Shopify Agentic Storefronts,
explained.

For most of the last year, the agentic commerce conversation has been theater. A lot of demos, a lot of protocol announcements, not much that actually moved a merchant's revenue. Late March 2026 changed the math. Shopify shipped Agentic Storefronts, and the change is not subtle: every Shopify store is now discoverable by default across AI assistants. ChatGPT, Microsoft Copilot, Perplexity, and others can surface your products without you lifting a finger.

I have watched a lot of platform shifts from inside the building and from the merchant side. This one is closer to the launch of the App Store than to most of the AI noise. It moves where the customer first meets you. And it does it in a way that, for once, keeps the merchant in the driver's seat.

Let me walk through what actually launched, how the mechanics work, and the part most people skipped past: who owns the checkout and the customer.

What actually
launched in
late March.

Agentic Storefronts is a default-on capability. You did not have to install an app or flip a setting for your products to become eligible. Shopify exposed the catalog of its stores to AI assistants through a universal product feed, and overnight the surface area for discovery expanded from search engines and social to the assistants people now ask first.

The framing matters. Shopify did not build a marketplace that sits between you and your buyer. It made stores legible to the agents that buyers already use. A shopper asks ChatGPT for a merino base layer under a hundred dollars, the agent can now consider your product as a candidate, and the shopper can move toward buying it.

Compare that to how merchants normally get onto a new surface. Usually there is an application, an approval queue, a fee negotiation, an integration sprint. Here the surface arrived to the merchant. That is rare, and it is the reason this deserves more of your attention than the average platform headline. The brands that understand it as a free expansion of their top of funnel will treat it like one. The brands that ignore it will simply lose share in the answers their buyers are already reading.

Discovery in the
agent, checkout
on your store.

Here is the part that separates Agentic Storefronts from the OpenAI approach that got the early headlines. The discovery happens inside the assistant. The purchase completes on your own storefront, through Shopify Checkout. The agent introduces you. It does not become the store.

Practically: the shopper finds you in the agent, then lands on your checkout, pays through your flow, and lives in your data afterward. You are the merchant of record. The customer is your customer. If you have built a brand worth returning to, you keep the relationship that makes that brand worth anything.

Under the hood, all of this rides on Shopify's Catalog API, the universal product feed and identifier system that lets agents understand what you sell. If your data is thin or inconsistent, agents have less to work with and you show up less often. The feed is the front door, and a sloppy front door costs you traffic. I dig into the practical work in the Catalog API playbook.

There is a subtle benefit to this design that gets overlooked. Because the buyer lands back on your store, your existing conversion work still does its job. Your reviews, your bundles, your upsells, your post-purchase flows, all of it stays in play. An in-agent checkout would have flattened your storefront into a single line item in a chat. Agentic Storefronts hands you a warm, high-intent visitor and lets you do what you already do well. That is the difference between a channel that respects your merchandising and one that erases it.

Why merchant of
record is the
whole ballgame.

I co-founded a brand group that scaled into the hundreds of millions in revenue. The single most valuable asset across all of those brands was not any one product. It was the customer relationship: the email, the purchase history, the permission to talk to that person again. The moment a middleman owns checkout, that asset starts leaking to them.

Agentic Storefronts keeps you as merchant of record. That sounds like a compliance detail. It is not. It means you keep the customer data, the order relationship, the ability to run retention, and the freedom to merchandise the post-purchase experience. You are renting attention from the agent, not renting your customer back from a platform.

The reported fee,
and how to read
it honestly.

There has been a roughly 4 percent fee reported on checkout sales that originate from ChatGPT. Treat that as reported, not official. The number may shift, the terms may differ by agent, and Shopify has not framed it as a fixed universal rate. So do not build a spreadsheet around a figure that is still moving.

The right way to think about it: agent-originated sales are a new acquisition channel, and like any channel it has a cost. If a 4 percent take brings you an incremental customer you would not have reached through search or social, the unit economics can still work, especially because you keep the customer for the next order. Compare it to your blended customer acquisition cost, not to a zero baseline.

What merchants
and app builders
should do now.

For merchants: clean your catalog. Structured titles, accurate attributes, real identifiers, complete product data. The feed is what agents read, and an incomplete feed reads as a weaker candidate. Then watch your analytics for agent-referred traffic so you can tell whether this channel is actually converting, not just appearing.

For app builders: the surface just expanded. Tools that improve catalog quality, enrich product data, monitor agent visibility, or help merchants understand agent-referred conversion all have a clearer reason to exist than they did a month ago. I spent years building the partner program at Shopify. When the platform opens a new surface, the apps that solve the boring data problem usually outlast the ones chasing the flashy one.

One more thing for both audiences. Do not wait for perfect understanding before you act. The merchants who waited for the search engine optimization playbook to be fully written in the early days lost a decade of compounding to the ones who started rough and improved. Agent discovery is at that same early stage now. The cost of starting is a weekend of catalog cleanup. The cost of waiting is showing up in fewer answers every month while a competitor with cleaner data takes the slot.

The companies that win this will not be the ones with the loudest AI announcement. They will be the ones whose data was already clean when the door opened.

If you want a clear-eyed read on whether your products are surfacing, start with why some products win in ChatGPT and most stay invisible, see what the end of ChatGPT Instant Checkout teaches about betting on one agent, then do the real work in the Catalog API playbook. Or skip the reading and tell me about your store.