Media Morpheus

Definitionby Media Morpheus Research

October 10, 2026 · Last fact-checked: October 10, 2026

What Is a Knowledge Graph? Definition, How Google's Works and What It Is Not

A knowledge graph is a database that stores facts as connections between entities (people, organizations, places and things), so that software can answer questions about the things themselves rather than matching words on a page.

The term is general, but in search it usually means Google's Knowledge Graph, launched in May 2012 with the slogan “things, not strings.” Google said in 2020 that it held “over 500 billion facts about five billion entities.” It matters to companies because knowledge panels and many direct answers draw on it. The common misconception is that the knowledge graph and the knowledge panel are the same thing: the panel is one display of the graph, and there is no form to submit an entity to the graph itself.

Knowledge GraphEntitiesGoogle SearchWikidata

At a Glance

  • Google, 2012: the Knowledge Graph “currently contains more than 500 million objects, as well as more than 3.5 billion facts about and relationships between these different objects” (Google).
  • Google, 2020: it “has amassed over 500 billion facts about five billion entities”, drawing on “materials shared across the web, as well as from open source and licensed databases” (Google).
  • Knowledge panels are “information boxes that appear on Google when you search for entities (people, places, organizations, things) that are in the Knowledge Graph” and are “automatically generated” (Knowledge Panel Help).
  • The general model behind most knowledge graphs is the W3C's RDF: “sets of subject-predicate-object triples” (W3C).
  • Wikidata, an open knowledge graph Google says is not its only source, “currently contains 123,583,264 items” (Wikidata, read October 10, 2026).

What Is a Knowledge Graph, in Its Defining Sources' Words?

No single standard defines “knowledge graph”, so we quote the main sources. Google introduced its Knowledge Graph on May 16, 2012 as a way to search for “things, not strings”, and said it “currently contains more than 500 million objects, as well as more than 3.5 billion facts about and relationships between these different objects” (Google). In 2020 Danny Sullivan, Google's public liaison for Search, called it “a giant virtual encyclopedia of facts” and “a system that understands facts and information about entities from materials shared across the web, as well as from open source and licensed databases” (Google).

The underlying data model is older. The W3C's RDF 1.1 specification describes graphs as “sets of subject-predicate-object triples”, where “An RDF graph can be visualized as a node and directed-arc diagram” (W3C). A fact such as “Acme Ltd, headquartered in, Leeds” is one triple: two entities joined by a named relationship. Google does not say its internal graph uses RDF, but its public Knowledge Graph Search API “uses standard schema.org types and is compliant with the JSON-LD specification” (Google for Developers).

Knowledge Graph at a Glance

QuestionAnswerSource
What it isA database of entities and the relationships between themGoogle; W3C
Best-known exampleGoogle's Knowledge Graph, launched May 2012Google
Size (Google)Over 500 billion facts about five billion entities (2020)Google
Where its data comes fromWeb content, open source and licensed databases, data partners, verified entitiesGoogle; Knowledge Panel Help
What people seeKnowledge panels and other facts in resultsKnowledge Panel Help
Open alternativeWikidata, a free, collaborative, multilingual knowledge baseWikidata
Can you submit to it?No public submission form; verified entities can suggest edits to their panelsKnowledge Panel Help
Entities in Google's Knowledge Graph, as reported by Google, with Wikidata for scale

Entities or items, millions

Google Knowledge Graph, May 2012
500M+
Google Knowledge Graph, May 2020
5,000M
Wikidata items, October 10, 2026
123.6M

Source: Google, “Introducing the Knowledge Graph: things, not strings” (May 16, 2012); Google, “A reintroduction to our Knowledge Graph and knowledge panels” (May 20, 2020); Wikidata:Statistics, read October 10, 2026.

Google's figures are its own round numbers (“more than”, “over”) and Google has not published a newer count we could find. Wikidata items and Google entities are counted differently, so the Wikidata bar is a scale reference, not a like-for-like comparison.

