What is a knowledge graph?
A knowledge graph is a database that stores things (entities) and the relationships between them, rather than pages or keywords. Each entity, such as a person, company, place or film, is a node, and each fact connects it to something else: "the Eiffel Tower" is located in "Paris", which is the capital of "France".
In SEO, "the Knowledge Graph" almost always means Google's, which Google describes as its "database of billions of facts about people, places, and things" [3]. Google uses it to answer factual questions such as "How tall is the Eiffel Tower?" directly in Search [3]. Other companies run their own knowledge graphs, and public ones such as Wikidata exist too, but Google's is the one that shapes search results most visibly.
How does Google's Knowledge Graph work?
Google's Knowledge Graph works by recognizing that a query or page refers to a specific entity, then drawing on stored facts about that entity and the entities connected to it. When Google launched it, Google named three jobs [1]:
- Telling apart things with the same name. The Taj Mahal monument versus the musician Taj Mahal.
- Summarizing the facts people want about an entity, in what became the knowledge panel.
- Helping people discover connected things, such as related people, works or places.
Google says its systems both gather and check Knowledge Graph facts automatically, that it accepts reports of errors from anyone, and that it acts fastest on errors about civic, medical, scientific and historical topics or where there is a risk of serious harm [2].
Where does Google's Knowledge Graph get its information?
Google's Knowledge Graph gets its information from public web sources, licensed data, public databases, structured data on websites and people who have claimed their knowledge panels. Google's 2020 explanation said Wikipedia is commonly cited "but it's not the only one", and listed hundreds of other web sources, licensed data for music, sports and TV, medical providers for health information, and structured data markup from content owners [2]. Google's help center adds public databases and licensed data such as sports scores [3].
At launch in 2012, Google named Freebase, Wikipedia and the CIA World Factbook as sources [1].
How has Google's Knowledge Graph changed since 2012?
Google's Knowledge Graph launched in the US in May 2012 with 500 million objects and grew to over 500 billion facts about five billion entities by 2020 [1] [2]. Along the way its main open data source, Freebase, was shut down, and Google added new layers for topics and new ways for entities to correct their own facts.
2010: Google buys Metaweb
In July 2010, Google acquired Metaweb, the company behind Freebase, an open, shared database of facts launched in 2007 [5]. Freebase later fed the Knowledge Graph [5].
2012: launch, "things, not strings"
Google announced the Knowledge Graph on May 16, 2012, starting with US English users, with more than 500 million objects and 3.5 billion facts and relationships [1]. In December 2012 it expanded to Spanish, French, German, Portuguese, Japanese, Russian and Italian [4]. Within about seven months of launch it had reached 570 million entities and 18 billion facts [4].
2013: Hummingbird
Google announced the Hummingbird algorithm on September 26, 2013, moving query handling toward the meaning of whole queries rather than individual words [6]. Understanding queries as being about things made the Knowledge Graph more central to Search.
2014 to 2016: Freebase closes
On December 16, 2014, Google announced it would close Freebase and help move its data to Wikidata [5]. Freebase.com and its API shut down on May 2, 2016 [5]. By mid-2016, Google reported 70 billion facts in the Knowledge Graph [4].
2018: the Topic Layer
In September 2018, Google added a Topic Layer to the Knowledge Graph, built to understand a topic space, its subtopics and how they relate, so Google could show the most relevant subtopics for a search [7].
2019: social profiles found automatically
In June 2019, Google deprecated social profile markup for knowledge panels and said it would discover social profiles automatically [9].
2020: 500 billion facts
In May 2020, Google said the Knowledge Graph held over 500 billion facts about five billion entities, and that people and organizations could claim their knowledge panel to give feedback directly [2].
What is the difference between the Knowledge Graph and a knowledge panel?
The Knowledge Graph is Google's underlying database of entities and facts, while a knowledge panel is the box in search results that displays some of those facts about one entity. Every knowledge panel draws on the Knowledge Graph, but not every entity in the Knowledge Graph gets a visible knowledge panel.
| Knowledge Graph | Knowledge panel | |
|---|---|---|
| What it is | Database of entities and relationships | Search result feature for one entity |
| Visible to searchers | No, only through features it powers | Yes |
| Can you claim it? | No | Yes, through Google's verification process [2] |
| How to check | Knowledge Graph Search API [8] | Search the entity's name |
This distinction matters for diagnosis. A brand with no knowledge panel may still be in the Knowledge Graph, and a brand with a panel showing wrong facts has a data problem, not a display problem.
How do you check if you're in Google's Knowledge Graph?
To check if an entity is in Google's Knowledge Graph, search its exact name in Google for a knowledge panel, then query the Google Knowledge Graph Search API. The API returns matching entities with an ID such as kg:/m/0dl567 (Taylor Swift's) and a resultScore showing how strong each match is [8].
Read the result with care. Google says the API returns only individual matching entities, not the connected graph, and is not suitable as a production-critical service [8]. The public API is a limited view of what Google holds, so a missing result is a hint that your entity is weak or absent, not proof.
A practical check when reviewing a brand: compare three things side by side. Does the API return the brand? Does a brand search show a knowledge panel? Does the panel, if any, show the right website, logo and profiles? A "yes, no, n/a" pattern means Google knows the entity but does not yet show it. A "no, no" pattern means identity work comes first.
How do you get into Google's Knowledge Graph?
You get into Google's Knowledge Graph by being described consistently on sources Google already trusts, since Google builds the Knowledge Graph from many sources rather than from submissions [2]. There is no form to add an entity. The work that helps:
- Publish a clear entity home, the page on your own site that states who or what the entity is.
- Add Organization or Person structured data with
sameAslinks to your official profiles. Google says Organization markup can help disambiguate an organization in search results [10]. See sameAs markup. - Keep facts identical across your site, profiles and listings.
- Earn independent coverage and listings on established sites. Search Insider's free Foundation Links list covers 300 business directories and profiles.
- Claim the knowledge panel once one appears, so you can suggest corrections [2].
This is the core of entity SEO.
Does the Knowledge Graph affect rankings?
The Knowledge Graph affects how Google understands queries and entities, but Google has not said it is a ranking factor for ordinary web pages and has published no figure on its effect. Its clearest roles are interpreting what a query is about, telling apart entities with the same name, and powering features such as knowledge panels [1].
For a business, the practical effect shows up on the brand search results page: whether Google shows a panel with the right facts, links the right profiles, and avoids confusing the brand with another one. Claims that being in the Knowledge Graph lifts rankings across a site are not backed by anything Google has published.
When should you care about the Knowledge Graph?
Care about the Knowledge Graph when people search for your brand or name by name: a business, a public figure, an author, a product line. If a brand search shows wrong facts, the wrong company or no panel, the fix lies in the sources the Knowledge Graph reads. For pages competing on generic queries, the Knowledge Graph matters far less than whether the page answers the query well.