For decades, the rhythm of online shopping was predictable, if often frustrating: search, scroll, cart, checkout. Then social platforms like Facebook and Instagram shook things up by dropping products directly into your feed for the ultimate impulse buy. Now OpenAI and Stripe are upping the ante, folding shopping into the flow of an AI chat — a kind of holy grail for consumers who want to go from idea to purchase in seconds.
On Wednesday, the companies unveiled an Instant Checkout feature in ChatGPT, powered by a new commerce protocol they co-developed. The feature is launching first with U.S.-based Etsy sellers and will soon extend to more than a million Shopify merchants, including buzzy brands like Glossier, Skims, Spanx, and Vuori. The protocol sits on top of an open standard for connecting AI models to business systems, developed by Anthropic, called MCP — but focuses specifically on commerce and payments. Stripe brings fraud prevention, global payment rails, and a vast merchant network, making the new Agentic Commerce Protocol (ACP) usable by millions of businesses right out of the gate.
But the implications go far beyond OpenAI. Because the protocol is open source, any AI assistant—from Claude and Gemini to TikTok’s AI and xAI—could use the same rails to let users shop within their chatbots. That means conversational shopping could spread across the internet far faster than brands are ready for.
The result: Google and Amazon’s days as the default starting points for online shopping may be numbered. Sure, those companies could adopt the same open protocol and join the in-chat shopping game too. But the bigger point is that, for the first time in two decades, there’s a real chance that the foundations of online shopping could shift. And that shift, if it catches on, will have profound implications for everything from the way that consumers make purchasing decisions to the way that brands market their products.
Still, why would OpenAI create a protocol that competitors can use too? ChatGPT product lead Michelle Fradin framed it as a merchant-first calculation: “The primary goal we had was making something incredibly easy for the entire ecosystem—merchants and developers—to adopt,” she said. “So yes, one aspect of that is, competitors or other players in the space can adopt it too.”
At the end of the day, she explained, it’s “net-beneficial” if merchants have to do less work to integrate with different platforms and grow sales. “We felt like the most merchant-friendly approach was making this available to everyone,” she added, noting there was so much demand from merchants that “we needed to build something that could scale.”
To be sure, not every brand or retailer will jump in. Even though ACP is open source, adopting it still means ceding some control of the direct relationship they’ve worked to build with their customers. There’s also the matter of fees: Stripe will take its cut on each transaction, which could be a deterrent for sellers operating on thin margins. But for many companies, especially bigger brands, the frictionless reach of AI shopping may outweigh the loss of control over customer data and brand experience.
And while brands and merchants will be able to take advantage of the ACP protocol through other platforms that adopt it (if Anthropic or Google were to adopt the protocol for example), there’s an obvious benefit to having products appear directly within ChatGPT’s search results, given the chatbot’s popularity. In addition to Etsy and Shopify, Fradin said that OpenAI is currently working with several other large retailers that will eventually be included in ChatGPT’s Instant Checkout, though she declined to name them.
To be included in ChatGPT’s search results for Instant Checkout, brands and merchants need to do some work, Fradin noted. That’s because if you want ChatGPT to recommend your product, you have to feed it structured, detailed data that the model can “see” and understand.
“The best way today to ensure as a merchant that your product has the best chance of being chosen is to make sure that we have the most up to date and richest amount of information about your products as possible,” she said. Fradin added that merchants are eager to provide detailed “product feeds” — essentially structured catalogs of their items with rich descriptions, updated prices, and availability — so ChatGPT has the fullest context when deciding what to recommend. In some cases, she noted, brands are supplying even more detail than they publish on their own sites, in hopes of improving their chances of being surfaced. It’s a sign that a new discipline is taking shape: what some are calling AIO, or AI Optimization, the successor to SEO, where the goal is to fine-tune product data so AI assistants surface your brand instead of a competitor’s.
Beyond the technical plumbing of payments, ACP effectively positions ChatGPT as a new arbiter of product recommendations. Instead of shoppers browsing Google search results, scrolling Amazon’s “customers also bought,” or consulting reviews on Wirecutter, the assistant itself will increasingly decide what to surface. That shift raises thorny questions: How will ChatGPT determine which product to recommend? Will it offer a menu of options or streamline to a single choice? And down the line, will OpenAI accept money from vendors to boost their placement — turning conversational commerce into a pay-to-play channel?
For now, Fradin noted that, unlike traditional search engines, where platforms carefully guard algorithms to prevent low-quality content from gaming the system, ChatGPT’s shopping recommendations are entirely AI-driven. There isn’t a fixed formula merchants can optimize against, she explained — instead, the more high-quality product information the model has access to, the more likely it is to surface relevant results.
“We expect to see a lot of evolution in the new version of AIO over time,” she said. “I think we’re just at the beginning of this space.”