You can do this yourself this evening. No agency. No expensive app. Just an hour of your time and your store’s admin.
Seven fields in Shopify that decide whether your product appears in answers from ChatGPT, Gemini, Copilot and Perplexity. No developer, no agency.
A barcode on every variant. A category from the standard taxonomy. A description that is not a copy. Category metafields. Clear policies. Clean images. At least 50 reviews. An hour of work on the top 20% of your catalogue is the highest return per unit of time available to you.
The two previous articles established two things. First: a third of shopping queries now end in an AI answer rather than a visit to a website. Second: Shopify already passes your product data to ChatGPT, Gemini, Copilot and Perplexity through the Catalog service — and the quality of that data decides whether your products appear in answers.
Now, what to do about it.
I have picked out seven fields that have the biggest effect on visibility in AI engines and that you can check and fix yourself. No developer, no agency, no app. Just open the store admin and spend an hour.
the barcode field on every product variant
products → pick a product → the variants section → each variant has its own barcode field
This is the shortest action with the highest return. The Barcode field holds the product’s international barcode — what the standards call a GTIN. Catalog uses it to work out what your product actually is and to cross-check it against the manufacturer’s database. Without a barcode, Catalog ranks the product one notch lower than it could.
The most common mistake: the barcode is filled in on one variant but not the rest. Or only on the “base” product. Catalog works with variants individually — an empty barcode field on a variant means, as far as it is concerned, that the product does not exist.
If you manufacture in-house and have no manufacturer’s barcode, it is acceptable to use the manufacturer’s article number or an internal code. What matters is that the field is not empty.
a category from the standard shopify taxonomy
on the product page, the product category field in the right-hand column
Shopify has a built-in hierarchical map of tens of thousands of categories — the Shopify Standard Product Taxonomy. Every product needs to be pinned to a point on that map at the deepest level available.
Not “Clothing → Other”, but “Apparel & Accessories → Clothing → Shirts & Tops → Men’s long-sleeve shirts”.
The more precise the category, the better Catalog understands which answers your product can belong to. A shopper asks about “a men’s long-sleeve shirt” — Catalog looks for exact category matches first, then works through the other fields.
a description that does not repeat the manufacturer’s
on the product page, the main description
If you resell products and take the description from the supplier’s site, Catalog can see that fifty other stores carry exactly the same text. And it has no idea why anyone should pick yours.
A description needs concrete specifications, facts and use cases. Not “this coffee machine will be a faithful helper in your kitchen”, but “brews espresso 90 seconds from switch-on, 19-bar pump pressure, a 1.8-litre reservoir, works with beans and ground coffee”.
category metafields
on the product page, the metafields section → category metafields
These are the most underrated fields in the whole of Shopify. For each product category the platform offers its own set of attributes: for clothing — material, size, fit, colour, length; for electronics — power, compatibility, warranty; for food — ingredients, allergens, country of origin.
Catalog uses these fields to match a product against a specific query. If a shopper asks for “a white linen shirt”, Catalog looks for the value “linen” in the “material” field. If the field is empty, your product drops out of the shortlist even if the word “linen” appears in the description.
For the top twenty per cent of your catalogue it is worth filling at least 80% of these fields. It is the most honest measure of how ready a store is for agentic commerce.
returns policy and delivery terms
settings → policies for the whole store; and make sure they are reachable from the product page
When a shopper asks an AI “can I return this”, the AI does not go and read your site. It asks Catalog: “What is this store’s returns policy?” Catalog answers with whatever you have written.
If the policy is generic and vague — “by agreement” — the AI gives the shopper a vague answer. If it is specific — “free returns within 30 days, courier collection, refund to your card within 5 business days” — the AI is specific too. And the shopper trusts it.
images that show the product, not just the marketing
on the product page, the media section
Catalog checks that images match the product. Images heavy with text, watermarks, busy backgrounds or collages showing several products at once lower a product’s standing in Catalog.
The main image should show the exact product being sold — ideally on a neutral background, at high resolution, with no captions.
And one thing that is often missed: if a product has variants in different colours, each variant needs its own image in that colour. Not one shared photo.
at least 50 reviews on every key product
depends on your reviews app, but the goal is a real flow of customer reviews
AI engines treat reviews as a signal of sentiment and quality. A product with five reviews is a “thin entity” to them, lacking the data for a confident recommendation. A product with five hundred reviews and replies from the store is “an entity with plenty of fresh signals”.
Fifty is the minimum threshold below which a product drops out of consistent inclusion in answers. If your key products have fewer, that is your first area for growth.
what comes next?
after the seven fields comes the audit
These seven fields are what you can and should do yourself. It is an hour or two of work for the top twenty per cent of your catalogue. With a thousand products it is a week or two for a content manager.
After that comes the next level, the one that usually needs an outside view: checking that data actually reaches Catalog from Shopify (it does not work by default on every theme); reconciling the markup on the site against what is in the admin (they often disagree); measuring your share of mentions in each AI engine and identifying which specific queries are dragging you down; configuring Catalog Mapping so that what goes to the AI engines is purpose-built data rather than the default fields.
But that is no longer an evening task. That is an audit.
a full AEO audit of your store
Once you have closed these seven fields, we have a next step for you. We check all seven layers of AI visibility, measure your share of mentions across four agent channels, and reconcile the data between the admin, the markup on the site and what Shopify actually sends to Catalog. You get two reports: a one-page summary for the owner and a full one for the team, with a list of specific actions and a time estimate for each.
see the audit →series · visibility in AI engines
- shoppers no longer come to your site
- shopify already sends your products to chatgpt
- seven fields in the shopify admin ← you are here
· · ·