Where Google's Knowledge Graph Comes From

Google's 2012 launch post said the graph was “rooted in public sources such as Freebase, Wikipedia and the CIA World Factbook” and “augmented at a much larger scale”, and that it was “tuned based on what people search for, and what we find out on the web” (Google). Freebase was a community-curated knowledge base managed by Google. Wikidata's project page records that on December 16, 2014 Freebase announced it was closing and wished to give its resources to Wikidata, and that it became read-only on March 31, 2015 (Wikidata).

By 2020 Google described a much wider set of inputs: “Wikipedia is a commonly-cited source, but it's not the only one. We draw from hundreds of sources from across the web, including licensing data that appears in knowledge panels for music, sports and TV.” It also cited medical providers for health panels and “special coding that content owners can use”, meaning structured data (Google). Our explainer on how an entity enters the Knowledge Graph follows those routes in detail, and does Gemini build the Knowledge Graph? covers what Google has said about AI models and the graph.

Facts in Google's Knowledge Graph, 2012 and 2020

Facts and relationships, billions, as stated by Google

May 2012
3.5B+
May 2020
500B+

Source: Google, May 16, 2012 and May 20, 2020 blog posts.

The 2012 figure is “facts about and relationships between” objects; the 2020 figure is “facts about” entities. Google does not publish how it counts either.

How a Knowledge Graph Works

A knowledge graph works by giving each entity a stable identifier and attaching typed facts to it. Three parts matter:

  • Entities. Each person, company or place gets an ID, so “Taj Mahal the monument” and “Taj Mahal the musician”, Google's 2012 example, are separate nodes.
  • Relationships. Named links such as founder, headquarters or parent organization connect entities. In RDF terms these are the predicates of triples.
  • Sources and reconciliation. Facts from different sources are matched to the same entity. Identity links such as the schema.org sameAs property help with this; see what the sameAs property is.

Google listed three early uses in 2012: finding the right thing when a query is ambiguous, summarizing key facts, and helping people explore related topics. Its Knowledge Graph Search API lists comparable uses for developers, such as “Getting a ranked list of the most notable entities that match certain criteria”, and warns: “This API is not suitable for use as a production-critical service” (Google for Developers). The same page says the API is being migrated to Cloud Enterprise Knowledge Graph. The broader practice of making a company legible to these systems is often called entity SEO.

Knowledge Graph vs Knowledge Panel

The graph is the database; the panel is one view of it. Knowledge Panel Help says knowledge panels appear “when you search for entities (people, places, organizations, things) that are in the Knowledge Graph”, that they “are automatically generated”, and that information “comes from various sources across the web” (Knowledge Panel Help). The same page says entities in panels are treated as “self-authoritative”, so “some of the information displayed may also come from verified entities who have suggested edits to facts on their own knowledge panels.”

Google also notes that on mobile, information from the graph “is often not presented through a single knowledge panel”; instead “one or more knowledge panels may appear interspersed among the overall results” (Google). An entity can therefore be in the graph without a visible panel for every query. Our guides on the Google knowledge panel and claiming a knowledge panel cover the panel side.

Knowledge Graph vs Wikidata

Wikidata is a public knowledge graph; Google's is private. Wikidata describes itself as “a free, collaborative, multilingual, secondary knowledge base, collecting structured data to provide support for Wikipedia, Wikimedia Commons, the other wikis of the Wikimedia movement, and to anyone in the world”, with data published under a public-domain dedication (Wikidata). Anyone can read and edit it, subject to its notability and sourcing rules. Google's graph cannot be edited directly, and Google does not publish which Wikidata items it uses. See how to create a Wikidata item for your company and is Wikidata easier than Wikipedia?

What a Knowledge Graph Means for a Company

For a company, the practical question is whether search systems recognize it as a distinct entity with correct facts. Google does not publish entry criteria, so the honest list is about consistency and evidence, not guarantees:

  • Describe yourself consistently with structured data on your own site; see how to add Organization schema markup. Google names “special coding that content owners can use” as one input.
  • Be described by independent sources. Google says Wikipedia is commonly cited and that it draws on hundreds of sources. Whether coverage is necessary is discussed in does a knowledge panel require media coverage?
  • Fix errors through the channels Google offers: feedback links on panels, and verification for entities that can claim their panel. Local businesses use a separate claiming process through what GoogleLocal businesses use a separate process through their Business Profile, according to Google's 2020 post.#x27;s 2020 post calls Google My Business, now Google Business Profile.

