Entity SEO: Entities, the Knowledge Graph and Semantic Search

By Christoph Hein · Head of SEO & Search Consultant · Updated August 2026

Search engines stopped working purely with strings a long time ago. They work with things: people, organisations, places, products, concepts. Those things are called entities, and they are why Google can answer a query in which the searched-for word never appears. Entity SEO means building content so that the entities within it are unambiguously identifiable and correctly connected – for classic search and for AI answer systems alike.

1. What is an entity?

An entity is a uniquely identifiable thing that exists independently of the label attached to it. It has three properties:

• Uniqueness: there is exactly one Cambridge in England – and another in Massachusetts. Two entities, similar name.
• Attributes: properties such as founding year, population, occupation, price.
• Relationships: connections to other entities – "located in", "works for", "part of".

The crucial difference from a keyword: a keyword is a string. "Apple", "Apple Inc." and "the iPhone maker" are three different strings but the same entity. Conversely, "Jaguar" is one string pointing at least three different entities: the animal, the car brand and a type of guitar.

2. From keywords to entities: what changed

In practice this means: a text about "crawl budget" in which Googlebot, robots.txt, log files, status codes, indexing and Search Console appear naturally is recognisably a text about the topic – without repeating the head term twenty times. Conversely, repeating it twenty times helps nothing if the accompanying concepts are missing.

3. The Knowledge Graph

Google launched the Knowledge Graph in 2012 with the line "things, not strings". It is a database of entities, their attributes and their relationships. Every entity has an internal identifier – the machine ID, recognisable in forms like /m/02y_9m3.

The main data sources:

• Structured open data: Wikipedia and Wikidata as the backbone. Freebase, the original foundation, was retired in 2016 and migrated into Wikidata.
• Verified first-party claims: Google Business Profile for local entities.
• Licensed databases: for films, music, sport and similar domains.
• Extraction from the open web: recurring, consistent statements about an entity – this is where entity SEO operates.
• Structured data on websites: Schema.org markup as a machine-readable first-party claim.

4. How Google recognises entities in text

The process has three steps, borrowed from natural language processing:

1. Named entity recognition: identifying that a sequence of words denotes an entity ("Popken Fashion Group" is an organisation).
2. Disambiguation: deciding which entity is meant. That happens through context – if "engine", "horsepower" and "gearbox" appear in the text, "Jaguar" is the car brand.
3. Entity linking: connecting the recognised entity to its record in the knowledge base and attributing the statements in the text to that entity.

Such systems additionally estimate salience: how central is an entity to the document? A text can mention fifty entities while genuinely covering only two. Salience is why long lists of terms do not work – they produce many mentions with low centrality.

If you want to see this in action, Google's Natural Language API has a public demo that outputs the recognised entities and their salience scores for any text you paste in. It is not the search algorithm, but it demonstrates the same class of processing.

5. Why entities matter for AI search

Generative answer systems have to pick a source for every statement. Their selection is not based on keyword overlap but on which source attributes a statement to an entity most clearly and consistently. Three consequences:

• Clarity beats reach: a text that states precisely who is making the claim and what it refers to gets cited more readily than a generically worded one.
• Consistency across sources: if your website, your LinkedIn profile and your Wikidata record all state the same facts, confidence in the attribution rises. Contradictions lower it.
• Verifiable authorship: original data, clearly named authors with checkable profiles, concrete numbers. Interchangeable text is worthless in a system that picks one source per statement.

How to offer AI crawlers additional structure is covered in the guide to llms.txt & AI crawlers.

6. Entity mapping: the practical groundwork

Before you think about markup, you need a map of your topic's entities. Five steps:

1. Define the primary entity. What is the page about? Exactly one entity, not three. On this page it is "entity SEO".
2. Collect secondary entities. Which concepts must someone know to understand the topic? Here: Knowledge Graph, named entity recognition, Wikidata, Schema.org, salience, disambiguation.
3. Name the relationships. Do not just list them, write the connection out: "Wikidata is a data source for the Knowledge Graph." Those sentences are the actual content.
4. Clarify the labels. Which synonyms and spellings exist? Use the most common form as the main label and mention the others at least once.
5. Write the boundary. Where does the entity end? A paragraph saying "entity SEO is not the same as keyword research, because …" creates more clarity than three more descriptive paragraphs.

Useful research sources for step 2: the Wikipedia article on the topic and its outbound links, the properties of the Wikidata record, plus "people also ask" and related searches in the search results.

7. Technical implementation with structured data

Schema.org offers three properties that are central to entities:

• @id – a stable, unique identifier for the entity within your data.
• sameAs – references to the same entity elsewhere on the web. This is the single most important property for entity SEO.
• about and mentions – what a document is about (primary entity) and what it merely references (secondary entities).

