OpenAI has introduced buying directly within the chat. Instant Checkout turns ChatGPT into a sales channel — Etsy is already live, and over a million Shopify merchants are in onboarding. Behind it all is the Agentic Commerce Protocol, an open standard developed by OpenAI together with Stripe. For retailers, this raises a new question: how do you sell to a customer who never visits your website? Here's what holds true today, what remains speculation, and how to prepare.
The purchase is moving to where the question is asked
Until now, a simple order prevailed. Customers research somewhere, but buying happens in the store. The website was the destination of every marketing effort, and whoever had the traffic had the business. That order is starting to crack. OpenAI has introduced Instant Checkout in ChatGPT. Users can buy products directly in the chat without ever visiting the retailer's website. Etsy is already live, and over a million Shopify merchants are in onboarding, including brands like Glossier, SKIMS, and Vuori.
This development deserves neither hype nor dismissal. Overstating it would mean claiming the traditional store is dead. Understating it would mean calling it just another channel. The honest assessment is this: a purchasing path is emerging where the advice, the selection, and the transaction all happen in one place the retailer doesn't control. Millions of people already ask ChatGPT which product to buy. Now they can buy it there too.
What the Agentic Commerce Protocol is and why it's open
Technically, Instant Checkout is built on the Agentic Commerce Protocol, an open standard developed by OpenAI together with Stripe and published as open source. The protocol defines how an AI agent shops with a retailer on behalf of a person: how products are handed off, how payment works, and how the retailer receives the order.
Two features of this protocol are worth noting. First, the retailer remains the retailer. The order runs through the retailer's own systems — payment, fulfillment, and the customer relationship all stay under their control. ChatGPT is where the purchase happens, but it is not the seller. Second, OpenAI has stated that participation in the payment system does not influence the ranking of product recommendations. Whether that remains the case long term is something to watch. But as a starting point, this model differs markedly from marketplaces where visibility is bought.
The fact that the standard is open carries strategic significance that extends beyond OpenAI. An open protocol can be implemented by other assistants and platforms as well. It is an attempt to set the rules for AI-assisted shopping before every provider builds its own closed system. For retailers, that is good news — a standard means: connect once, use many times.
Why Shopify merchants have a head start
For an individual retailer, connecting to a channel like this would be a substantial project. Product data must be available in machine-readable form, order intake must be automated, and payment must be agent-ready. This is exactly where the value of a platform becomes clear. Shopify is building the infrastructure for agentic commerce at the platform level, and the connection to Instant Checkout comes to merchants as a feature, not a custom development effort. Bringing over a million merchants into a new channel simultaneously is precisely the kind of task a shared platform exists to handle.
That is also the straightforward reason why this topic matters on our Shopify page. If you're on the platform, the agentic era arrives as an update. If you run a proprietary solution, it arrives as a project. Both are manageable, but the difference in cost and speed is real — and it grows with every additional channel that emerges.
What changes for your content
The most consequential implication of agentic commerce isn't technical — it's about content. An AI agent shops differently from a person scrolling through a store. It doesn't read banners. It isn't impressed by a beautiful header. It works with what is structured and available: product data, attributes, prices, availability, reviews, shipping terms.
That shifts the priorities. A product with incomplete attributes is essentially invisible to an agent, no matter how well it's photographed. A size chart that only exists as an image cannot be evaluated. Contradictory information between a title and a description makes a product useless for recommendation. In short: product data quality moves from a back-office concern to a sales question. Retailers who have it in order will be found and recommended by agents. Those who don't simply won't exist in this channel. Why clean product data determines the success of digital projects in any case is something we covered in our Pimcore article on data strategy. Agentic commerce turns that argument into a deadline.
What holds true today and what is still speculation
With a topic this new, being precise is part of being honest. What is established is the current state: Instant Checkout currently supports single-item purchases; multi-item carts and additional regions are planned to follow. The rollout starts with US users and US merchants. For a retailer outside the US, this means the channel isn't part of everyday business yet — but the direction is set, and the lead time is a gift.
Speculation covers everything related to revenue shares. No one knows today what share of commerce will run through assistants in five years. Anyone who gives you a percentage is guessing. The sensible approach is that of a pilot: stay prepared, test early, learn from real data. Panic is as unwarranted as ignoring the signal.
Three things you can do right now
The first is an audit of your product data. Take your twenty most important products and review them through the eyes of a machine. Are all attributes filled in? Do the title, description, and technical specifications align? Are sizes, materials, and compatibility details available as structured data, or only as running text and images? This review takes an afternoon and gives you an honest picture of how agent-ready your catalog is.
The second is the platform question. If your store runs on Shopify, keep an eye on the Instant Checkout onboarding and sign up early once it opens for your region. If your store runs elsewhere, agent-readiness belongs in every platform discussion from now on. It's a criterion that didn't exist two years ago, and it's not going away. If a platform switch is already on the table, read our analysis of why enterprise brands are moving to Shopify.
The third is a look at your overall channel strategy. Agentic commerce is one more channel alongside your store, marketplaces, and social commerce. The businesses that will do well with it are those that serve all their channels from a single product source and a single system, rather than building a silo for each channel. That work pays off regardless of how quickly the agents arrive.
What a purchase through ChatGPT actually looks like
To keep this concrete, it's worth walking through the experience from the customer's perspective. A user asks ChatGPT for a gift for a friend who enjoys hiking. The assistant suggests specific products, including one from a Shopify merchant. She asks follow-up questions about the material and delivery time and gets answers drawn from the merchant's product data. Then she taps Buy. Payment runs through the payment method saved in the chat, and the order lands in the merchant's system like any other — with shipping, tracking, and returns handled by the merchant.
From the merchant's side, the same transaction looks entirely unremarkable — and that is exactly the point. It's one more order in the system, with a new source channel. No new back end, no separate inventory, no separate payment processing. The complexity lives in the protocol underneath, and that is the platform's responsibility. What the merchant needs to contribute is what the agent needs to make a recommendation: complete, consistent, machine-readable product data.
The questions retailers ask us
Three questions come up repeatedly in conversations. The first: who owns the customer? Under the current model, the retailer does. The order runs through their systems, they know the buyer, they handle shipping and service. Whether assistant platforms hold that line over time is something to watch carefully.
The second question: how do I get recommended? The honest answer is that there's no guaranteed formula. OpenAI states that payment participation does not influence rankings. What you can influence are your own fundamentals: data quality, availability, competitive terms, and reviews. That sounds like the same old homework — because it is.
The third question: will this channel cannibalize my store? In the short term, hardly — volumes are too small for that. In the long term, the more important question is: if some purchases shift to assistants, do you want to be there or not? The channel doesn't cannibalize your store. It cannibalizes the retailers who aren't in it.
A final perspective
Commerce history has a recurring pattern. A new channel emerges, some dismiss it as a novelty, others call it the end of everything, and in the end it simply becomes part of doing business. That's how it went with the online store itself, with mobile, with marketplaces. There's every reason to think agentic commerce will follow the same path — only faster, because this time the infrastructure is being built as an open standard from the start.
The good news is that what you need to do doesn't require betting on the future. Clean product data, an agent-ready platform, and a unified system instead of isolated solutions pay off even if the agents arrive more slowly than expected. These are the fundamentals that always make sense. How we can help is laid out on our Shopify page. You'll find all our analyses in our Shopify Insights.