Knowledge graph

A knowledge graph is a database that stores real-world entities, such as people, places, companies and products, with facts about each one and the relationships between them. Google's Knowledge Graph fills the knowledge panels in Google Search and is one of the sources Google AI Mode can draw on to answer a prompt.

Why the knowledge graph matters for AI answers

When Google launched its Knowledge Graph in 2012, it held "more than 500 million objects" and more than 3.5 billion facts about them and their relationships. By 2020, Google said the graph had "amassed over 500 billion facts about five billion entities." It is the store of settled facts Google keeps about the world.

That store now feeds AI answers. Introducing AI Mode in March 2025, Google wrote that it can "tap into fresh, real-time sources like the Knowledge Graph" alongside web pages. For a brand, being in the graph means Google holds structured facts about what you are, what you make and which site is yours, not just pages that happen to mention you.

How a knowledge graph works

A knowledge graph stores "things, not strings," to borrow Google's launch phrase. Three parts do the work:

Entities. Each node is one real-world entity with its own ID: a person, an organization, a place, a product. Google's API lists types such as Organization, LocalBusiness, Person and WebSite, described with schema.org vocabulary.
Facts and relationships. The edges link entities: a company was founded by a person, is based in a city, makes a product.
Sources. Google says the facts come "from materials shared across the web, as well as from open source and licensed databases." Wikipedia is one source among hundreds.

Knowledge panels are the visible part: the information boxes Google shows when you search for an entity that is in the graph. They are generated automatically, and the subject of a panel, such as a company, can claim it and suggest changes.

Example: one name, several entities

Google's 2012 launch post used the search [taj mahal], which could mean the monument, a Grammy Award-winning musician, a casino in Atlantic City or the nearest Indian restaurant. An AI assistant meets the same problem with a prompt like "What is Taj Mahal's best album?": the word "album" points to the musician, and the engine has to settle on that entity before it can answer.

The Knowledge Graph Search API shows what a stored entity looks like. In Google's documentation, a request for "taylor swift" returns the ID kg:/m/0dl567, the types Thing and Person, the description "Singer-songwriter", a summary from Wikipedia, the official website and a result score.

How to check and improve your knowledge graph presence

Search your brand name on Google. A knowledge panel shows Google treats you as an entity. Check every fact, then claim the panel to suggest corrections. Local businesses manage theirs through a Business Profile.
Query the Knowledge Graph Search API. It tells you whether an entity matches your name, with which type and description. Google says the API is not suited to production use, so treat it as a spot check.
Keep your facts identical everywhere. Use the same name, category, founders and links on your site, your social profiles and the directories and articles that cover you.

Then check whether recognition turns into recommendations. Hawkscan tracks whether ChatGPT, Gemini, Microsoft Copilot and Google AI Mode mention your brand for the prompts your buyers ask, every day, with your mention rate and average position.