Marking up an organisation unambiguously:

<script type="application/ld+json">
{
  "@context": "https://schema.org",
  "@type": "Organization",
  "@id": "https://example.com/#organization",
  "name": "Example Ltd",
  "url": "https://example.com/",
  "logo": "https://example.com/logo.png",
  "foundingDate": "2011-04-01",
  "address": {
    "@type": "PostalAddress",
    "streetAddress": "1 Sample Road",
    "postalCode": "BS1 4TR",
    "addressLocality": "Bristol",
    "addressCountry": "GB"
  },
  "sameAs": [
    "https://en.wikipedia.org/wiki/Example_Ltd",
    "https://www.wikidata.org/wiki/Q12345678",
    "https://www.linkedin.com/company/example-ltd/",
    "https://github.com/example-ltd"
  ]
}
</script>

And connecting a document to its primary and secondary entities:

<script type="application/ld+json">
{
  "@context": "https://schema.org",
  "@type": "Article",
  "headline": "Entity SEO: Entities and the Knowledge Graph",
  "about": {
    "@type": "Thing",
    "name": "Entity SEO",
    "sameAs": "https://www.wikidata.org/wiki/Q1136290"
  },
  "mentions": [
    {
      "@type": "Thing",
      "name": "Knowledge Graph",
      "sameAs": "https://en.wikipedia.org/wiki/Knowledge_Graph"
    },
    {
      "@type": "Organization",
      "name": "Google",
      "sameAs": "https://www.wikidata.org/wiki/Q95"
    }
  ],
  "author": {
    "@type": "Person",
    "@id": "https://example.com/#author-js",
    "name": "Jane Sample",
    "jobTitle": "Head of SEO",
    "sameAs": [
      "https://www.linkedin.com/in/jane-sample/",
      "https://example.com/team/jane-sample"
    ]
  }
}
</script>

The general rules for implementation are in the guide to JSON-LD and structured data. For entities, three additional points apply: sameAs may only reference profiles you genuinely own; the @id must stay stable for years; and the markup must not assert anything that is not visible on the page.

8. Establishing your own brand as an entity

The goal: Google should know your brand as an entity in its own right, not as a string on a website. The steps, in order of impact:

1. One canonical self-description: an "about" page stating founding, location, activity and the people responsible in plain sentences. This page is the reference everything else points to.
2. Consistent details everywhere: company name, address, founding year and spelling must be identical across website, legal notice, business profile and social profiles. Discrepancies are the most common reason an entity stays unclear.
3. Organization and Person markup with complete sameAs references.
4. Verified external reference points: Google Business Profile, company registers, trade associations, professional profiles.
5. Mentions in independent sources: trade press, studies, interviews. For corroborating an entity, independent statements count for more than your own.
6. Treat authors as entities: a real author page with credentials and links to external profiles, connected via author and a stable @id.

9. Internal linking as an entity graph

Your internal linking is the statement Google can read most easily about which topics belong together. Three principles:

• One term, one page: every entity in your topic area should have exactly one target page. Two pages on the same term create duplicates and dilute the attribution.
• Link at first mention: when a secondary entity first appears, link to the page that explains it. The anchor text should be the entity's name – the most precise form of relevance transfer there is.
• Connect clusters both ways: not only hub to detail, but detail to detail where a genuine relationship exists. That is what mirrors the graph.

10. Common mistakes

1. Listing entities instead of explaining them. A list of related terms at the end of a text is not semantic optimisation, it is keyword stuffing with a new name.
2. sameAs pointing at other people's profiles. References to Wikipedia articles about other organisations, or to profiles you do not control, are false claims.
3. Markup without a visible counterpart. Information that exists only in the JSON-LD violates Google's structured data guidelines.
4. Contradictory details across channels. Different legal names or addresses in the legal notice, the markup and the business profile.
5. An unstable @id. If the identifier changes with every relaunch, you create a new claim each time.
6. Several primary entities per page. Cover three topics at once and you will be recognised as the definitive source for none of them.
7. Treating entity SEO as a replacement for keyword research. Demand is still measured through search terms. Entities determine how content is built, not whether there is demand for it.

11. Checking entity visibility

• Brand search: does a knowledge panel appear? Are sitelinks and social profiles shown? That is the most direct indication of entity recognition.
• Knowledge Graph Search API: a query shows whether and under which identifier an entity is held. The API covers a subset, so a negative result is not proof.
• Natural Language API demo: paste your own text and check whether the intended primary entity is recognised with high salience.
• Structured data testing: the Schema Markup Validator shows whether your annotations parse without errors.
• Citation test in AI systems: ask the answer systems relevant to your field about your subject area and check whether and how you are cited.
• Search Console: rising impressions for topically related queries you never explicitly optimised for are a good sign of semantic attribution.

12. Entity SEO checklist

1. Define exactly one primary entity per page.
2. Collect secondary entities from Wikipedia, Wikidata and related searches.
3. Write out the relationships between the entities in full sentences.
4. Use synonyms and alternative spellings at least once.
5. Write a boundary section: what the entity is not.
6. Add Organization or Person markup with a stable @id.
7. Point sameAs only at your own verified profiles.
8. Add about and mentions for your key documents.
9. Check company details across all channels for consistency.
10. Create author pages with verifiable credentials and connect them.
11. Structure internal linking so each entity has exactly one target page.
12. Validate the markup with the Schema validator.
13. Review brand search and knowledge panel quarterly.

13. Frequently asked questions

Read next

About the Author

Christoph Hein

Head of SEO at Popken Fashion Group & independent Search Consultant

Christoph has spent 10+ years in search, currently steering organic strategy for 5 fashion brands across 13 countries and more than 30 domains. Alongside his in-house and consulting work, he founded niche content portals such as Angelmagazin.de and BaristaCompass.com, and built the Rank-O-Saur extension to make technical SEO audits effortless. Every guide here is grounded in hands-on, data-driven practice rather than theory.

christophhein.com ↗ LinkedIn ↗