Knowledge Graph vs Related Terms

TermWhat it isRelationship to a knowledge graph
Knowledge panelAn information box in Google resultsA display of Knowledge Graph facts
WikidataAn open, editable knowledge baseA public knowledge graph; one of many possible sources
Schema.orgA shared vocabulary for structured dataSupplies types and properties; Google's API uses its types
EntityA uniquely identified thingThe node a knowledge graph stores facts about
Featured snippetAn excerpt from a web page shown firstDrawn from a page, not from the graph; see what a featured snippet is
Search indexGoogle's store of crawled pagesHolds documents; the graph holds facts about things

Common Misconceptions About Knowledge Graphs

MythWhat the source actually saysSource
The knowledge graph and the knowledge panel are the samePanels appear for entities “that are in the Knowledge Graph”; the panel is a displayKnowledge Panel Help
Google's graph is just Wikipedia“Wikipedia is a commonly-cited source, but it's not the only one”Google
You can submit your company to the Knowledge GraphPanels are “automatically generated”; verified entities can suggest editsKnowledge Panel Help
The Knowledge Graph API shows how Google ranks entitiesGoogle calls it read-only and “not suitable for use as a production-critical service”Google for Developers
The graph is always correctGoogle: “Inaccuracies in the Knowledge Graph can occasionally happen”Google

Frequently Asked Questions

What is a knowledge graph in simple terms?

It is a database of things (people, companies, places) and the facts that link them, such as who founded a company or where it is based. Software uses it to answer questions about things rather than matching words.

How big is Google's Knowledge Graph?

Google said in May 2020 that it had over 500 billion facts about five billion entities. In 2012 it reported more than 500 million objects and 3.5 billion facts. We found no newer official figure.

Where does Google get Knowledge Graph data?

Google says it draws on materials across the web, open source and licensed databases, data partners for topics such as music and sport, and suggestions from verified entities. Wikipedia is commonly cited but is not the only source.

Is a knowledge panel the same as the Knowledge Graph?

No. A knowledge panel is an information box Google shows for entities that are in the Knowledge Graph. The graph is the underlying database.

Can I add my company to the Knowledge Graph?

There is no public submission form. Google says panels are generated automatically. Consistent structured data, independent sources and, where eligible, panel verification are the documented ways to influence the facts shown.

Is Wikidata a knowledge graph?

Yes. Wikidata is a free, collaborative, multilingual knowledge base that anyone can read and edit. It is separate from Google's Knowledge Graph, though Freebase content was offered to Wikidata when Google closed Freebase.

Sources and Methodology

This explainer is built from Google's 2012 and 2020 blog posts on the Knowledge Graph, Knowledge Panel Help on knowledge panels, the Knowledge Graph Search API page, the W3C RDF 1.1 Concepts specification and three Wikidata project pages, all read on October 10, 2026. Definitions are quoted verbatim. Google's size figures are its own rounded statements and have not been independently verified; we found no newer official count. The Wikidata item count is the live figure on the day we read it. We did not query the Knowledge Graph Search API or test how any entity enters the graph.

  1. official Google, Amit Singhal, Introducing the Knowledge Graph: things, not strings, May 16, 2012. Read October 10, 2026.
  2. official Google, Danny Sullivan, A reintroduction to our Knowledge Graph and knowledge panels, May 20, 2020. Read October 10, 2026.
  3. official Knowledge Panel Help, About knowledge panels. Read October 10, 2026.
  4. official Google for Developers, Google Knowledge Graph Search API, last updated April 26, 2024. Read October 10, 2026.
  5. standards W3C, RDF 1.1 Concepts and Abstract Syntax, W3C Recommendation. Read October 10, 2026.
  6. official Wikidata, Wikidata:Introduction. Read October 10, 2026.
  7. official Wikidata, Wikidata:Statistics, item count as displayed on October 10, 2026. Read October 10, 2026.
  8. official Wikidata, WikiProject Freebase, Freebase closure timeline. Read October 10, 2026.