Back to blog

On Shopify, AI Agents Now Get Your Product Fields, Not Your Product Page

Nov 2, 2026

A brightly dressed glass shop window at night with nobody in front of it, while a line of small lanterns carrying flat glowing cards slips out through a plain side door in the brick wall.

Shopify shipped its Fall '26 Edition on 28 October, and most of the coverage went where you would expect: checkout inside AI agents, Shop Pay in Claude Duet and Muse, a rebuilt Sidekick. Underneath all that is a sentence that matters more for anyone who owns a catalog. In Shopify's own words, ChatGPT, Copilot, Meta, Instinct and other agents “get product information from Shopify Catalog”.

Read that again with your product pages in mind. The page you designed, with the lifestyle photo, the size chart and the three-paragraph story under the fold, is not what those agents read. They read a structured record of each product, assembled from your admin data. The theme never enters the conversation.

Three doors into your catalog, and none of them is your theme

Three doorways of different shapes cut into one long indigo wall, each glowing a different warmth, with faint trails of light running from them to a single low shelf of product shapes.
Catalog, browser tools and the knowledge base all reach the same products. None of them passes through the storefront design.

The Edition describes three ways an agent now reaches a Shopify store. The first is Shopify Catalog itself, which Shopify's help centre calls “a structured source of product information gathered from eligible products”. Shopify's new shop.md skill lets a shopper point any agent at it and search more than a billion products, and the Edition's merchant post names ChatGPT, Copilot, Meta and Instinct among the agents already drawing from it.

The second is WebMCP, a set of tools every Liquid storefront and checkout now exposes to browser agents. Shopify's developer post puts the difference plainly: agents “get structured data back instead of reading screenshots and simulating clicks”, and “every Shopify store is agent-ready by default”. An agent that lands on your store in a browser no longer has to interpret your layout. It asks the store for products and gets fields.

The third is the knowledge base, now a page in the admin under Sales channels and Agentic. It holds the non-product facts, return windows, shipping destinations, gift services, drafted from your storefront, and agents “only get facts the merchant has published”. Of the three, it is the one where Shopify asks you to approve before anything goes out. The other two are on by default.

What actually travels

Shopify's help centre is specific about what Catalog supplies: “titles, descriptions, images, pricing, options, and availability”. It asks merchants to keep “titles, descriptions, images, categories, variants, options, prices, and inventory” up to date, and to provide “accurate, detailed, and descriptive product information”. That list is the whole product as far as an agent is concerned.

Shopify does some work on top. Its own explainer on Catalog says products are mapped to the Shopify Standard Product Taxonomy and structured into key fields such as key features, technical specifications and a value proposition. The sentence to underline is this one: Shopify “will infer missing attributes when the product data is clear enough”. Clear enough is doing a lot of work. A description that says the mug is 350 ml, stoneware, dishwasher safe and made for slow mornings gives the inference something to hold. “Great mug, ships fast” does not, and no amount of theme design rescues it, because the theme is not in the record.

Two more facts from the same help page, because honesty is cheaper than surprise. You cannot opt out of Shopify Catalog itself, only switch individual agentic storefronts off. And inclusion “doesn't guarantee that a product will appear in a specific AI answer”, since each channel controls its own ranking and wording. Your fields decide what an agent could say about you. The agent decides whether it does.

The new mapping page is an audit you did not have to build

A grid of identical translucent cards laid out on a table seen from above, most filled with layered amber and teal light and a few left as clear empty frames.
Coverage is the honest number: how many products have anything at all in the field an agent is about to read.

The quietest item in the Edition may be the most useful. A changelog entry from 28 October adds a channel mapping page where you manage the product data sent to each sales channel and preview what a channel receives for any product. Each channel field pulls from a default source, you can change that source for every product on the channel at once, and you can combine sources; the example Shopify gives is adding the brand name to the product title. You can override a field for one product on one channel, and bulk edit the rest.

The part to go and look at today is the coverage percentage beside each field. It tells you how many of your products actually have a value for that field, and Shopify says it is there to help you find and fill missing values. For most stores the first reading of that number will be uncomfortable. It is also the most honest audit of agent readiness you can get, because it measures the record the agents receive rather than the page shoppers see.

  1. Open the new admin view that shows whether your store is eligible for Shopify Catalog. It says what to change if it is not.
  2. On the mapping page, read coverage for description, images and category before anything else. Those are the fields the inference leans on.
  3. Fix the top sellers by hand this week, then use the coverage view to see how long the tail really is.
  4. Open the knowledge base and publish the facts you recognise. Unpublished drafts help nobody.

Filling the fields at catalog size

A wide loom strung with hundreds of thin glowing threads, a shuttle passing through the gaps and leaving a dense bright finished cloth behind it.
Drafted for every product at once, approved one at a time, written back into the record the agents read.

This is where CortexCommerce fits, and it fits because the job has changed shape. When the agent reads a field rather than a page, the quality of every description, title and tag counts for every product, not just the twenty on your homepage. CortexCommerce reads the product data and images you already have and drafts the missing fields for the whole catalog: descriptions in your brand voice that say what the product is, who it is for and what it is made of, unique SEO titles and meta descriptions, consistent tags, and alt text written from what the AI actually sees in each photo.

Every draft waits in a review inbox with the before and after side by side, nothing reaches the live store until you approve it, and price, SKU, barcode and inventory are locked, so an enrichment run cannot touch the numbers Catalog cares most about. Approved copy is written straight back to Shopify, which means the coverage percentage on the mapping page moves the moment you approve, with no export and no re-import. For a 3,000-product store, filling the description field by hand is weeks. Approving drafts is afternoons.

We cannot tell you how ChatGPT or Copilot will rank what they find, and neither can Shopify. What we can say is that when an agent asks for your product record, every field will have something worth reading in it. To see how your store reads today, run a free scan at cortexcommerce.ai/scan and start with the products it shows as thin.

Sources

  1. Shopify News: Exponential, Shopify's Fall '26 Edition (28 October 2026)
  2. Shopify News: Bringing shopping into the future, the Fall '26 Edition for developers
  3. Shopify changelog: Map product data by sales channel (28 October 2026)
  4. Shopify Help Center: Shopify Catalog
  5. Shopify blog: What Is Shopify Catalog and How Does It Work?