Definition
The Knowledge Graph is Google's database of real-world entities (people, organizations, places, products, creative works and concepts) and the facts that connect them. Google uses it to work out which 'thing' a search refers to, to populate knowledge panels, and to answer factual questions directly. This entry covers Google's system specifically; knowledge graphs explains the general technology.
Google launched the Knowledge Graph on May 16, 2012, under the line 'things, not strings'. Amit Singhal's launch post described more than 500 million objects and 3.5 billion facts, built from sources such as Freebase, Wikipedia and the CIA World Factbook. In a May 2020 post, Danny Sullivan put it at over 500 billion facts about five billion entities. Google has not published a newer figure, so larger counts quoted online come from third-party tracking. One of those trackers, Kalicube, reported that the graph lost more than 3 billion entities (6.26%) in two updates on June 13 and June 20, 2025, mostly expired events and vaguely typed 'thing' entries.
Each entity gets a machine ID (MID) and one or more Schema.org types, plus attributes and relationships such as founder, headquarters or parent company. Google says facts come from hundreds of web sources, licensed data for areas like music, sports and TV, vetted medical content, and markup that site owners add. Owners who claim a knowledge panel can also suggest corrections. Developers can look up entities through the Knowledge Graph Search API, which returns JSON-LD with the MID and a resultScore. Google is moving that API to Cloud Enterprise Knowledge Graph.
Google's AI Overviews and AI Mode retrieve pages from the Search index and generate answers from them, and Google describes both as rooted in its core ranking and quality systems. Google has not documented exactly how the Knowledge Graph feeds those answers. The practical role is entity resolution: before Google can retrieve pages about 'Mercury', it has to decide whether the prompt means the planet, the element or a company called Mercury. A brand that Google cannot resolve cleanly is more likely to be confused with a namesake, described with stale facts, or left out of AI Overviews and AI Mode answers.
To strengthen your presence:
- Publish one clear entity home (usually the About page) that states the name, category, founders, location and founding date in plain text.
- Add Organization or Person schema markup to that page, with sameAs links to official profiles and, if one exists, the Wikidata item.
- Keep the same facts on LinkedIn, Crunchbase, Wikidata, Google Business Profile and other listings so independent sources agree.
- Claim the knowledge panel once it appears, and use its feedback tool when a fact is wrong.
Wikipedia and Wikidata entries help when the entity is notable enough to meet those communities' rules. Creating promotional entries that get deleted does more harm than good. Entity SEO covers the wider practice of making an organization machine-readable.
Examples of Knowledge Graph
- A software company queries the Knowledge Graph Search API for its brand name and gets two results: its own entity and a 2014 film with the same title. It adds a disambiguating description and sameAs links to its About page so the two stop being merged.
- Eight months after a rebrand, a retailer's knowledge panel still shows the old name. The team updates the About page, Wikidata label, LinkedIn page and Crunchbase profile in the same week, then flags the old name through the claimed panel's feedback form.
- A museum sees AI Mode give the wrong opening year. The error traces back to its Wikidata item, which an editor corrects with a citation to the museum's own annual report.
- Before pitching a knowledge panel project, an agency checks whether the client's founder already has a Knowledge Graph MID. She does not, so the plan starts with an author page and press coverage rather than markup.
