9 of 20 stores give AI agents no catalog to read

What an agent is looking for

An AI shopping agent does not browse. It requests a machine-readable list of your products, with prices, variants and stock, and builds its answers from that. On Shopify-class platforms that list has a standard address, and product feeds serve the same job for shopping systems. If the list is not reachable, the agent sees your storefront the way a courier sees a locked warehouse.

A human never notices this. The store looks perfect in a browser, sells in a browser, ranks in a browser. The gap only exists for the machine readers, and they do not send a complaint.

The finding

9 of the 20 stores in our July sample expose no machine-readable catalog at all. On the access side that splits into 7 passive defaults, where the catalog endpoint is simply closed, and 2 active blocks, via robots.txt rules or bot protection.

The intent side is the part that stays with me. Only 2 of the 9 show any provable, deliberate decision about AI access. For the rest, everything points to a default nobody has looked at. That is a reading of public signals, not a survey of the merchants. And because two of the nine provably acted deliberately, it is most of them. Not all.

Why it goes unnoticed, and how to check

Nothing in a store's daily numbers flags this. No error log, no dashboard warning, no customer complaint. The channel where the gap costs visibility is one the store owner does not watch, because it never sent traffic before.

Checking it takes a minute. Open your own store's catalog endpoint in a browser, on Shopify that is your domain followed by /products.json, and see whether products come back. Or let the free scan read your store the way an agent does, catalog access included.