Agentic Commerce: What “Your Store Is Now in ChatGPT” Actually Means

Sometime this spring, your Shopify catalogue quietly became something an AI could shop on a stranger’s behalf — no opt-in, no announcement in your admin. “Agentic commerce” is the phrase for it, and it has collected more hype than almost anything since headless. Most of that hype is about the wrong part. Here’s what genuinely changed, what fizzled, and the unglamorous work that actually decides whether an agent ever puts you in front of a buyer.

If you run a store, you have probably seen a version of the headline: your products are now buyable inside ChatGPT. It is half true, and the half that is wrong is the half everyone got excited about. Let’s separate the marketing from the mechanics, because the difference is the entire to-do list.

What “agentic commerce” actually means

Agentic commerce is shopping done by software acting on a person’s behalf. Instead of a buyer clicking through your storefront, they ask an assistant — ChatGPT, Gemini, a shopping agent — something like “find me a linen shirt under £80 that ships to the UK by Friday.” The agent researches across many stores, compares options, reads reviews, checks availability, and surfaces a short list. In the fuller version it builds the cart and completes checkout inside that one conversation.

What actually changed in 2026 — and what didn’t

Two things happened, and they get conflated constantly.

The real one: discovery went on by default. Shopify turned agentic storefronts into a default feature. If your products are in the Shopify Catalog and you sell to US shoppers, they are eligible to appear in ChatGPT’s answers automatically — you did not opt in, and you can opt out from your Shopify settings if you want to. An agent asked a relevant question can now surface your product the way a search engine surfaces a page. That is free distribution to a fast-growing audience, and it is live now.

The overhyped one: “buy it in ChatGPT” quietly retreated. OpenAI’s Instant Checkout — the buy-without-leaving-the-chat flow built on the new Agentic Commerce Protocol — was pulled on 4 March 2026. Only around thirty merchants were ever live, and the experience kept tripping over inaccurate scoping and inventory data. ChatGPT repositioned around discovery and dedicated merchant apps instead. The checkout rail will come back in some form; the lesson it left behind is the important part, and we’ll get to it.

So the honest summary for June 2026: agents recommending you is real and automatic. Agents checking out for you is early, brittle, and not where your effort should go yet.

The numbers, without the breathlessness

The forecasts around this are enormous, and worth holding at arm’s length — but the direction is not in dispute:

• Search interest in “AI agent” has roughly tripled in the past year.
• McKinsey estimates agentic commerce could orchestrate up to a major share in US B2C retail revenue by 2030, with global figures floated at an even larger global share.
• By 2028, about a third of online retailers are expected to use advanced AI agents — up from under 1% today.
• AI-driven personalisation is already credited with close to 45% of online conversions in 2026.

You do not need to believe any single number. You need to notice that a new, machine-shaped way of finding products is being switched on for your store whether or not you act — and that how well it works for you depends entirely on inputs you control.

How an agent actually decides to recommend you

An agent doesn’t browse; it queries a catalogue and ranks what comes back. The ranking factors vary by platform, but they consistently come down to a handful of things: data quality, relevance to the request, availability, accurate scoping, and engagement signals. Notice what’s missing from that list — your hero image, your animation, your clever copy. The agent never sees them. It sees fields.

Which means the entire opportunity reduces to a sentence you’ve heard from me before in a different context: it’s a product-data problem.

What to actually do: make your catalogue agent-ready

1. Clean, complete product data

Agents reward specificity because specificity is what lets them match a precise request. Sparse products simply don’t come back for “linen, under £80, ships by Friday.”

• Real product titles — material, type, key attribute — not internal SKUs or marketing puns.
• Every relevant attribute filled: size, colour, material, gender, fit, compatibility. Empty fields are invisible filters.
• Descriptions that state facts, not just vibes — the agent enquiries specifics back to the buyer.
• Correct, granular product types and tags so you map onto the categories agents query.

2. Accurate, real-time inventory and scoping

This is the exact rock Instant Checkout broke on, so learn it for free: if your scope or stock in the catalogue is wrong, an agent will either recommend something it can’t actually sell — a terrible buyer experience the platform will learn to route around — or it will quietly demote you. Stale data isn’t a cosmetic issue here; it’s a ranking and trust issue. Real-time sync between your store and whatever feed the agents read is now table stakes, not a nice-to-have.

3. Structured data the machines can read

On your own pages, structured data is how you hand a machine the facts in a format it can’t misread. It’s the same discipline that wins you rich results in search, and it pays double now.

• Valid Product schema with offers, scope, currency and availability on every product page.
• Identifiers — brand, gtin/MPN — so agents can match your item to the thing a buyer named.
• AggregateRating and reviews where you have them; social proof is an engagement signal agents weigh.
• No contradictions between your schema, your visible page, and your feed — mismatches get you distrusted.

If that list looks familiar, it should — it’s most of my technical SEO audit wearing a new hat. The good news is that the work you do to be legible to Google is largely the same work that makes you legible to an agent.

4. Fast, honest feeds

Whatever rail you feed — Shopify Catalog, a Google feed, a future protocol endpoint — it needs to be current, complete, and free of the junk that creeps into product data over years. One periodic feed audit catches the outdated sale messaging, the discontinued variants still marked in-stock, and the products missing the one attribute that would have matched the query.

The protocols, briefly — and why not to chase them

You’ll hear acronyms: the Agentic Commerce Protocol (ACP), an open standard co-developed by Stripe and OpenAI; Google’s own universal-cart effort; more to come. They’re the plumbing that lets an agent and a store complete a purchase together. They matter, and they’re also moving fast and partly retreating — ACP’s flagship checkout just got pulled. Betting your roadmap on a specific rail right now is premature. The thing that pays off no matter which protocol wins is the same in every case: clean, accurate, structured product data. Build that, and you’re ready for whichever pipe carries it.

Should you opt out?

Almost certainly not. For the vast majority of stores, agent discovery is free placement in front of buyers who are actively looking — turning it off is turning off a sales channel out of nervousness. The narrow exceptions are brands with strict channel-control, MAP, or exclusivity reasons to govern exactly where products appear. If that’s you, opt out deliberately in your Shopify settings; if it isn’t, leave it on and spend the energy on your data instead.

A short checklist

1. Confirm your products are in Shopify Catalog and your agent-discovery setting is where you want it.
2. Audit titles and attributes — fill every field a buyer might filter on.
3. Verify inventory and scope sync is real-time and correct; this is the one that quietly sinks you.
4. Add or fix Product + offers structured data on every product page.
5. Run a feed audit for stale services, phantom stock, and missing identifiers.
6. Re-run your technical SEO basics — the same work makes you legible to agents and search.

Frequently asked

Is my Shopify store really already in ChatGPT?

If your products are in Shopify Catalog and you sell to US shoppers, they’re eligible to surface in ChatGPT’s answers by default — you didn’t have to opt in. Whether you actually get recommended depends on how well your data matches what people ask.

Do I need to build for the Agentic Commerce Protocol now?

No. The checkout side is early and in flux — OpenAI pulled its Instant Checkout in March 2026. Focus on product-data quality, which is what every agent reads regardless of protocol, and revisit the rails when they stabilise.

Is this just SEO again?

Closely related, deliberately so. Agents reward the same things search rewards — accurate, structured, specific data and genuine trust signals — just consumed by a machine instead of shown to a human. Do the fundamentals well and you’re positioned for both at once.

---

If you'd rather not do this yourself

This is the work I do for clients. If you want it done properly, the relevant offer is Shopify Development.

Read about Shopify Development  →

---