TLDR;
OpenAI launched Instant Checkout for selected US merchants, letting people buy inside ChatGPT. It later shifted towards product discovery and checkout on merchants' own websites. Retailers should now focus on accurate product information, a smooth route to purchase and whether the channel makes money.
What happened
OpenAI introduced Instant Checkout and the Agentic Commerce Protocol (ACP), initially supporting eligible US purchases from selected merchants. The announcement moved ChatGPT beyond product recommendations into transactions; in March 2026, OpenAI subsequently shifted its emphasis toward product discovery and merchant-owned checkout.
Why it matters
A purchase initiated inside an assistant crosses more systems than a recommendation. Product identity, availability, payment, order creation and customer support must remain consistent. A successful interface demonstration tells you little about retries, fulfilment failures or the cost of servicing orders acquired through the new channel.
Buying inside an assistant compresses the gap between recommendation and transaction. That can reduce the effort required to complete a purchase, but it also moves part of the selling experience away from pages the merchant controls. A customer may see fewer opportunities to check delivery conditions, compare variants or understand a return restriction. The merchant still carries the consequences when the order arrives incorrectly or the customer expected something different. The launch therefore raised a commerce operations question as much as a visibility question. Selected merchant participation also matters: a new checkout format should not be described as a route every retailer can immediately use. The subsequent move towards merchant checkout reinforces the need to build around reliable product and order information, rather than assume one interface will remain the permanent destination.
How your brand can benefit / be affected
Assess eligibility and start with a limited product range. Reconcile feed identifiers with order identifiers, confirm current prices and stock, and test duplicate submission and cancellation. Keep customer communications, returns and support ownership explicit.
Evaluate incremental contribution after fees, returns and fulfilment costs. The later move toward merchant checkout changes the implementation priority: maintain strong product discovery and a reliable hand-off, rather than treating the original in-chat transaction model as a permanent requirement.
Review the buying journey as a series of promises that must agree. The product recommended, selected variant, quoted total, available stock and delivery expectation should all refer to the same purchase. Write down which system is authoritative for each promise and how changes reach any external shopping interface. A common failure to investigate is a product that remains attractive in discovery while the relevant size or configuration is unavailable. The brand should decide what the customer sees in that case before optimising for more exposure. Increased demand is only useful if the handoff can support the purchase the assistant has helped the customer choose.
For any eligible integration, test fulfilment and support as carefully as payment. Check order confirmation, cancellation, refunds and the information available to customer service when the buyer refers to an assistant conversation. Compare the contribution margin and support burden with orders through existing channels. For brands without access, improve product descriptions and merchant checkout rather than delaying work until an integration becomes available. Clear variant information, current availability and transparent conditions remain useful through a discovery-led handoff. Treat platform availability as a planning input and revisit it periodically; avoid a large programme that depends on a launch configuration already superseded by later developments.
News date: 29 September 2025. Editorial review: 16 September 2026. Analysis includes subsequent developments where stated.