Live session Oct 22: AI Search for Clients
Promptwatch Logo

Knowledge Graphs

Knowledge graphs store facts as entities joined by typed relationships. AI systems use them to tell entities apart, check facts and answer multi-step questions.
Updated October 15, 2026
AI

Definition

A knowledge graph is a way of storing information as entities (nodes) connected by typed relationships (edges), so that a fact like 'Acme Corp was founded by Jane Doe in 2015' is recorded as linked, queryable data rather than a sentence. This entry covers the technology in general. For Google's own system, see Knowledge Graph.

Most knowledge graphs are built from three parts. Entities have stable identifiers, such as a Wikidata Q-ID or a Google machine ID. Relationships and attributes are expressed as subject, predicate and object statements. An ontology defines which types and relationships are allowed, for example that a 'founded by' edge connects an organization to a person. Schema.org, RDF and property graph databases such as Neo4j are common ways to express this.

Public knowledge graphs are large and openly reusable. Wikidata, the structured sibling of Wikipedia, listed about 123.6 million items on its statistics page in October 2026, and its data is reused by search engines, voice assistants and research datasets. Companies also build private graphs that link products, customers, documents and teams for internal search and assistants.

Large language models store knowledge differently. What a model learns in training is spread across its weights as parametric knowledge, with no explicit list of entities, which is one reason models can mix up similarly named companies. Knowledge graphs get added back in at retrieval time. Microsoft Research's GraphRAG, described in an April 2024 paper and open-sourced on GitHub in July 2024, uses an LLM to extract an entity graph from a document collection and then summarizes clusters of related entities. The paper reports better comprehensiveness and diversity than standard RAG on broad questions about a whole corpus. Entity linking is a related step: Google's Cloud Natural Language API, for example, tags entities it finds in text with a Wikipedia URL and Knowledge Graph MID when it can match them.

For AI search, graphs matter in three places:

  • Disambiguation: deciding whether a prompt about 'Jaguar' means the carmaker or the animal before retrieving sources.
  • Fact checking: comparing a generated claim, such as a founding date, against a structured record.
  • Multi-hop questions: answering 'Which CRM tools were founded by former Salesforce employees?' requires following several relationships, which query fan-out and graph traversal handle better than a single keyword search.

The consumer assistants (ChatGPT, Claude, Perplexity, Gemini) do not publish how, or whether, they consult knowledge graphs when answering. What a brand can control is the input: a consistent entity home, accurate entries in public graphs such as Wikidata where the brand qualifies, Schema.org markup that matches visible text, and third-party sources that state the same facts. Those inputs are what entity SEO works on.

Examples of Knowledge Graphs

  • A Wikidata item for a software company records 'instance of: business' (P31), 'founded by' (P112), 'inception' (P571) and 'official website' (P856), giving every tool that reads Wikidata the same founding facts.
  • A legal firm runs GraphRAG over 40,000 case files so associates can ask which judges appear most often alongside a specific expert witness, a question plain vector search answered poorly.
  • An electronics retailer builds an internal product graph linking each SKU to compatible accessories and replacement parts, which its support assistant queries before answering 'Will this charger work with my model?'
  • An SEO analyst runs a product page through Google's Natural Language API and finds the brand name tagged as a generic 'OTHER' entity with no Knowledge Graph MID, a sign Google has not linked the name to the company yet.

Terms related to Knowledge Graphs

Knowledge Graph

The Knowledge Graph is Google's database of entities and the facts linking them. It fills knowledge panels and helps Google resolve which entity a query means.

SEO

Entity SEO

Entity SEO makes a brand, its people and products identifiable as distinct entities to search engines and AI systems, with consistent facts and clear topics.

SEO

RAG (Retrieval-Augmented Generation)

RAG is how AI answer engines like ChatGPT search, Perplexity and AI Mode fetch web pages at question time and cite them, deciding which brands appear.

AI

Parametric Knowledge

Parametric knowledge is what a language model learned in training and stores in its weights. It answers without searching, but cannot cite or update itself.

AI

AI Grounding

AI grounding ties a model's answer to verifiable sources, such as live search results or supplied documents, so each claim can be traced and cited.

AI

Semantic Search

Semantic search returns results that match the meaning and intent of a query, not just its words, using language models, embeddings and entity understanding.

SEO

Query Fan-Out

Query fan-out is how AI search splits one question into several sub-queries, runs them at once and merges the results, so pages compete on each sub-query.

AI

Entity Home

An entity home is the one page a brand or person controls that states the canonical facts about them, which every other profile and markup points back to.

SEO

Vector Search

Vector search finds content by comparing embedding vectors, returning items closest in meaning to a query. It is the semantic half of AI search retrieval.

AI

Frequently Asked Questions about Knowledge Graphs

Learn about AI visibility monitoring and how Promptwatch helps your brand succeed in AI search.

A relational database stores records in tables with fixed columns, and relationships are reconstructed through joins. A knowledge graph stores relationships as first-class data: every fact is an entity linked to another entity or value by a named relationship. That makes it easier to add new kinds of facts, follow chains of relationships, and merge data from different sources about the same entity.

Be the brand AI recommends

Monitor your brand's visibility across ChatGPT, Claude, Perplexity, and Gemini. Get actionable insights and create content that gets cited by AI search engines.

Promptwatch Dashboard