For the past year, the pitch for "agentic commerce" has been that AI assistants would soon buy things on your behalf, start to finish, without you ever opening a browser tab. The most visible test of that idea just failed. OpenAI quietly discontinued ChatGPT's Instant Checkout feature, the one that let people complete a purchase without leaving the chat window, after it launched with real retail partners in late 2025. That's a useful signal for any small business owner trying to figure out whether "selling through AI" is something to act on now or something to watch from a safe distance.
Here's what actually happened, what's still moving forward, and what to do about it.
What "agentic commerce" means in practice
Two competing approaches have been racing to let AI assistants shop on a customer's behalf. OpenAI built the Agentic Commerce Protocol with Stripe, which powered Instant Checkout: a buy button inside ChatGPT that let shoppers pay for a product without leaving the conversation. Separately, Shopify and Google announced the Universal Commerce Protocol, an open standard meant to let AI agents discover and transact with any participating merchant's catalog, rolling out through Google's AI Mode in Search and the Gemini app.
Both protocols are trying to solve the same problem from different directions: giving an AI assistant a standard way to read your product data and, eventually, complete a sale. Where they differ is what they've actually shipped and whether it's working.
The part that didn't work
Instant Checkout launched with Walmart and Target as partners and was framed as the first real step toward chat-based shopping. By March 2026, OpenAI had scaled it back, citing weak uptake, and shifted its stated focus from in-chat payment to product discovery. The new direction sends shoppers who find something in ChatGPT back out to the retailer's own site or app to actually pay. Shopify, notably, confirmed it would not build a dedicated checkout app for the feature, and other partners had no firm timeline to build one either.
The lesson isn't that AI shopping is dead. It's that the hardest, most trust-sensitive part of a purchase, actually handing over payment and shipping details, turned out to be the part customers weren't ready to do inside a chat window. People were comfortable asking an AI assistant for product recommendations; they weren't comfortable paying through it. That's a meaningful distinction if you've been wondering whether you need to urgently integrate your store with any of this.
What's still moving forward
While in-chat checkout stalled, the discovery side of agentic commerce kept building. The Universal Commerce Protocol launched with backing from major retailers and is explicitly aimed at the earlier stage of the funnel: letting an AI assistant read a merchant's live product catalog, pricing, and availability accurately enough to recommend it, compare it, and point a shopper toward it. That's a lower bar than handling payment, and it maps onto something your business can act on today regardless of which protocol eventually wins: whether your product information is accurate, current, and machine-readable in the first place.
What this means for your store, right now
You don't need to pick a side in the OpenAI-versus-Google protocol race, and you definitely don't need to build a custom checkout integration for a feature that was just scaled back. What's worth doing instead:
- Keep your product data clean and current. Accurate titles, descriptions, pricing, and stock levels matter more than ever, because every version of this technology, in-chat or not, depends on an AI being able to read correct information about what you sell and whether it's actually in stock.
- Use the feeds you already have access to. If you're on Shopify, Google Merchant Center, or a platform with a product feed, make sure it's connected and up to date. That's the same plumbing AI shopping surfaces are reading from, whether or not checkout ever happens inside the chat itself.
- Don't build custom integrations for features that aren't stable yet. Instant Checkout went from launch to scale-back in under six months. Chasing every new agentic commerce feature as it appears is a good way to spend time on something that gets pulled before it pays off. Let the big platforms settle on a standard first.
- Watch where the traffic actually comes from. If you start seeing referral traffic from ChatGPT, Gemini, or Google's AI Mode in your analytics, that's your real signal to invest further, not a press release about a new protocol.
The practical takeaway is almost boring: the businesses best positioned for AI-driven shopping, whenever it actually arrives at scale, are the ones with accurate, well-structured product information today. That was true before anyone coined the term "agentic commerce," and it will stay true no matter which checkout protocol eventually wins.