Entity

Search engines store facts about things, not strings. Google describes its Knowledge Graph as a database of billions of facts about people, places and things, used to answer factual questions and shown in knowledge panels, with facts drawn from public sources, licensed data and content owners, including those who claim a panel and suggest changes^[1].

An entity has a type and an identity. Google’s Knowledge Graph Search API returns entities with standard schema.org types, in JSON-LD^[2]. Schema.org’s most generic type, Thing, carries sameAs: the URL of a reference page that unambiguously indicates the item’s identity, such as its Wikipedia page, its Wikidata entry or its official website^[3].

A site can describe its own. Google says organization markup on a home page can help it understand an organisation’s administrative details and disambiguate it in search results, and can influence what appears in a knowledge panel, such as the logo^[4]. There are no required properties; Google recommends those useful to people, such as the name, an address or telephone number, the URL and the logo^[4]. How such markup works in general is covered under structured data.

AI answers lean on the same clarity. An assistant that recommends a brand has to tell it apart from others with similar names, and Bing’s advice for content cited in its AI answers is to align text, images and video so that they consistently represent the same entities, products or concepts^[5].

A name or an address is a clue, not an identity. QueryRev’s study reports the share of sites that block an AI crawler by name for each country-code ending, from 44.6% of the sites read on .it to 3.7% on .cn, and warns that an ending is not a country, since some are used worldwide^[6].

Search for the brand by name and with its category, and see what the results page treats as one thing: a knowledge panel, the facts in it and the profiles it links. A panel that mixes two companies, or none at all for a well-known brand, shows the gap.

Audit the facts the brand states about itself on its own pages, in its structured data and on its profiles: name, legal name, address, founding date, logo and products. Every mismatch is a reason for a system to doubt which entity it is reading.

Ask AI assistants who the brand is and what it sells, as part of your tracked prompts, and record the errors. An answer that confuses the brand with a namesake is an entity problem, not a ranking one.

Sample data: an invented site, to show the method.

Two businesses called Larkfield sell in the same country: a bed retailer and a maker of garden tools. Asked about Larkfield’s returns, an assistant gives the tool maker’s 14-day policy with the bed retailer’s address. The retailer’s site never states its legal name or links its profiles, and its Organization markup holds a name and nothing else.

The fix is one set of facts, the legal name, the address, the logo and links to its official profiles, stated the same way in the markup, on the about page and on every profile.