Who this article is for: Merchants using Plug in AI who want to understand what AI shopping agents see, and what a product needs before agents can recommend it properly.
What you'll learn: The two places agents read your products from, the four things a product needs, and where to go for each one.
AI shopping agents — ChatGPT, Copilot, Gemini and others — read your products and answer shoppers' questions with what they find. This article explains what they can see, and what you can do about it.
Where agents read your products from
There are two separate surfaces, and they don't work the same way.
The Shopify Catalog
The cross-store feed that ChatGPT and other agentic storefronts read. This surface reflects your Catalog Mapping, so it's the one that can show your agent description instead of your storefront description.
For a product to be there at all, your store has to be enrolled in the Agentic sales channel in your Shopify admin, and the product has to be published to it.
Your store's own agent search
Your store's agent endpoint, which any agent can query directly. It always uses your storefront description and ignores Catalog Mapping.
That difference catches people out: changing the agent description does not change what your own store's agent search returns. If the agent catalog doesn't return a product at all, it may be unpublished from your Online Store.
The four things a product needs
1. An agent description — a separate, fact-dense description written for machines. It never changes your storefront description.
2. A product category — Shopify syncs products to agents by category, so a product without one doesn't sync properly.
3. Answered questions — the specifics shoppers ask about, in a form agents can quote. Aim for at least three.
4. The two one-off setup steps — adding the Q&A block to your theme so shoppers can see the questions, and mapping the Shopify Catalog to your agent description so agents receive it. You only do each of these once.
How to tell what's missing
Two places show you where a product stands:
The Optimize products list gives you an Agent score per product, a Q&A count, and the state of the description and meta fields. Sort by agent score to find the products with the most to gain.
Agent readiness issues, on the product page itself, lists what's still missing — for example "No agent description yet" or "No product category."
Work through those two and you'll know exactly what's left.