Definition: What is the Knowledge Graph?
The Google Knowledge Graph (KG) is a structured knowledge base that stores entities as nodes and their relationships as edges. It was introduced in 2012 under the slogan “Things, not strings” and marked Google's transition from purely textual search to semantic search. According to Google's last official figure (May 2020), the KG contains over 500 billion facts on more than 5 billion entities; Google has not published newer figures. It covers people, organizations, places, products, events and concepts. Every entity carries an internal identifier (Machine-ID or kgmid), properties and typed relations.
The KG is not the same as a Knowledge Panel: the graph is the database, the panel is the visual SERP rendering. Every entity in the panel originates from the KG, but not every KG entity gets a panel. The KG is also a data source for direct answer boxes and semantic query interpretation. How much AI Overviews draw on the KG is not documented by Google; officially, they rely via RAG on the search index and the core ranking systems (Google's guide). The KG influences which entities are recognized as relevant for a query.
No entity in the KG, no AI visibility
Generative systems benefit from unambiguously anchored entities. Without KG anchoring a brand remains a text fragment that cannot be reliably identified at inference time. Entity engineering is therefore the foundation of any GEO strategy.
Data sources of the Knowledge Graph
The KG draws on several input streams. At launch in 2012 Google named, among others, Freebase, Wikipedia and the CIA World Factbook; in 2020 Google described content from the open web, open and licensed databases, licensed data for music, sports and TV, and health information vetted with medical professionals. One of the most important open sources is Wikidata, the Wikimedia Foundation's knowledge base with currently over 120 million items (Wikidata statistics). How Google weights individual sources is not disclosed; internally the KG uses its own identifiers.
On the web side, structured data from Schema.org markup, consistent sameAs references, official brand homepages and co-occurrence patterns in authoritative trade portals are added. Google merges these sources internally; details of that processing are not public. The KG is therefore not a static dataset but a continuously updated graph.
Google KG vs. Wikidata - the difference
Wikidata is open, collaboratively editable and fully publicly queryable (SPARQL, REST API). The Google Knowledge Graph is proprietary, only partially public (via the Knowledge Graph Search API) and algorithmically curated. Wikidata holds facts that have not yet propagated into Google. Conversely, the Google KG contains internal facts that never reach Wikidata - for example from licensed third-party databases.
Operationally that means: Wikidata is the controllable lever. In our experience, a structured, well-referenced Wikidata entity increases the probability of being absorbed into the Google KG. Wikidata is not a guarantee, but it is the most reliable entry point. For companies without a Wikidata entry, that is the first step of any entity strategy - before fine-tuning schema markup and before co-occurrence campaigns.
Operational path into the Knowledge Graph
The standard path from enterprise advisory practice:
- Schema.org Organization markup on the homepage with a stable
@id, logo, founder, contact details and asameAsarray. - Wikidata entry with at least three independent secondary sources (press articles, commercial registry, trade literature). Secure the Q-ID.
- Build sameAs density: LinkedIn company, Crunchbase, official social profiles, GitHub org (if available). All profiles link back to the main domain.
- Authoritative co-occurrence in trade portals (industry studies, guest contributions, interview formats). Target: tier-1 media with a domain rating > 70.
- Monitoring via the Knowledge Graph Search API and branded SERP analysis. If a Knowledge Panel appears, the entity is stable in the KG.
Typical timeline in our projects: 3-9 months from project start to a first stable Knowledge Panel. Inclusion cannot be guaranteed, and promises of fixed dates are not credible.
Typical mistakes in KG strategies
- Wikidata entry without references. Deleted within weeks. Secure at least three independent editorial sources beforehand.
- Inconsistent name forms. "Example GmbH" vs. "Example" vs. "Example GmbH & Co. KG" across different platforms - weakens MSA.
- sameAs without a return reference. The website links to LinkedIn but LinkedIn does not link back. In our experience bidirectional references are the more robust signal (a hypothesis, not documented by Google).
- Schema markup without
@id. Every page declares a newOrganization. Consequence: fragmentation, no consolidated entity. - Single-point publishing. Own content without external validation. Without co-occurrence in authoritative sources the entity stays weak.
Related terms
The Knowledge Graph is tied to Entity (Schema.org), Knowledge Panel, Wikidata, sameAs and Schema.org. For AI visibility, clean KG anchoring is a foundation for GEO strategies and for unambiguous recognition in AI Overviews. See also E-E-A-T for the trust dimension.
FAQ on the Knowledge Graph
What is the Google Knowledge Graph? ▾
The Google Knowledge Graph is a structured knowledge base that captures entities (people, places, organizations, concepts) along with their properties and relationships. Launched in 2012, it spanned over 500 billion facts about more than 5 billion entities according to Google's last official figure (2020). It underpins Knowledge Panels and direct answers.
Which sources feed the Knowledge Graph? ▾
Google names content from the open web, open and licensed databases and vetted expert content, for example for health. Among the best-known open sources are Wikipedia and Wikidata; at launch in 2012 Google also named Freebase and the CIA World Factbook. How individual sources are weighted is not disclosed.
How do I get my brand into the Knowledge Graph? ▾
Via a Wikidata entry, Schema.org Organization markup with sameAs, authoritative co-occurrence in trade portals and, where possible, a Wikipedia article. In our projects the process typically took 3-9 months from the start of entity anchoring; inclusion cannot be guaranteed.
Is the KG different from a Knowledge Panel? ▾
Yes. The Knowledge Graph is the database. The Knowledge Panel is the visual SERP rendering of a KG entity on a brand search. An entity can exist in the KG without a Knowledge Panel being shown - that depends on the relevance score.
Can I query the Knowledge Graph directly? ▾
Partly. Google offers the Knowledge Graph Search API, which returns entity stubs with name, type, description and detailedDescription. The full internal KG structure is not public. For structured queries, Wikidata with SPARQL is the appropriate alternative.