Knowledge graph — a database of things and the relationships between them, rather than a database of documents.
Things, not pages
A search index stores pages and what words are on them. A knowledge graph stores entities — a business, a person, a place, a category — and the facts that connect them. Google’s Knowledge Graph is one well-known instance. It is not the only one; the major model providers all maintain something that plays the same role.
The difference matters because a recommendation is a statement about a thing, not about a document. “The best HVAC company in Naperville” is a claim about an entity. A system that has not resolved you into an entity has nothing to make that claim about.
What this means for a local business
If you exist in the graph as a resolved entity — with a category, a location, and relationships to places and people — you are available to be recommended. If you exist only as a set of pages, you are available to be ranked. Those are different outcomes and they are won in different ways.
Related: entity resolution, structured data, retrieval-augmented generation.