Knowledge Graph

A knowledge graph is a network of entities (people, organisations, places, products, concepts) connected by labelled relationships. Google launched its Knowledge Graph in May 2012 to move search “from strings to things”, and it now underpins knowledge panels, direct answers and much of how Google and AI systems understand what a query or page is about.

How it works

A graph has four building blocks. Nodes store details about entities, with labels for their type and properties for their attributes. Relationships connect two nodes, for example Apple FOUNDED_BY Steve Jobs, and are what let a graph answer “who founded Apple?” without a sentence containing both terms. Schemas set which nodes and relationships are valid, and an ontology formally specifies the concepts and relationships in a domain. Facts are often stored as tuples: subject, predicate, object.

At launch, Google said its graph held more than 500 million objects and more than 3.5 billion facts and relationships. According to Google patents covered on MLforSEO, the graph is kept up to date through reconciliation: new facts are only added when several independent, authoritative sources agree, and Google can generate its own questions to fill gaps it detects from user queries.

How it’s used in SEO and marketing

• Checking whether your brand, people and products are recognised as distinct entities, using the Knowledge Graph Search API.
• Disambiguation when a brand shares a name with something else.
• Covering the factual attributes users search for (founder, headquarters, price) clearly enough for extraction.
• Grouping keywords by shared entities and relationships, and planning internal links around them.
• Keeping entity descriptions consistent across your site, Wikidata, LinkedIn and directories, since panels draw on multiple sources.

Example

A search for “height of Empire State Building” is an entity-attribute question: the entity is the Empire State Building and the attribute is height. Google can answer it straight from the graph and suggest related facets such as floors, year built or architect, which is also a useful map of the subtopics a page about that entity should cover.

Limitations and things to watch

• Google’s graph is not public; the Search API shows a limited view, and its result score is not an explained ranking metric.
• Wikidata and other open graphs feed into the ecosystem but are not the same as Google’s graph.
• Entities can lose knowledge panels when Google prunes thin or contradictory facts, so recognition needs maintenance.
• Structured data helps disambiguation but cannot create notability you have not earned.

Learn it on MLforSEO

• What Is a Knowledge Graph? How to Use It for Smarter Keyword Strategy
• Entity-Based Search in the LLM Era: How Google Structures and Interprets Entities
• What It Takes to Earn a Google Knowledge Panel
• Knowledge Graph API Extraction and Entity Relationship Analysis (Notebook)
• AI Search & LLMs: Entity SEO and Knowledge Graph Strategies for Brands (course)