In search, an entity is a single, distinct thing, such as a person, place, organization or object, that a search engine can recognize as itself whatever words are used to name it. Google summed up the shift in 2012 as moving from strings (the letters someone types) to things (what those letters refer to).
How entities work
Google introduced its Knowledge Graph on 16 May 2012 as a model of real-world entities and the relationships between them. Before it, a query was a set of keywords to match. After it, Google could work out which thing a query meant and connect it to related things: the launch post linked Marie Curie to her husband, Pierre, and to her two children, one of whom also won a Nobel Prize.
By May 2020 Google said the Knowledge Graph held over 500 billion facts about five billion entities, which it describes as people, places and things. The facts come from material across the web, open-source and licensed databases, and markup that site owners add. Wikipedia is a common source but not the only one.
Google's Knowledge Graph Search API shows how an entity is recorded: the documentation's sample record for Taylor Swift carries the ID kg:/m/0dl567 and the schema.org types Thing and Person.
Why entities matter for AI answers
Ask an AI engine which tools, brands or products to consider and the answer is a list of entities. Google says Google AI Mode can tap the Knowledge Graph alongside web pages. Microsoft's February 2026 advice for publishers cited in Copilot and Bing's AI answers includes aligning text, images and video so they consistently represent the same entities, products or concepts.
Names drift in practice. Our scan records of AI answers about AI visibility software hold at least five spellings of one vendor, Otterly.AI: "Otterly.AI", "Otterly.ai", "Otterly AI", "OtterlyAI" and plain "Otterly". Counted by exact string, one company would look like five. That is the problem entity recognition solves, and it is why a tracker that reports AI share of voice has to fold name variants together before counting. If an engine attaches the wrong facts to your name, see how to correct wrong information about your brand in AI answers.
An example: Taj Mahal
Google's 2012 announcement used the query [taj mahal]. To a keyword engine it is two words. To a person it could mean the monument, the Grammy Award-winning musician or a casino in Atlantic City. Treating each as a separate entity let Google narrow the results to the one the searcher meant.
Common misconceptions
- "An entity is a keyword." A keyword is text. An entity is the thing the text points to, and several spellings can point to the same one.
- "Only a Wikipedia page makes you an entity." Google says Wikipedia is a commonly cited source but that it draws on hundreds of sources, including licensed data and site markup.
- "Entities are a Google-only idea." Schema.org, founded by Google, Microsoft, Yahoo and Yandex, is built to describe entities and the relationships between them, and Microsoft's AI citation guidance uses the same word.