AI-commerce: your product catalog becomes an interface for AI

SEO, Google Shopping, marketplaces, social commerce. As if your product catalog didn’t already have enough channels to keep track of.

And AI? You can now consider that a shopping channel, too.

People have been using ChatGPT and other AI systems for a while now to search for, compare, and research products. You don’t ask which online stores sell good hiking boots; you simply say:

“I’m looking for a waterproof hiking boot under 150 euros, suitable for wide feet.”

AI helps with the search, compares features, and helps you determine which option best fits your needs. That in itself isn’t new.

What is interesting, however, is what happens behind the scenes. AI systems are getting better and better at working directly with product catalogs, prices, availability, and other e-commerce data.

And right now, the traditional search landscape is changing as well.

TL;DR

• AI-powered commerce is already here. People are already using ChatGPT and other AI systems to search for, compare, and research products and services.
• Meanwhile, Google is also changing. Organic product carousels have disappeared from standard search results in Europe, while AI is playing an increasingly larger role in product discovery. Read about what’s changing with Google Shopping.
• ACP and UCP make commerce data usable for AI platforms. Think of product information, prices, availability, and ultimately even checkout or reservations.
• MCP makes it even more interesting. It gives AI access to your own data and tools, for example, to retrieve product information, check inventory, or verify the availability of accommodations.
• We’re working on this ourselves. With our Commerce MCP, we make up-to-date product information immediately usable for assistants and agents within Sterc.ONE and beyond.
• The real takeaway: don’t build for just one AI platform. Make sure your product catalog, PIM, CMS, or reservation system is structured, up-to-date, and accessible to AI.

Meanwhile, Google Shopping is changing, too

Since September 2026, Google Shopping has looked quite different in Europe.

Around September 16, the free product carousels, such as the well-known organic “Popular products” blocks, virtually disappeared from standard Google search results within the European Economic Area.

It's not because Google has suddenly stopped featuring products.

The change is related to the European Digital Markets Act, the DMA. That law prohibits large platforms from systematically favoring their own services over competing services. Google was fined by the European Commission for this in July 2026, and part of that fine specifically related to favoring its own services, such as Shopping, Hotels, and other specialized search results.

The result?

Price comparison sites and other so-called Comparison Shopping Services are now given more prominent space in European search results.

An important caveat: free product listings haven’t disappeared entirely from Google. They can still appear on other Google platforms, such as the Shopping tab, Images, Lens, YouTube, and Gemini.

But in the regular search results, something has indeed changed fundamentally.

And that makes this development particularly interesting.

You can no longer assume that product discovery takes place primarily on Google.

Consumers are searching differently. Platforms are changing. And AI is becoming an increasingly prominent part of that.

From searching to asking questions

The traditional online store starts with filters.

Brand. Price. Color. Size. Category.

An AI interface starts with intent.

• "Which office chair is suitable for someone who is 1.95 meters tall and works from home eight hours a day?"
• “I’m looking for a laptop for video editing that weighs no more than 1.5 kilograms.”
• “Which running shoe is suitable for wide feet and distances over 20 kilometers?”

That’s a fundamentally different way of searching. You’re no longer just searching by features, you’re describing your situation.

The AI then has to understand which products fit that description.

And that’s when product data suddenly becomes quite important.

ChatGPT already does this

ChatGPT now has a dedicated Shopping Research feature that lets users explore and compare products.

To do this, the system can use various sources, such as product information provided directly by merchants, public information on online stores, and other relevant retail sources.

This lets ChatGPT compare prices, features, and reviews, then point users to a seller.

In theory.

But if you use it in the Netherlands, you'll notice it's still far from perfect.

Sometimes you get nice product cards. Sometimes it’s mostly text. Sometimes you can easily click through to an online store.

And sometimes it takes a bit of searching to figure out where you can even buy that item.

There’s a reason for that.

Why doesn’t shopping in ChatGPT always work smoothly in the Netherlands yet?

Shopping Research itself is widely available, but OpenAI’s direct commerce infrastructure hasn’t been rolled out equally everywhere yet.

OpenAI runs a merchant program that lets online stores submit their catalogs directly, but as of this writing that direct integration is still rolling out from the United States.

As a result, ChatGPT often relies on information it finds on the web or gets from other sources when it comes to Dutch online stores.

That can work just fine, but it has limitations.

An online store may block automated access. Product information may be poorly structured. A price may have changed in the meantime. Inventory levels may be out of date. Or it may simply be unclear which variant corresponds to which price.

And that brings us to an important development:

online stores are gaining more and more options to make that information available to AI systems directly and in a structured format.

Enter: the Agentic Commerce Protocol

OpenAI uses, among other things, the Agentic Commerce Protocol (ACP for short) for this purpose.

Through ACP, merchants can make product information available in a structured format for AI-powered commerce.

Not:

“Here’s our online store, good luck crawling it.”

But rather:

"This is product X. This is the variant. This is the price. This is the current availability. These are the features. This is the image. And you can buy it here."

That gives an AI system much more to work with, and it gives merchants more control over what information is used.

That’s why OpenAI requests details such as product ID, title, description, URL, brand, images, price, and availability in its product specification. On top of that, much richer information can be added.

And that’s where the bigger trend lies:

Product data is no longer just data for your online store, Google Merchant Center, or a marketplace. It’s becoming data for AI.

ChatGPT isn’t the only one doing this

And that’s where it gets interesting.

OpenAI is using ACP to build commerce capabilities within ChatGPT, but you can see the same trend emerging in other well-known AI platforms as well.

Google, for example, has the Universal Commerce Protocol (UCP). With this, Google aims to enable commerce actions within AI Mode and Gemini.

This goes beyond simply displaying a product. UCP includes support for features such as shopping carts, checkout, fulfillment, and orders.

Google is even working on a separate version for accommodations, which would allow an AI to find a place to stay based on availability, dates, and preferences, and ultimately initiate a reservation.

In other words: this isn’t just about sneakers and coffee machines. The same principle can apply to vacation homes, hotel rooms, cars, tickets, or other services.

UCP is not yet available to merchants in Europe and the Netherlands, but it does clearly show the direction in which AI-commerce is heading.

The protocols differ.

The development does not.

AI shouldn’t just be able to read information about your offerings. AI needs to be able to work with it.

Hey, didn’t we talk about MCP earlier?

Absolutely.

And here’s where our previous blog post on MCP and A2A suddenly ties in nicely with e-commerce.

The Model Context Protocol(MCP) wasn’t developed specifically for online stores. It’s an open standard that allows an AI system to access external data and tools.

For example, you can use it to give an AI a tool to search for products, check inventory, retrieve up-to-date information about a specific product, or verify the availability of a vacation home.

That might sound like a distant dream, but Booking.com already has an official MCP server. An MCP-compatible AI can use it to search for available accommodations based on location, dates, number of guests, and other preferences, among other things.

That’s exactly why MCP is so interesting.

You’re not building for just one chatbot.

You’re making a system accessible to AI.