What is the Google Knowledge Graph?

Google's database of entities and the facts and relationships between them, introduced in May 2012. At launch it held more than 500 million objects and 3.5 billion facts, drawing on sources such as Freebase, Wikipedia and the CIA World Factbook. It powers knowledge panels in Search.

Can I add my business to Google's Knowledge Graph?

Not directly. Google builds it from sources across the web. You can make your entity easier to recognise with consistent facts, Organization structured data, and profiles on sources Google reads. Once a knowledge panel exists, you may be able to claim it and suggest edits.

What is the difference between a knowledge graph and a database?

A regular database stores records in tables designed for one application. A knowledge graph stores entities and typed relationships between them, so new kinds of facts can be added and connected without redesigning the structure.

How do knowledge graphs relate to LLMs?

They complement each other. LLMs are good at language but can invent facts; knowledge graphs hold explicit facts. Some systems ground model answers in graph data, and newer approaches such as Microsoft Research's GraphRAG build graphs from text to improve retrieval.

AI search glossary

What is a knowledge graph? Entities, relationships and AI search

A knowledge graph is a structured database that stores facts as entities (people, companies, places, products, ideas) connected by labelled relationships, so software can look up what something is and how it relates to other things.

Also called: Google Knowledge Graph, knowledge base graph, KG

Why it matters for AI search

Knowledge graphs give search engines and AI systems a fact layer that does not depend on reading prose each time. Google's Knowledge Graph feeds knowledge panels, and open graphs such as Wikidata are reused widely. If your brand is a clear entity with correct facts in these graphs, engines can identify you and describe you consistently.

How a knowledge graph is built

A knowledge graph has two basic parts:

Nodes, the entities: "Douglas Adams", "St John's College", "The Hitchhiker's Guide to the Galaxy".
Edges, the labelled relationships: "educated at", "author of".

Each fact is a small statement: Douglas Adams, educated at, St John's College. Because entities have stable identifiers, the graph knows that "Adams" in one source and "Douglas Noel Adams" in another are the same person. A schema (often schema.org types such as Organization, Person or Product) defines what kinds of entities and relationships are allowed.

Google's Knowledge Graph

Google introduced its Knowledge Graph in 2012 with the phrase "things, not strings". It said the graph would help Search pick the right meaning of an ambiguous query, summarise key facts about an entity, and show related topics. The visible result is the knowledge panel beside search results.

Google offers a read-only Knowledge Graph Search API that returns matching entities in JSON-LD using schema.org types. Google notes it is not suitable as a production-critical service, is migrating to Cloud Enterprise Knowledge Graph, and recommends Wikidata dumps for anyone who needs graph data rather than single entities.

Open and other knowledge graphs

The term predates Google. Wikipedia traces it to 1972, and earlier graphs include WordNet, GeoNames, DBpedia and Freebase. Today the largest open graph is Wikidata, released under CC0 so anyone can reuse it. Microsoft runs its own graph, Satori, for Bing.

Why it matters for AI visibility

When an AI engine is asked "what is [brand]?", it is helped by an unambiguous entity with consistent facts. Knowledge graphs are one of the places those facts live. To be represented well:

Describe your organisation consistently on your site and mark it up with schema markup.
Link your official profiles with sameAs, so engines connect them to one entity.
If you qualify, keep a referenced Wikidata item accurate.
Ask ChatGPT or Gemini who you are and correct the sources behind wrong answers.

Knowledge graphs also help with ambiguity. If your brand shares a name with a town, a film or another company, an entity with a clear type, location and website is what lets an engine tell you apart. This is the core of entity SEO.

Questions

What is the Google Knowledge Graph?: Google's database of entities and the facts and relationships between them, introduced in May 2012. At launch it held more than 500 million objects and 3.5 billion facts, drawing on sources such as Freebase, Wikipedia and the CIA World Factbook. It powers knowledge panels in Search.
Can I add my business to Google's Knowledge Graph?: Not directly. Google builds it from sources across the web. You can make your entity easier to recognise with consistent facts, Organization structured data, and profiles on sources Google reads. Once a knowledge panel exists, you may be able to claim it and suggest edits.
What is the difference between a knowledge graph and a database?: A regular database stores records in tables designed for one application. A knowledge graph stores entities and typed relationships between them, so new kinds of facts can be added and connected without redesigning the structure.
How do knowledge graphs relate to LLMs?: They complement each other. LLMs are good at language but can invent facts; knowledge graphs hold explicit facts. Some systems ground model answers in graph data, and newer approaches such as Microsoft Research's GraphRAG build graphs from text to improve retrieval.