. OpenAI notes that Stripe's Shared Payment Token (SPT) is the first Delegated Payment Spec-compatible implementation.
Stripe's Instant Checkout announcement also highlights SPT as a way for sellers to participate without major changes to their stack, even if they are not using Stripe as their primary processor.
What merchants control (and what the agent controls)
A common concern is "Does the agent become the storefront?" With Instant Checkout, the conversational agent is the customer-facing UI, but the merchant remains authoritative for:
- pricing and inventory rules
- shipping options and fees
- order creation and fulfillment
- post-purchase support and returns policies
The protocol is designed so the merchant backend returns the canonical checkout state, and the agent guides the user based on that state.
Rollout and ecosystem signals
At launch, Instant Checkout rolled out to U.S. users with early commerce partners. Reuters reported OpenAI's partnerships with Etsy and Shopify, with an initial single-item purchase flow and plans to expand features and regions over time.
The bigger signal for product and commerce teams is that Instant Checkout changes what "being discoverable" means. It is no longer just about ranking on a marketplace or optimizing on-site search. It is about providing AI systems with data and endpoints that are:
- structured
- current
- semantically searchable
- transaction-ready
What you need before you go live
Instant Checkout is not something you bolt on at the very end. It depends on two layers being production-ready:
AI-ready catalog data
Even the best checkout flow cannot compensate for incomplete or ambiguous product information. You need clean titles, strong attributes, consistent variants, reliable availability, and channel-ready policies. OpenAI's commerce specs and guidance are explicit that the agent relies on structured outputs and authoritative states.Checkout endpoints that behave predictably
You must handle create, update, and complete operations, respond with a full cart snapshot, and support cancellation and order updates so the agent can keep the user experience consistent through retries and edge cases.
Why testing matters before publishing feeds
Stripe and OpenAI's approach makes one thing clear: agentic commerce rewards merchants who treat product data quality and checkout correctness as first-class systems. Before you publish feeds into agent ecosystems or enable a live Instant Checkout flow, you should validate how models interpret your products, and whether your catalog supports accurate retrieval and confident recommendations.
That is where tools like ventic.ai fit naturally: test your catalog and supporting materials across AI providers, identify gaps (missing attributes, weak taxonomy, unclear policies), and ship an AI-ready product knowledge layer through an API before it becomes a revenue-critical channel.