Last week OpenAI quietly dropped a bomb. A leak from The Information revealed they’re killing Instant Checkout - their first real agentic commerce product, which let users buy products directly inside ChatGPT.
We decided to wait a few days until the dust settled before writing an analysis, and let the “spicy” takes on LinkedIn - people dunking on OpenAI, declaring agentic commerce dead, doing victory laps - cool off a bit. We felt most takes were not really thinking this through from first principles (although there was one really great analysis by Jason Goldberg that we’ve used below). Now that the dunking is cooling off a bit, we can start with our analysis.
Background
Instant Checkout was announced back in September 2025 in partnership with Stripe and their co-developed Agentic Commerce Protocol (ACP). The idea was that users already come to ChatGPT to research products, so why not let them buy right there? It launched with Etsy sellers first, with Shopify’s merchant network supposed to follow.
There was one great take from Jason Goldberg, Chief Commerce Strategy Officer at Publicis. So we’ll start with his summary:
“Here’s what actually happened. OpenAI launched Instant Checkout on September 29. It supported single-item purchases from Etsy sellers only. No multi-item carts. No promo codes. No shipping promises. They hadn’t even set up sales tax remittance. Only about a dozen Shopify merchants ever went live….
…The real story isn’t that consumers don’t want agentic commerce. It’s that OpenAI is fighting for survival on multiple fronts and couldn’t afford to build a commerce stack on the side.”
Jason’s take centers on OpenAI shifting resources as they lose market share in the consumer AI landscape. He’s right that Instant Checkout was an extremely early beta. Almost no merchants were on board. It was highly basic, with zero real effort to push consumers to use it. And the plug was pulled before it ever had a fair shot.
We think this is a good explanation. Here are our additional takes:
The real focus is ads and ChatGPT Apps. Fidji Simo - who came from Instacart to become OpenAI’s CEO of Applications - is laser-focused on building ChatGPT into a consumer platform. The big monetization play right now is advertising (which officially launched in February 2026) and the ChatGPT app ecosystem (Instacart, Expedia, DoorDash, Target, etc.). That’s where the energy and resources are going.
Instant Checkout competes with ChatGPT Apps internally. If you're building an app ecosystem where Target and Instacart build dedicated ChatGPT experiences that handle their own checkout, having a separate native checkout flow that bypasses those apps creates internal tension. You can't push both at the same time.
We actually wrote about this back in October 2025, a month after Instant Checkout was announced:
ChatGPT’s Commerce Dilemma
“An interesting angle that would be interesting to follow is internal competition within ChatGPT. The platform itself will need to decide when to trigger an app versus the instant checkout interface. For example, if someone builds a sports app and a user asks for “running shoes,” does ChatGPT launch its own commerce flow or open the sports app instead?
This creates a new kind of internal platform tension - between ChatGPT’s standardized commerce and app-based commerce. Users might be prompted to choose, but given the stakes, it’s likely that ChatGPT will eventually lean in one direction.”
The ROI for OpenAI wasn't worth the effort. Building a real commerce stack - live pricing, inventory sync across millions of SKUs- is incredibly hard to do from scratch, while the potential outcome on OpenAI top line is probably two orders of magnitude lower than ads, and significantly lower than ChatGPT Apps. The juice wasn't worth the squeeze. Plus, OpenAI needs merchants and brands on their side because the ChatGPT Apps platform is the strategic bet - and competing with those same merchants over who owns the checkout was creating friction with some of the larger players they probably need as partners.
Meanwhile, Shopify Didn’t Get the Memo
To prove the point that agentic commerce is far from dead - just days after the Instant Checkout “death” was announced, Shopify sent an email to its merchants announcing “ChatGPT in Agentic Storefronts,” launching later in March. The email (attached below) tells merchants that buyers will be able to find their products and complete purchases inside ChatGPT - and that this channel will be turned on by default for their store. We believe that what Shopify refers to as “Agentic Storefronts” is actually built on the ChatGPT Apps framework, using ACP under the hood as OpenAI specifies.
In many senses, Agentic Storefronts deliver the same promise Instant Checkout had - merchant keeps merchant-of-record status, you can buy directly through ChatGPT - but it benefits the merchant or brand better. The experience is branded, lets the merchant build custom apps with custom control, and retain ownership of the checkout for crucial things like upsells, loyalty programs, and their own business logic.
We’d also remind our readers that Anthropic (Claude) has announced they’ll be accepting the MCP standard that powers ChatGPT Apps, plus entering agentic commerce (yours truly wrote about this last month) - so we expect them to add support as well at some point this year.
Google Is moving ahead
While OpenAI is stepping back from native checkout, Google is doing the exact opposite - doubling down. Back in January at NRF 2026, Sundar Pichai personally took the stage to announce the Universal Commerce Protocol (UCP), a new open standard for agentic commerce co-developed with Shopify. UCP checkout module is also richer in its capabilities, supporting merchant embedded checkout from day one. The difference between Google's approach and what OpenAI tried is structural. Google has the Shopping Graph, Merchant Center with millions of retailers already feeding product data, Google Pay for frictionless payments, and years of ad infrastructure. They're not building a commerce stack from scratch - they already have one.
Closing Thoughts
Technology adoption is never linear. It’s not a smooth line - especially when it’s not just technology but the intersection of different business incentives and basically a redistribution of power around e-commerce.
We expect there will be many more back-and-forth moments between merchants, the agents they’re building, and AI surfaces like Claude, Copilot, ChatGPT, and Gemini. It’s going to be a bumpy road.
But the potential is there. We feel very strongly that there will be commerce embedded in chat interfaces, and it will grow in complexity around the use cases we’ll be able to support. This is the starting point, and it’s a rough ride - but don’t let clickbait social media posts take your eyes off the real target.
If you want to read more agentic commerce deep dives, please feel free to subscribe below:
Nekuda are literally the number 1 voice in agentic commerce, and have been for some time now.
The single-item, no-cart limitation you described tells the whole story. That's not a commerce product - that's a demo. I was testing agent shopping this week with a different setup: give an AI a budget and a vague brief, let it find and buy something.
The actual blocker wasn't payment auth. It was that every serious e-commerce site has bot detection sophisticated enough to stop a headless browser cold before checkout even loads. OpenAI's pivot away from Instant Checkout may be resource constraints, but the underlying infrastructure problem would've bitten them anyway.