For fashion brands
← Blog

Field guide7 min readNovember 17, 2026

Shopify Fall '26 put checkout inside AI agents. Fit has to be data, not just a widget.

Shopify's Fall '26 Edition lets shoppers check out inside AI agents. Agents read your catalog and knowledge base, not your size widget. Here's what to fill in before they sell your sized products.

A man in an olive overshirt and cream trousers sits on a green sofa at dusk, shopping on his phone, with a city skyline behind him.

Shopify released its Fall '26 Edition on 28 October 2026. Most of it is about selling where the shopper is talking to an AI rather than browsing a store. Per Shopify's Edition page, as of November 2026 shoppers can check out with Shop Pay inside Muse, Instinct and Claude Duet. Merchants can sell in Google Search's AI Mode and the Gemini app with checkout powered by Shopify. A "shop skill" lets ChatGPT, Claude and other agents find products and check out. Shopify's newsroom says agents can search "over a billion products from Shopify merchants, with current prices and inventory."

In September we wrote about what happens to fit when the product page disappears. Back then it was one brand selling handbags inside Gemini. Now the plumbing ships with the platform. So the question has changed. It used to be "will this happen to my store?" Now it is "what will an agent know about fit when it does?"

What the agent can read about your clothes

An agent can't scroll your product page. It reads structured data. The Fall '26 Edition is clear about where that data comes from:

  • Your catalog. Products, variants, prices and stock. You can now "customize and map your product data for each sales channel from your admin."
  • Your knowledge base. You can "review and approve the facts that get shared with agents, including policies and FAQs."
  • The Ask endpoint. Shopify's developer docs describe one tool, ask_business, that "returns an answer from the store's knowledge base, policies, and product information."

Look at what that list leaves out. A size-guide pop-up, a fit quiz and a try-on button are all things that open on a page. The agent never loads the page. If your fit answer exists only as a widget, the agent can't see it.

What Shopify's own data model says about fit

Shopify already has a place for fit facts. Its Standard Product Taxonomy gives each category a list of attributes, and you fill them in as category metafields. As of November 2026, the Jeans category lists 21 attributes, including size, size type, fit, waist rise, stretch level and pants length type.

Look at what those fields hold. Fit is a style label: slim fit, boyfriend, loose fit, classic fit. Size is the label on the tag. Both describe the garment. Neither says anything about the person buying it.

The same taxonomy file also lists return reasons for jeans: waist too big, waist too small, too long, too short. Since January 2026, Shopify suggests reasons like these for each category, in the admin, at POS, in self-serve returns and in the Shop app. So the platform has a precise way to record a fit failure after the sale. Before the sale, it has no way to record which size fits which body.

That gap is the whole point of this post. An agent can truthfully say "these are slim-fit jeans with a mid rise, in 28 to 36." It can't say "you're a 31 in these" unless something gives it that answer.

Widget versus data

For the last decade, a fit tool has been a thing on the product page. A shopper opens it, answers a few questions or uploads a photo, and gets a size or a picture. That works while the shopper is on the page. In an agent flow, three things break.

  1. The input has nowhere to go. A widget gathers measurements on the page. An agent conversation has no page, so the measurements must come from data the shopper already gave, or be asked for in the chat.
  2. The answer has nowhere to show. A size shown in a pop-up never reaches the chat. To be used, the recommendation has to come back as a value an agent can read and repeat.
  3. The return loses its cause. Shopify's newsroom says an order from an AI channel "lands in the merchant's admin with the customer's name." The return reason will land there too. But nothing will tell you whether the shopper got a size recommendation before buying, so a jump in "too small" can't be traced back to anything you could fix.

"Fit as data" means three things you can check. Garment facts are filled in as structured fields. A shopper's size for that garment comes back as an answer from a service, not a screen. And each order keeps a record of the size that was recommended, so returns can be matched against it.

What to do before agents sell your sized products

Most of this costs nothing and doesn't need a fit vendor.

1. Fill in category metafields on every sized product

Set the most specific product category, then fill in size, size type, fit, rise and length type. Use your own words in the product description, but put the facts in the fields. That's what gets mapped to each sales channel.

2. Put your fit notes in the knowledge base

If your trousers run small in the waist, or your knits are cut for layering, write it down as a fact you approve. The Ask endpoint answers from the knowledge base, so a fit note there can reach an agent. A fit note inside a widget can't. Keep it specific: "size up one in the 1960s cut" is useful, "fits beautifully" is not.

3. Publish real garment measurements

Flat measurements per size (waist, inseam, chest, length) let any system turn a shopper's body into a size. Without them, the best an agent can do is repeat a size chart. We explain why girth beats height and weight in this sizing post.

4. Start with unsized products

Bags, scarves, jewellery and one-size knits have no fit question. If you're testing Gemini or a shop skill, list those first and keep sized products on your own site until the fit answer can travel with them.

5. Tag the channel on every return

Compare too-big and too-small returns for agent orders against store orders, by category, each month. That's how you'll know whether agentic checkout is costing you on fit.

Where a widget is still the right call

Most apparel orders still come through your own product pages, and that will be true for a while. If agent orders are a rounding error for you this quarter, a good on-page size finder is still the best spend. If your sizing is simple and your returns are low, a clear size chart may be all you need. Our comparison of Shopify size finder apps covers cheaper tools that do that job well. Large retailers with years of purchase history may be better served by platforms built on that history. Ten sizing platforms compared explains which approach suits which store.

The argument here is narrower. Whatever tool you use, make sure the answer exists as data, not only as something drawn on a page.

Where Garu sits

Garu is the personalization layer and fit-data platform for fashion commerce. A shopper gives a photo and measurements. Garu returns a size matched to the merchant's own chart, with a fit note and a rendered try-on, front and back. The widget is one way to show that. Underneath, the size and the render come from services that don't depend on a product page. Opening them to agent surfaces is work in progress, and we'll write about it when it ships, not before.

What we can show today is what shoppers do with a size on a real store. On DEWEY, a petite brand, Garu sized 749 shoppers between 4 September and 2 November 2026, and 98% of shoppers who opened Garu got a size. Between 24 September and 14 October 2026, 14 of 27 orders (52%) came from shoppers who got their Garu size first, and those orders were 38% larger. These numbers come from matching Garu sessions to Shopify checkouts. They show what Garu shoppers did. They are observational, not a controlled test, and don't prove Garu caused it. The full details are on our case studies page.

Those shoppers got their size on a product page. The aim is for the same answer to reach them wherever they check out.

Questions merchants ask

What did Shopify's Fall '26 Edition change for checkout?

As of November 2026, shoppers can check out with Shop Pay inside Muse, Instinct and Claude Duet, and merchants can sell in Google's AI Mode and the Gemini app with checkout powered by Shopify. A shop skill also lets ChatGPT, Claude and other agents find products and check out.

Can an AI agent use my size chart app?

Not if the app only runs as a widget on the product page. Agents read structured catalog data, the knowledge base you approve and your policies. A size chart or quiz that only opens on the page isn't part of that.

What fit data does Shopify already store?

Category metafields from the Standard Product Taxonomy. For jeans, that includes size, size type, fit style, waist rise, stretch level and length type. These describe the garment, not the shopper.

Should I keep dresses and trousers out of agentic checkout?

For now, consider starting with unsized products and watching return reasons by channel. Move sized products over once your fit facts are filled in and a size answer can reach the agent.

Does Garu work inside AI agents today?

Garu runs on your product pages today. Its sizing and render come from services that don't depend on the page, and opening them to agent surfaces is in progress. We'll say so when it ships.