How Shopify Catalog groups products across stores
Shopify explains cross-store product matching, daily cluster builds and why a missed match does not mean a listing has disappeared.
Shopify Catalog groups related listings from different stores so AI agents can recognise the same underlying product. A missed match does not necessarily mean a product has been removed from Catalog.
Shopify’s October 20 engineering explanation says an unmatched listing remains searchable under its own identifier until a later build finds the connection. That is a useful distinction when investigating an apparent discrepancy in AI product discovery.
Product identity comes before comparison
The system brings related listings under a Universal Product Identifier, exposed as UPID in Catalog. Shopify describes a process that retrieves plausible pairs, scores their relationship and builds clusters. Text, images and other metadata provide different pieces of evidence. It favours precision when combining groups because an incorrect merge can combine genuinely different products.
Our merchant takeaway is to make the identity of each item unambiguous. Check model names, compatibility, pack size and variant details across the title, description and photographs. Do not copy a competitor’s wording merely to look similar: the goal is an accurate description of the item you actually sell.
Updates are not an instant matching promise
Shopify says clusters are built daily while more expensive comparisons continue in background queues. A daily build therefore does not imply every possible comparison has finished. The explanation is not a ranking guarantee or a promise that any particular edit will produce a match by the following day.
When reporting an issue, keep the original product URL, the exact variant, the conflicting result and the time observed. Distinguish an incorrect grouping from missing availability, a stale price or an unexpected recommendation. Those observations give support a more precise problem than a general claim that AI cannot find the store.
Keep a controlled record
Change one clear data defect at a time and record the before-and-after values. Repeat the same lookup with the same product and market context. If the result changes, do not automatically credit a title edit: the catalogue and the agent’s response can also change independently.
For developers managing an integration, identity matching and endpoint migration are separate jobs. See our Catalog REST API migration coverage for the November 2 deadline.
Shopify Engineering’s October 20 Catalog announcement
Read the original announcement on X.
More from AI
See all →
