Entity based SEO: how search engines model things rather than strings

Entity based SEO is optimising for the things a page is about instead of the words it happens to contain. A search engine stores an entity as a node: a stable identifier, a type such as Organization or Person or Place, a set of attributes, and relationships to other nodes. The query "jaguar service interval" resolves to the car maker, not the animal, because the engine matched a thing rather than a string. The practical consequence is that your business is either a resolvable entity, with one name, one home, confirmed attributes and corroboration from sources the engine already trusts, or it is a phrase that appears on some web pages. The work below moves you from the second to the first.

This page covers the model and the work it implies. For the retrieval mechanics underneath it, see how search engines work.

String matching and entity matching, side by side

	String based	Entity based
Unit of meaning	The word and its variants	The identified thing
"cheap flights to turkey"	Matches pages containing those words	Resolves Turkey the country, flight the service, price the intent
Synonyms	Handled by matching more words	Handled by one node with many labels
Your lever	Where the phrase appears on the page	Whether the engine can tell which thing you are
Failure mode	Thin pages stuffed with variants	Two companies with similar names merged into one

How an entity gets built

• Extraction. The engine reads a page and pulls out candidate entities: names, places, products, dates, people.
• Disambiguation. It decides which known node each candidate refers to, using the surrounding text. "Apex Legal, Chennai, insolvency" is a different node from "Apex Legal, Leeds, conveyancing", and the words around the name are what separate them.
• Corroboration. Attributes repeated across independent sources get treated as settled. A founding year that appears on your site, your LinkedIn page and two industry directories is a fact. One that appears only in your footer is a claim.
• Identifier. The node gets an internal ID. Public identifier systems such as Wikidata are the closest thing the open web has to the same idea, which is why a Wikidata item is a useful anchor rather than a vanity entry.
• Relationships. Who founded it, what it sells, where it operates, who it is part of. Relationships are what make a node answerable to questions you never wrote a page for.

The five jobs that make your business resolvable

1. One name, spelled one way. Decide between "Apex Legal", "Apex Legal LLP" and "ApexLegal" and use the chosen form in your title tags, your schema, your directory listings, your social profiles and your press. Variants split corroboration across two nodes.
2. One entity home. Pick the URL that represents the company, normally the home page, and make every other reference point at it. That page carries the name, the description, the location, the founding facts and the links out to your profiles.
3. Declare the node in markup. Organization or Person schema with sameAs is how you say, in a form no parser can misread, that these accounts are all the same thing.
4. Earn corroboration you do not control. Industry bodies, regulator registers, trade press, review platforms, conference speaker pages. These matter more than anything you publish about yourself, because independence is the whole point of corroboration.
5. Cover the relationships. A page per person, a page per service, a page per location, each linked to the others. Internal linking is how you express those relationships in a form a crawler follows.

The markup, with the properties that do the resolving

sameAs is the important line. Each URL is a separate confirmation that the node and the profile are one thing.

<script type="application/ld+json">
{
  "@context": "https://schema.org",
  "@type": "Organization",
  "@id": "https://example.com/#organization",
  "name": "Apex Legal LLP",
  "alternateName": "Apex Legal",
  "url": "https://example.com/",
  "logo": "https://example.com/logo.png",
  "foundingDate": "2011-04-01",
  "areaServed": { "@type": "Country", "name": "India" },
  "sameAs": [
    "https://www.linkedin.com/company/apex-legal-llp",
    "https://www.wikidata.org/wiki/Q000000"
  ]
}
</script>

Replace the Wikidata URL with your real item, or leave the property out until you have one. People resolve the same way, and knowsAbout plus worksFor is what ties an author to a company and a subject:

<script type="application/ld+json">
{
  "@context": "https://schema.org",
  "@type": "Person",
  "@id": "https://example.com/team/priya-nair#person",
  "name": "Priya Nair",
  "jobTitle": "Partner",
  "worksFor": { "@id": "https://example.com/#organization" },
  "knowsAbout": ["insolvency law", "corporate restructuring"],
  "url": "https://example.com/team/priya-nair",
  "sameAs": ["https://www.linkedin.com/in/priya-nair-example"]
}
</script>

The @id values are what turn two separate blocks into a connected graph. Without them you have published two unrelated facts; with them you have published one relationship. Google's structured data introduction covers the general rules, and schema.org sameAs defines the property.

Entity thinking changes the keyword work

• Plan by topic, publish by question. One page per decision a buyer makes, not one page per phrase variant. Three near identical pages compete with each other and dilute the node.
• Name the related things on the page. A page about insolvency resolution that never names the tribunal, the statute or the roles involved reads as thin to an extraction step, regardless of word count.
• Give every entity a page it owns. If you want to be associated with a product, a method or a location, that thing needs a URL.
• Write the definition early. The first paragraph that says plainly what a thing is does most of the extraction work for you.

How to test whether you resolve

1. Search your brand name alone. Count how many of the first ten results are yours or about you.
2. Search the name plus your city and plus your category. If a different company appears, you have a disambiguation problem, not a ranking problem.
3. Search Wikidata and your main industry registers for your name and note which hold a current entry.
4. Check whether a knowledge panel appears for the brand, and whether the facts in it match your own pages.
5. Run your home page through the Rich Results Test and confirm the Organization block parses with its sameAs list intact.

Questions buyers ask about entity based SEO

Is entity based SEO different work, or the same work described differently?

Much of it overlaps with good technical and content practice. What changes is the priority order: naming consistency, markup and third party corroboration move ahead of on page phrasing.

Do I need a Wikidata item?

It helps as a public anchor with a stable identifier, and it is worth creating when your company meets the notability expectations there. Industry registers and association directories do similar work and are available to everybody.

What is entity salience?

How central a thing is to a document, as opposed to merely mentioned in it. A page where your main entity appears in the title, the opening paragraph, the headings and the markup reads as being about that entity.

How does this relate to E-E-A-T?

Experience, expertise, authoritativeness and trust are assessed about entities. Author pages, credentials and consistent attribution are the mechanism that connects a quality signal to a node rather than to a URL.

What breaks an entity fastest?

A rebrand done halfway. New name on the site, old name in the directories, two LinkedIn pages and a logo mismatch gives an engine two candidate nodes and no way to merge them. Change every reference in one pass and fix the citations in the same week.

Where entity work pays twice

You build an entity so Google can resolve your business as a named thing. The same work serves assistants, which also need to know who you are before they can recommend you.

The test is the same too. Ask an assistant "what is" and "who is" about your company and your founders, then ask for a recommendation in your category. A clear, correct description and a mention means the entity resolved. A blank or a confusion with a similar name means it did not. AI Knows Us handles both kinds of question across the assistants and shows where the facts were sourced.