A knowledge graph is a network representation of facts: entities such as people, products, or companies become nodes, and the relationships between them become edges. “Company A supplies component B” or “drug X interacts with drug Y” each become a stored, queryable connection. Storing knowledge this way lets systems answer questions that flat tables make painful, such as multi-hop relationships and hidden links.
Search engines popularized the idea, and enterprises now use knowledge graphs for master data, fraud detection, recommendations, and grounding AI systems. Paired with a language model, a graph supplies verified relationships the model can cite instead of guessing, which reduces hallucinated answers.
Building one is mostly a data exercise: entities must be resolved across systems and relationships kept current. A graph built on stale or inconsistent records repeats those flaws in every answer it grounds.
Related terms: Retrieval-Augmented Generation (RAG) · Ground Truth · Master Data Management · Semantic Layer