When AI agents shop for your customer: how to prepare your webshop

Your next customer may be an AI agent

More and more people start their shopping by asking an AI service instead of running a search. "Find me a waterproof hiking jacket that survives the washing machine and doesn't cost too much." The answer is a handful of products, and only the stores whose product data can be read and compared make the list.

This is not a future scenario to wait for. Shopify has built it into the platform, and whether your store keeps up depends largely on how well your product data is organised. The good news is that this preparation pays off no matter which AI service wins.

What has happened since spring: Shopify has made AI shopping a standard feature

Since spring 2026, products from eligible Shopify stores can be shown and bought in AI services through Agentic Storefronts. The feature is switched on by default for stores that meet the requirements, according to Shopify Help Center. Merchants don't have to actively opt in.

Still, the picture is more nuanced than the headlines. Some parts have already been pulled back, and most of it only applies to buyers in the US for now. That is why we look at both what is real and what is hype.

How it works: Agentic Storefronts, Shopify Catalog and UCP

Three terms keep coming up: Agentic Storefronts, Shopify Catalog and UCP. They sound technical, but the idea is simple. Your product data should be as easy for a machine to read as your product page is for a customer.

Shopify Catalog is the data layer that makes products readable to AI agents. It standardises products in a shared taxonomy, checks price and inventory in real time and distributes the data to connected AI platforms. Merchants don't need to install any apps, according to Shopify.

Which AI channels are included

Products can appear in ChatGPT, Google AI Mode, Gemini, Microsoft Copilot and Meta. They reach the channels via Shopify Catalog. For Google AI Mode and Gemini, they can also go through the Google & YouTube channel or Google Merchant Center, according to Shopify Help Center.

Shopify states that there are no channel fees for selling through Agentic Storefronts, only the usual payment fees, according to eSEOspace. The fee that a secondary source reports for ChatGPT's Instant Checkout belonged to a feature that has since been retired, so don't mix the two up.

UCP explained without the jargon

UCP stands for Universal Commerce Protocol, an open standard developed jointly by Shopify and Google and presented on 11 January 2026. Think of it as a common language between AI services and stores. With it, an AI agent can find products, build a cart, complete a purchase and handle post-purchase matters without every store needing its own custom setup.

Support comes from companies including Walmart, Target, Etsy, Visa, Mastercard and Stripe, according to Shopify. The standard is evolving quickly, and a new version arrived in late summer with, among other things, more secure payments. That is mainly a matter for developers, and Shopify handles the updates for the store, according to UCP's own documentation.

As part of this development, some older interfaces were retired in August. That mainly concerns developers and headless stores with their own integrations. A regular Shopify merchant normally doesn't need to do anything.

Who owns the customer and the checkout

In ChatGPT, the purchase happens in the store's own checkout. The order appears in Shopify Admin with a reference to ChatGPT, and the merchant remains the seller. Google, Copilot and Meta, on the other hand, may use a Shopify-powered checkout inside the AI channel itself, according to eSEOspace.

Shopify states that orders from AI channels appear in admin with channel or referral attribution, and that the merchant keeps ownership of the customer relationship and the post-purchase experience, according to Shopify Help Center. Your checkout, your terms and your customer contact therefore still matter.

Hype or reality?

When a new channel launches, it's easy to believe everything will change overnight. What happened with buying inside the chat is a good reminder that reality moves at a different pace.

Buying directly in ChatGPT never took off at scale

ChatGPT's Instant Checkout, meaning purchases directly in the chat, launched in September 2025 and was pulled back in March 2026. OpenAI now sends the user to the seller's own website to complete the purchase. Forrester describes this as the answer engine becoming a referral, just as search engines have always been, in its analysis on Forrester.

The feature also never reached the scale that was promised. Naturaily writes that Forrester analyst Emily Pfeiffer counted fewer than 30 live merchants on Shopify in February 2026, compared with the million that had been promised. Other sources give even lower figures, so think of it as a few dozen. This is from Naturaily's guide.

What actually works: AI that finds and recommends products

Naturaily's conclusion is that discovery through AI is real and growing, while purchases inside the chat have mostly stalled. The article also reports that Walmart's in-chat purchases converted at roughly one third of the rate of purchases on walmart.com. That is a single report, but it points in the same direction.

According to several secondary sources, AI-referred traffic and AI-attributed orders have grown strongly since early 2025, mainly in the US and from low levels. The figures should therefore be used with care. The point is that AI is becoming a new way into your store, not that it replaces your store.

The lesson: merchants who invested in their own store and their own catalog benefited, while those who rebuilt everything around a single feature had to throw away months of work. That is one blogger's assessment, but it is worth taking on board, see Taylor Sicard.

And Sweden? Mostly the US today, but the development is coming

Today, purchases in the AI channels are only available to buyers in the US, and built-in checkouts are only shown to customers there. The features are globally available to businesses that sell to US buyers, and Shopify says merchants should be ready by default as AI partners expand to more regions, according to Shopify.

