Google Shopping Graph
The Shopping Graph is Google's product data layer: a continuously updated model of products, sellers, prices, reviews and availability, assembled from Merchant Center feeds, crawled retailer pages, manufacturer data and other sources. It is what shopping results, Shopping in AI Mode, and Google's agentic shopping features draw on, which makes feed accuracy and page accuracy two inputs to the same system.
What it is made of
It is an entity graph over products rather than an index of pages. A product exists as a node, with sellers, offers, prices, images, reviews and attributes attached, and Google reconciles many sources into that single node.
The inputs are the ones you would expect: Merchant Center feeds, crawled retailer and manufacturer pages, structured data on those pages, and Google's own data. Public statements about its scale have been made at keynotes and change frequently; the size claim matters less than the mechanic, which is that your feed and your page are both evidence about the same node and are expected to agree.
Matching is the whole game
For your offer to attach to the right node, Google has to match your listing to a known product. That is a product matching problem, and it runs on identifiers first: GTIN where one exists, brand plus MPN where one does not.
B2B catalogs fall out here more often than anywhere else. Plenty of industrial SKUs never received a manufacturer-assigned GTIN, so brand and MPN carry the entire identity load, and an MPN mangled by stripped dashes or leading zeros fails to match. Google's own guidance is that GTIN is strongly recommended where the manufacturer assigned one, and that products genuinely lacking one should submit brand and MPN instead.
A mismatched offer does not error. It quietly attaches to nothing, or worse, to the wrong node, and you find out through a support ticket about the wrong part.
Where the industrial catalog strains it
The Shopping Graph is built around retail assumptions, and several of them do not hold in distribution.
- Identity without GTIN. Handled, but less reliably than the GTIN path.
- Selling unit versus pack. Feeds distinguish
unit_pricing_measure,multipackandis_bundlefor good reason; a catalog with one flat pack-size column cannot express its own commercial shape. - Technical specifications. The
product_detailattribute acceptssection_name : attribute_name : attribute_valuetriples, which is the right place for specs and is very widely left empty. - Branch-level availability. National stock is what the feed expects, and it is not how distribution works.
- Gated pricing. Feeds require a price. Distributors who publish price only to logged-in accounts are largely outside this system by choice.
None of those are reasons to ignore it. They are the reasons a distributor's Shopping Graph presence is usually thinner than their catalog would justify.
Frequently asked questions
What is the Google Shopping Graph?
Google's continuously updated model of products, sellers, offers, prices and availability, built from Merchant Center feeds, crawled pages, structured data and other sources. It underlies shopping results, Shopping in AI Mode, and Google's agentic shopping features.
How does my product get into the Shopping Graph?
Primarily through a Merchant Center feed, supplemented by crawled product pages and their structured data. Your offer then has to match to the correct product node, which is driven by identifiers: GTIN where one exists, brand plus MPN where it does not.
What happens if my products have no GTIN?
Google's guidance is to submit brand and MPN instead for products the manufacturer genuinely never assigned a GTIN to, which covers a large share of industrial SKUs. The matching is less reliable than the GTIN path, so consistent MPN formatting across every system becomes critical.
Can distributors with gated pricing participate?
Only partially. Shopping feeds require a price, so catalogs that publish price only to authenticated accounts are largely outside the feed-driven surfaces by design. Product pages and structured data still contribute to general retrieval and AI answers.