Guides

AI shopping agents: what changed after ChatGPT’s Instant Checkout, and what store owners should do now

By Nick Acquah, founder of Mercenta · 6 minute read

Published 29 Oct 2026 · Updated 29 Oct 2026

What do AI shopping agents mean for my store?

The buy button flopped. AI shopping didn’t.

When I first heard you’d be able to buy inside ChatGPT, I thought: if they pull this off, they’ll have the most powerful closed system there is. AI is already changing how people research, and people are adopting it fast. Close the sale inside the app too, and a lot of other marketplaces were going to get hurt.

When it got pulled back, it made things a lot clearer. ChatGPT is being used as a next-generation Google.

Here’s what I mean. On Google, you type a phrase like “best headphones” and land on someone’s comparison site. On ChatGPT, you ask long, specific questions much earlier, while you’re still working out what you want.

So the failed checkout just proved how much people use it for research. You speak into your phone, have a conversation and reach a conclusion. Compare that to typing into Google and reading an article that may or may not answer your question. It’s night and day.

It failed at checkout. But that showed us how well it works as a research tool.

ChatGPT launched Instant Checkout in September 2025. In March 2026, OpenAI said merchants could use their own checkout, and that it would focus on discovery. Its staff had found people were researching products in ChatGPT, but not buying there. (Source: Modern Retail)

That doesn’t mean agents don’t buy. In September 2026, Shopify said it would let Meta’s Muse shopping agent check out on all Shopify stores with Shop Pay. (Source: Implicator) And Google is rolling out one shopping cart across Search and Gemini in the US, where you pay in a few taps. (Source: PPC Land) The buying is moving. The research is where the decision gets made.

Why people research in AI but buy on your store

People want research to be easy, and they trust what comes back because it’s presented so intelligently. It doesn’t read like something written for someone else. It reads like it was made for you, in words you understand. If you don’t understand something, you ask the same chat to explain it. As a research tool, it can’t be beaten.

Buying is different. This is just my hypothesis, I haven’t got the data to back it, but I think it depends on what you’re buying. The more expensive the item, the more you want to touch and feel it.

Price matters too. ChatGPT might not show you the best price. Five or six retailers might sell the same product, with different delivery times, colors or versions you only see on their own sites.

Personally, I’d buy inside the tool. But the market wasn’t ready. Will they come back to it? I’m not sure. They might. Either way, the real strength is the research. That can’t be replaced.

Walmart saw the same thing. It said conversion was about three times lower when people bought inside the chatbot than when they clicked through to its own site. (Source: Modern Retail)

What an AI shopping agent actually is

An AI shopping agent is software that searches, compares and shortlists products for a shopper. At an immense scale.

You give your agent a goal: find me this. It isn’t limited to certain places. It looks across lots of sites, then comes back with a shortlist. It can buy for you, if you let it. Or it just brings you the list and you choose.

And it’s intelligent. It doesn’t move in a straight line. It thinks as it goes, solves problems along the way, and works out the best route to the goal the shopper set. If one search comes back thin, it tries another. If a page doesn’t answer something, it goes and finds the answer somewhere else.

Think of it as a living online organism that reports back to its boss: you. Either it hands you the shortlist, or you let it buy.

What does that mean for your store? Keywords alone won’t carry you. Every gap on your product page is a question the agent has to answer somewhere else. And where it finds that answer might be your competitor’s page. So is the sale. Make its job easy, with clear facts and straight answers, and you stay on the list.

What the agent needs to know about your product

The agent needs to know everything about your product.

Take headphones. A buyer might be weighing up:

• Over-ear, or earbuds?
• Working out and running? Then they need to sit comfortably in the ear.
• Studying? Then they want noise cancellation.
• Swimming? Then they need a waterproof rating, and proof of it.
• Lots of calls? Then they need a great microphone.
• Swappable tips for different ear sizes, or fixed?
• How comfortable are they, according to other people?
• How long does the battery last? Does the case charge wirelessly?
• Do they work well with their phone?
• How’s the bass? Is it strong?

The agent uses all of this at different stages of the research. The first question might be general. The last one might be very specific. Miss any of those stages and your product can drop out.

So it needs the granular details. Why someone would use your product. And, more importantly, the things you take for granted and never think to mention. Those need to be there too.

The one thing to fix this month

First, make sure AI can see you. Make sure your site is crawlable. Then make sure what AI finds about your brand and products is true.

Your articles, your specs, whatever you have online: keep it up to date and correct. Fill out your schema. When AI crawls your site, it should get your products exactly right. If it misquotes or misses things, that hurts your reputation with the LLMs further down the line. And a product AI gets wrong is a product it doesn’t sell.

How to check the basics: see GEO vs AEO vs SEO: what actually changes for online stores.

What to ignore for now

Rushing to build an agent checkout program. I wouldn’t bother building a bot on your site to help people buy, or diving into chatbots and agents to make your site more interactive. I’d ignore that completely.

The one exception is Shopify. If you’re on Shopify, showing your products in AI channels is managed from your Shopify admin. That’s a setting, not a build. (Source: Shopify)

Who should care right now?

Everybody should worry about this now. The sooner you fix these things, the easier it is to rank.

Despite what you might read online, most stores aren’t optimized. They haven’t really thought about it. They’re living in yesterday. And every time a shopper asks AI and hears a competitor’s name, that’s a sale that went somewhere else. Make changes today and it’s a lot easier to get to the top of the pile.

And if you’ve got decent SEO but no AEO, and you’ve always lost out to certain brands, you might be able to outcompete them right now. A lot of terms and angles are up for grabs. Exploit that now and you could take a lot of market share, and a lot of sales, from your rivals, just by moving quicker.

No one can really wait. Waiting just prolongs the inevitable. The market is open now. Take advantage.

What surprised me in our own tests

We crunch a lot of data, across a lot of industries and different LLMs. Each LLM cites different sources and weighs things differently. There are patterns, but they’re hard to spot. Once you spot them, you can pretty consistently cover the bases of what each LLM is looking for.

But they change how they look at sources over time. What worked yesterday might not work today. That’s why you’ve got to stay on it.

The biggest surprise: if you think a couple of changes will do it, you’re sorely mistaken. Put the time and effort in, with the right strategy, and you’ll be in a great position. But getting that strategy right means crunching a lot of data. Not just for now. For tomorrow too.

FAQ