Entity SEO: How Search Engines Resolve Things, Not Strings

Entity SEO is the practice of stating every person, organization, product, place, and concept on a site so that a search engine resolves each one to a single identified entity instead of matching character strings. This article defines what an entity is, explains how entities are stored in the Knowledge Graph, and shows how entity resolution works on a single page. It then covers entity attributes and relations, entity consistency across a site, and a worked example of entity SEO on holisticradar.com. It closes with entity SEO in semantic SEO services and with how entity SEO relates to schema, Wikipedia, and brand entities.

What an entity is

An entity is a single, uniquely identifiable thing, such as a person, an organization, a place, a product, or a concept, that exists independently of the words used to name it.

A string is a sequence of characters. An entity is the thing the string refers to. The string “Mercury” can refer to at least four different entities, and nothing in the six letters says which one a page means.

Entity SEO is the discipline built on that distinction. It treats a page as a set of statements about entities and makes each statement resolvable: the entity is named in full, its type is clear, and at least one attribute in the same sentence separates it from every other entity that shares its name. Keyword optimization asks which strings a page should contain. Entity SEO asks which things the page is about, what it states about them, and whether a machine can confirm those statements.

Every entity has four properties a search engine can work with: an identifier, a type, a set of attributes with values, and a set of relations to other entities. A page that exposes all four is cheap to understand. A page that exposes only the string leaves the engine to guess.

Entities in the Knowledge Graph

The Knowledge Graph is Google’s database of entities and the facts that connect them, introduced on 16 May 2012 under the phrase “things, not strings”.

Google’s launch announcement, Introducing the Knowledge Graph: things, not strings, described a graph of more than 500 million objects and more than 3.5 billion facts about and relationships between them. Its early sources included Freebase, Wikipedia, and the CIA World Factbook. In May 2020 Google reported that the graph had amassed more than 500 billion facts about five billion entities.

The 2012 announcement named three jobs for the graph. The first was finding the right thing: a search for “Taj Mahal” could mean the monument, a casino, or a local restaurant, and the graph lets Google present the options as distinct entities. The second was summarizing the entity, which became the knowledge panel. The third was going deeper and broader, by following relations from one entity to the next.

Each entity in the graph carries a machine identifier that is independent of language and spelling. Two pages that call the same company by slightly different names can still be linked to one identifier, provided each page gives enough evidence. That is the practical point for SEO: a search engine does not rank a page for a string alone. It interprets the query as a request about entities, interprets the page as statements about entities, and compares the two.

Entity resolution on a page

Entity resolution on a page is the process that maps each mention in the text to one entity, using the words around the mention, the topic of the page, and any structured data as evidence.

Entity linking systems generally work in four steps.

1. Mention detection. The system finds spans of text that name a thing, such as “Mercury” or “Holistic Radar”.
2. Candidate generation. The system lists every known entity the span could refer to.
3. Disambiguation. The system scores each candidate against the context: co-occurring entities, stated attributes, the page’s main subject, and the site’s subject.
4. Linking. The system assigns the mention to the highest-scoring candidate, with a confidence level, or leaves it unresolved.

A page controls the evidence available at step three. A sentence that says “Mercury has an atomic number of 80” resolves on the spot. A sentence that says “Mercury is fascinating” does not. Five page habits raise resolution confidence: name the entity in full on first mention, place a disambiguating attribute in the same sentence, keep one term for one entity throughout the page, name the entity in the heading rather than in a pronoun, and make structured data state the same type and values as the visible text.

Resolution failures are usually silent. The page is indexed and may still rank for some strings, but the engine cannot connect its claims to the right entity, so the page contributes little to the source’s standing on that subject.

Entity attributes and relations

An entity attribute is a property with a value, such as a founding date or a price, and an entity relation is a typed link from one entity to another, such as founder or parent organization.

Both are expressed as triples: a subject, a predicate, and an object. The subject is the entity, the predicate is the attribute or relation, and the object is the value or the other entity.

Attributes fall into three groups by how widely they are shared. Root attributes belong to every entity of a type: every organization has a name and a location. Rare attributes belong to some entities of the type: some organizations publish a named method. Unique attributes belong to one entity alone and identify it. A page that states only root attributes describes a category. A page that states rare and unique attributes describes a specific entity, and it gives the engine facts no other source repeats.

Relations matter because they let an engine move through the graph. A page that states an organization’s founder, its services, and its parent company connects that organization to a Person entity, several Service entities, and another Organization entity. Each connection adds context that helps resolve the next mention.

