Knowledge Graph

A knowledge graph is a database of entities, the people, places, organizations and things a system knows about, and the relationships between them, stored as nodes and edges rather than as pages. Google’s Knowledge Graph is the best known, and it is what its knowledge panels are filled from.

Nodes, edges and names

Each entity is a node with a stable identifier; each fact about it is an edge to another node or to a value. Google’s Knowledge Graph Search API shows the shape from outside: a query returns matching entities as JSON-LD, each typed with a schema.org type such as Person, Place or Organization, with a name, a description, an image and a score for how well it matched. The same page says the API returns single entities rather than the graph around them, is read-only, is not to be relied on in production, and is being moved to a newer product. It is a window on the graph, not the graph.

How a business becomes a node

An engine adds an entity when it can reconcile the same thing across sources it trusts: the business’s own site, its profiles, references elsewhere, all agreeing on the name, the address, the phone number and what it does. Disagreement is the failure mode. Across 9,483 small-business sites audited by this site in September 2026, 43.1% were inconsistent about name, address or phone across their own three main pages, 36.9% published more than one phone number, and 34.1% had no postal address a machine could parse. A system reconciling entities has nothing firm to reconcile against. Schema helps here: an Organization or LocalBusiness block with the identity stated once, and sameAs pointing at the profiles that describe the same entity, tells the graph builder which records are the same thing. The audit is in what 9,483 small business websites serve to an AI crawler.

A common question

What stops a search engine from adding a business to its knowledge graph?

Disagreement between sources. An engine creates an entity when the business's own site, its profiles and references elsewhere line up on the name, the address, the phone number and the trade; when they contradict each other there is nothing firm to reconcile. In this site's audit of 9,483 small-business sites, 43.1% contradicted themselves on name, address or phone across their three main pages, 36.9% gave more than one phone number, and 34.1% had no postal address a machine could parse. An Organization or LocalBusiness block stating the identity once, with sameAs pointing at the matching profiles, is how a site settles the question.

Sources

Google, Knowledge Graph Search API. Property definition: schema.org/sameAs. The audit figures are this site’s own; sample drawn per metro and trade, percentages describe the sample.