For our AI SEO work, the smallest irreducible thing a client wants to be visible for, treated as a cluster of the many query and prompt variants that stem from it.
Entity, for our AI SEO work, is the single most nuclear and irreducible version of something a client wants to be visible for: a product, a service, a platform, a process. Many different queries and prompts stem from one entity, so we treat the entity as a quantized cluster of a wider set of query and prompt variants. Anchoring to the entity rather than to any one phrasing absorbs part of the stochastic behaviour of language models, where the same intent surfaces through countless wordings.
This is a working definition, not the entity of named-entity recognition or a knowledge graph. Those identify and link real-world things (people, places, organisations) as nodes with attributes and relationships. Our sense is narrower and purpose-built: the smallest unit of visibility a client cares about, from which the query fan-out of prompts expands, and against which we measure AI Visibility.
Entity disambiguation is the step of resolving an ambiguous name to the one entity it refers to: "Apple" the company against the fruit, "Jaguar" the car against the animal, one person against a namesake. Systems settle it from context, from structured data, and from a knowledge graph that records which entity carries which attributes. For a brand it is the difference between being recognised as a distinct, resolved entity and being blurred into a common word or a competitor, so consistent naming, clear context, and corroborating references across the web are what make a brand easy to disambiguate.