Entity consistency across a site

Entity consistency is the condition in which every page on a site gives the same name, type, description, and attribute values for the same entity.

A search engine reads a site as a collection of statements. When two pages give the same entity two founding years, two service names, or two descriptions, the engine has to choose one, average them, or lower its confidence in both. None of those outcomes helps the site.

1. One canonical name per entity. Variants such as abbreviations are declared once as alternate names, not rotated for variety.
2. One description per entity. The sentence that defines the organization, a service, or a method is the same wherever it appears.
3. One profile URL per entity. Each person, service, and method has a page that represents it, and mentions point to that page through internal links where links are allowed.
4. One value per attribute. A duration, a price, or a count is stated with the same number and unit on every page.
5. Structured data that mirrors the text. Markup states only what the visible page states, with the same values.

Consistency extends beyond the site. Business profiles, social profiles, directory listings, and press mentions are also statements about the entity. When they match the site, they corroborate it. When they conflict, they compete with it.

Entity SEO on holisticradar.com

Holistic Radar applies entity SEO to its own site through one Organization entity with one description, author entities linked to team profiles, named method entities with fixed attribute values, and one schema graph output by the theme in JSON-LD only.

The schema graph contains Organization, WebSite, a founder Person, WebPage, Service, BlogPosting, FAQPage, and BreadcrumbList nodes. The theme outputs it, so no page can introduce a second description of the organization through a plugin or a hand-written snippet.

The methods show why entity consistency is a discipline and not a one-off task. The Six-Layer Pyramid is always six layers, from L1 source context and central entity to L6 task completion and revenue. The Radar Scan is always free and always five business days. Because each method is treated as an entity with one attribute set, a search engine or an AI system that reads any page on the site receives the same facts.

Entity SEO in semantic SEO services

Entity SEO is the first layer of Holistic Radar’s semantic SEO services, because the source context and central entity sit at L1 of the Six-Layer Pyramid and every layer above them depends on that definition.

A weak layer caps every layer above it. A query network, a topical map, and page semantics all describe attributes of the central entity, so a site that has not defined its central entity cannot decide which attributes deserve pages, which pages are core, or which facts must stay consistent.

On a client site, entity work therefore comes before content work: the central entity and source context are defined, the entity inventory is listed with one name and one description each, attribute values are fixed, and the schema graph is designed to state them. The rest of the semantic SEO service is built on that record.

Entity-attribute-value graphs and triples

An entity-attribute-value graph is the structure search engines build from triples: each entity linked to its attributes, and each attribute to a value.

Search engines decompose documents into subject, predicate and object triples and store them in knowledge graphs. Retrieval-augmented generation in AI search reads the same triples back out to compose answers. A page written as clean triples feeds both systems directly. The sentence rules that produce clean triples are set out in microsemantics.

→ Semantic SEO services: the central entity defined at L1 and carried through the topical map and content network

Entity SEO and schema, Wikipedia, and brand entities

Entity SEO meets three neighbouring subjects that are often confused with it: schema markup, Wikipedia, and brand entities.

Is schema markup required for entity SEO?

Schema markup is not required for entity SEO, but it removes ambiguity by stating entity types and relations in a machine-readable form. Google Search Central’s Introduction to structured data markup explains that Google uses structured data to understand page content and to gather information about the people, books, or companies the markup describes. The markup supports the visible text; it does not replace it, and it must describe content the page actually shows.

Does an entity need a Wikipedia page?

An entity does not need a Wikipedia page to be recognized by search engines. Wikipedia was one of the Knowledge Graph’s early sources and remains a strong corroborating reference, but its notability and conflict-of-interest rules mean a page created for marketing purposes is likely to be removed. Consistent statements across the entity’s own site, its official profiles, and independent sources build recognition without it.

How does a brand become an entity?

A brand becomes an entity in a search engine’s systems when enough consistent, corroborated statements exist for the system to identify it as one thing. Those statements include Organization markup on the official site, sameAs references to official profiles, one description repeated everywhere, and independent mentions that match it. A brand that appears in a knowledge panel can be claimed by its verified representatives, which lets them suggest corrections.

Is entity SEO different from technical SEO?

Entity SEO is different from technical SEO: technical SEO makes a page retrievable, and entity SEO makes what the page says resolvable. A page can pass every crawl and render check and still fail entity resolution because its mentions are ambiguous or its values conflict with other pages.

→ Semantic SEO: the process that builds on resolved entities

→ Semantic search: how Google moved from strings to meaning, system by system

→ Structured data: how markup states entity types and relations

Entities are resolved from the query network before any page is planned. → Entity and query research