A knowledge graph stores information as a network of entities connected by explicit, labelled relationships: people, places, products, and concepts as nodes, with edges describing how they relate — who founded what, which part belongs to which product, how one event led to another. It captures structured, factual connections that can be traversed and queried.
This makes a knowledge graph complementary to a vector database. Where a vector database stores implicit relationships learned as geometry — items that are similar end up close together — a knowledge graph stores explicit relationships defined as facts. One excels at fuzzy semantic similarity, the other at precise, structured reasoning over known connections.
The two are increasingly combined, as in GraphRAG, where vector search finds relevant entry points and graph traversal expands outward along relationships to gather connected facts. This pairing gives an AI system both the recall of semantic search and the structured, multi-hop reasoning of a graph, which is valuable for complex questions that depend on how facts link together rather than just on topical similarity.