A knowledge graph stores facts as connected entities: a company, its founders, its products, its headquarters, and how each relates to the others. Google introduced its Knowledge Graph in 2012 and uses it to show knowledge panels, answer factual questions and interpret what a search is about. Other search engines and AI companies maintain their own graphs, and public graphs such as Wikidata feed many of them.
How a knowledge graph works
Each entity has an identifier and a set of attributes. Relationships link entities, for example “Quillbrook” is a “software application” made by “Quillbrook Inc.”, which is “headquartered in” a city. Facts are gathered from many sources, including licensed data, public databases, structured data on websites and the web pages themselves.
Because the graph is about things rather than pages, it lets a search engine answer questions where the answer is a fact rather than a document.
Knowledge panels
The most visible output of Google's Knowledge Graph is the knowledge panel, the box that appears for searches about a well-known entity. If your company has one, you can claim it through Google's verification process and suggest corrections. Google decides whether a panel appears; you cannot create one directly.
Why it matters for AI search
AI systems need to know which thing a name refers to and what is true about it. Knowledge graphs give search systems a structured way to do that, and AI features built on search can draw on them. Language models also learn entity relationships from text. The exact way any AI product uses a knowledge graph is not published, but consistent, well-sourced facts about your company across the web help every system that builds one.
Feeding the graph
- Organization and Product structured data on your site, consistent with visible content.
- sameAs links to your official profiles.
- Accurate listings on major business, review and marketplace sites.
- Coverage in reputable publications that state facts about your company.
- A Wikidata item, if your company meets its notability criteria.
Common mistakes
Editing public graphs promotionally
Wikipedia and Wikidata have conflict-of-interest rules. Promotional edits are often reverted and can hurt trust.
Expecting a panel on demand
Google decides whether an entity gets a panel. Focus on consistent facts rather than the panel itself.
Questions
Not directly. Google builds it from many sources. You can make facts clear and consistent, and claim a panel once one exists.
No. Structured data is markup on a page. A knowledge graph is the database a search engine builds, partly using that markup.