The requirements are clear. Products must be eligible for Shopify Catalog and the store must sell to customers in the US. Google AI Mode, Gemini and Copilot also require a US-based store, according to Shopify's requirements. So it is not possible to promise Swedish merchants that their products appear in these channels today.

We haven't found a confirmed timeline for Sweden. A secondary source says Google has announced an expansion of UCP-based shopping to, among others, the UK, but Sweden isn't mentioned, so treat that as unconfirmed. That is why we say "when the features reach Europe" rather than naming a date.

How to prepare your webshop, on Shopify and WooCommerce

What makes preparation a sound move is that product data is reused everywhere: in Google Shopping, in on-site search, in AI search and in product feeds. Parts of the payment experience in the AI channels aren't here yet, but the data is useful today. This reasoning is supported by WisdmLabs, and it is where SEO and GEO meet the webshop.

Shopify syndicates title, description, options, images, price, inventory status and other key attributes. With Shopify Catalog Mapping, you can connect your own metafields and metaobjects to the attributes the AI channels read, according to Cuebase. Start with your 20 to 50 most important products, not the whole range.

Clean product data: titles, categories, attributes and GTIN

This is the foundation. According to OptiPhoenix, these measures have the most effect in Shopify:

1. Use clear titles with product type, material, size and use instead of creative names.
2. Choose the right product category, as specific as possible.
3. Fill in the category's metafields and move facts from running text into fields.
4. Add barcode (GTIN) and brand.
5. Give unique details for each variant.
6. Make sure price and inventory are correct.
7. Use images with descriptive alt text.

A common Swedish case is merchants with their own or private-label brands that lack barcodes. That doesn't necessarily mean bad data. Systems often let you mark the product as having no GTIN instead of leaving the field empty.

Clear descriptions: facts before sales copy

A description like "a luxurious feel for everyday life" gives an AI agent nothing to work with. Concrete details about material, weight, washing instructions and sizes, on the other hand, can be matched against the customer's question. The example comes from Seresa.

The sales copy doesn't need to disappear, it just needs to come after the facts. Give each detail its own place in a field, and don't leave facts hidden in images or long paragraphs.

Correct inventory and prices everywhere

Price errors and incorrect inventory levels are among the biggest risks. If the price in the store doesn't match what the AI channel showed, the purchase fails, according to eCommerce Development Pros. Prices and inventory must therefore be the same in every system: ERP, webshop and feeds.

This job tends to get neglected when there are many systems. Go through where the figures actually come from and which system is the source of truth.

Structured information: schema, metafields and Knowledge Base

AI agents read structure, not mood. Metafields, attributes and schema put the information where machines look for it. Shopify Knowledge Base, with questions and answers about shipping, returns and warranty, as well as templates like llms.txt and agents.md, is documented in Shopify Help Center, according to Brandactive.

Return terms and shipping rules should be clear and correct, especially given EU consumer rights. That applies today, and becomes more important once an AI agent is relaying them to the customer.

Especially for WooCommerce

AI agents read structured fields such as attributes, feeds and price fields, not sales copy. If the data doesn't answer the customer's question, the agent recommends a competitor whose data does. WooCommerce automatically maps many core fields, such as title, description, image, SKU, price, currency and stock status, to Schema.org JSON-LD. Richer fields and variant data may require extensions or custom code, according to WooCommerce.

According to a secondary source, WooCommerce has no built-in UCP support, only third-party plugins. That is why data quality is the safest investment no matter which protocol wins. Check that:

• every product has a GTIN, MPN or brand plus SKU
• size, colour and material are stored as attributes
• every variant has its own price, stock status, SKU and image

This comes from WisdmLabs. WooCommerce itself recommends a short audit of your product data and testing your pages with Google's Rich Results Test. Plugins such as Yoast create Product schema automatically, but the question is whether that is enough, so check price, availability, SKU, GTIN and brand. See WooCommerce.

Common mistakes to avoid

We see these often, and they also show up in the sources we've reviewed, including eCommerce Development Pros:

1. Building around Instant Checkout or promising purchases in the chat.
2. Creative product names and sales copy without facts.
3. Facts hidden in images or long texts.
4. Missing GTINs and categories.
5. Different prices or inventory in the store, feed and schema.
6. Doing the whole range at once instead of the bestsellers first.
7. Assuming Sweden is already included.

An eighth mistake is believing that visibility in the AI channel is the same as being competitive. Products can exist in Catalog and still not be recommended if the data is thin.

What to take with you

• Shopify has made AI shopping a standard feature, but buying directly in ChatGPT never took off at scale. The purchase happens in the store's own checkout.
• The AI channels mainly target US buyers today, and we don't know when they will reach Sweden. Don't assume your store appears there today.
• Preparation isn't a bet on a single technology. Clean product data works in Google Shopping, on-site search, AI search and feeds.
• Start with your 20 to 50 most important products. Clear titles, the right category, facts in fields, GTIN and correct prices and inventory go a long way.
• This applies to both Shopify and WooCommerce. Don't build for a single AI channel, build product data that works everywhere.

Want a store that is ready?

If you'd like help going through your product data, on Shopify or WooCommerce, get in touch and we'll look at your bestsellers together. You can reach us via our contact page.