What is an entity in SEO?

"SEO entities" sounds like consultancy jargon, but the question underneath is simple: does a search engine know what you sell, or only which words your page uses? Here, entity has nothing to do with a legal entity or a banking entity. It comes from a shift Google announced in 2012, when it started treating searches as real-world things rather than strings of text. It matters to a store because its two main entities, its brand and each of its products, are what a search engine or an AI assistant has to recognize in order to recommend it without confusing it with another.

How it works

Things, not strings. That is how Google introduced its Knowledge Graph on May 16, 2012. For forty years, searching had meant matching keywords to queries, and to a search engine "taj mahal" was just two words, when a person might be thinking of a monument, a musician or a casino. The graph is a model that understands real-world entities and their relationships, and Google summed it up this way: it is not just a catalog of objects, but the relationships between them, which are the key.⁠^{[1]Google, official blog (Amit Singhal). Introducing the Knowledge Graph: things, not strings} Today it describes the graph as its database of billions of facts about people, places and things.⁠^{[2]Google Knowledge Panel Help. How Google's Knowledge Graph works}

A machine finds them in a text. Google Cloud's Natural Language API has a dedicated entity analysis: it looks in a text for known entities, both proper nouns (public figures, landmarks) and common nouns (a restaurant, a stadium), and returns for each one its type, the places in the text where it is mentioned and, when there is one, the reference to a knowledge repository. Its example shows what it is for: in a sentence about "Lawrence of Arabia," it tells the film, typed as a work, from the soldier T. E. Lawrence, typed as a person.⁠^{[3]Google Cloud. Natural Language API basics} That API is a tool for developers, not Google Search, but it shows well what recognizing an entity means.

A page declares them with a vocabulary. In structured data, that vocabulary is usually Schema.org, which says of itself that its vocabulary covers entities, relationships between entities and actions.⁠^{[4]Schema.org. Schema.org} All of its types descend from Thing, "the most generic type of item," which already carries the properties an entity uses to tell itself apart from another: name and alternateName for its names, disambiguatingDescription for a short description that separates it from similar ones, identifier for any identifier (an ISBN, a GTIN) and sameAs, the URL of a reference page that unambiguously indicates the identity of the item, such as its Wikipedia or Wikidata entry or its official website.⁠^{[5]Schema.org. Thing} Google says it can make general use of sameAs and the rest of Schema.org beyond the features it documents.⁠^{[6]Google Search Central. Introduction to structured data markup in Google Search}

And each one carries its identifier. In JSON-LD, the keyword @id uniquely identifies each node with an address (an IRI). The W3C specification explains why: to refer to a node from outside, it matters that it has an identifier, and ideally that address leads to a description of the node itself.⁠^{[7]W3C. JSON-LD 1.1} That is what lets you say "this brand" once and point to it from every product, instead of repeating its name with variations.

In a store

The entities on a product page and how each one is made unmistakable. Fictitious example: the brand, the store and the product are invented.

And the block that declares it, with the brand written once and the product pointing to it through its @id identifier:

<script type="application/ld+json">
  {
    "@context": "https://schema.org",
    "@graph": [
      {
        "@type": "Organization",
        "@id": "https://www.example.com/#organization",
        "name": "Calella",
        "url": "https://www.example.com/",
        "sameAs": ["https://example.net/profile/calella"]
      },
      {
        "@type": "Product",
        "@id": "https://www.example.com/calella-linen-shirt#product",
        "name": "Calella linen shirt, blue",
        "material": "linen",
        "brand": { "@id": "https://www.example.com/#organization" }
      }
    ]
  }
</script>

The line that matters is brand: it does not repeat "Calella" as text, it points to the same node that describes the organization. To a machine, that turns two mentions of a word into a single entity with two relationships. If the brand is written differently on each product page ("Calella," "CALELLA Apparel," "Calella Shirts"), an automated reader has to guess whether they are the same.

How to check it

1. List the entities on a product page: the brand, the product, the manufacturer if it is a different one, the material, the place. If any of them lacks a unique name across the whole store, start there.
2. Search the page source for @id and check that the brand and the product each have a stable one, and that brand points to the brand's.⁠^{[7]W3C. JSON-LD 1.1}
3. Review the identifiers: the product's GTIN if it has one, and the brand's sameAs links, which have to lead to profiles that exist and are yours.⁠^{[5]Schema.org. Thing}
4. Run a description through the entity analysis of Google Cloud's Natural Language API to see which things a machine recognizes in your text and of what type.⁠^{[3]Google Cloud. Natural Language API basics}
5. Search for your brand on Google and see whether a knowledge panel appears: it is the visible sign that Google has it as an entity.⁠^{[2]Google Knowledge Panel Help. How Google's Knowledge Graph works}

Related concept: Knowledge Graph

The entity is the unit; the Knowledge Graph is the collection of entities and relationships that Google maintains, and the knowledge panels come from it.⁠^{[2]Google Knowledge Panel Help. How Google's Knowledge Graph works} Your site can describe its entities with all the care in the world, but what gets into Google's graph is Google's call, using information from many sources besides yours. What is usually called semantic SEO starts from this same idea: write and mark up for things, not for words.

To go on, the Knowledge Graph entry explains how Google's graph works, and the structured data reading path puts the rest of the terms in order.