What is an entity in search?

An entity is a single real-world thing that a system can tell apart from everything else. Companies, people, products, places and organizations are the common types. The term named entity comes from language processing, where software picks the names of people, organizations and places out of plain text.

The key difference is between a string and a thing. A keyword is a string of letters. An entity is what that string refers to. "Mercury" can mean a planet, a metal or a company. Each is a separate entity with separate facts.

Search entities are the things a search engine has catalogued, usually in a knowledge graph that links each entity to its attributes and to other entities.

An AI engine can learn about your entity through two channels.

• Parametric knowledge. The model absorbed facts during training. That knowledge is frozen at a cutoff date, so nothing you publish this quarter changes it.
• Retrieval. The engine fetches pages when it answers a question and reads the facts on them. This is the channel your website can affect.

Public knowledge bases also store entities. Wikidata, for example, records each entity as an item with a stable ID and a list of statements about it. Whether any given engine reads that entry for a given answer is not documented.

AI platforms get specific facts wrong about specific brands: an old price, a renamed product, the wrong headquarters. One cause sits entirely on your side. Your own site may state the same fact in different ways.

Say you run a hypothetical project-management app. The pricing page says $49 a month. A 2024 blog post still says $39. The About page says you were founded in 2019, the footer says 2018. An engine retrieving those pages has two answers for each question.

What to check first:

1. List the facts a buyer asks about: pricing tiers, product names, founding year, location, integrations.
2. Find every page that states each one, including old posts and templates.
3. Make each fact say one thing everywhere, and state it plainly in a sentence.
4. Confirm the fact is in the raw HTML. Vercel's crawler study found ChatGPT and Claude crawlers fetch JavaScript files but don't execute them, so client-rendered facts may be invisible.

A specific, fetchable sentence gives an engine something it can quote. It does not guarantee the engine will quote it. Mixed signals also raise the risk of brand conflation, where AI blends you with a similarly named company. The free checker on this site shows what engines currently say for three buyer questions, so you can see which facts come back wrong.

• "Organization schema makes you an entity." You are an entity whether or not you add markup. Google states that structured data isn't required for generative AI search, and no markup has evidence behind it as a route into AI answers.
• "Raise your entity authority score." No engine publishes one. Domain Authority is a Moz metric, and Google publishes no site-level authority score.
• "You need Wikipedia to exist." Not as a rule. See whether you need a Wikipedia page.
• "A consistent site fixes every answer." It removes one source of error. Engines also read third-party pages, which you don't control. See how AI weighs your site against other sites.

• Knowledge graph: the store that links entities to facts.
• Wikidata: a public, editable knowledge base of entities.
• Brand conflation: when AI mixes up two entities.

For the practical side, read how to manage your brand as an entity for AI search.