Semantic SEO: Entities, Topics and a Method

Semantic SEO is the practice of optimizing content for meaning: the entities a topic involves, how they relate, and the questions people ask about them, instead of for repeated keyword strings. It is how you write for search systems that understand things rather than match words, which since 2012 has meant Google, and now also means every AI answer engine.

This guide is the practical companion to our explainer on semantic search. That post covers the technology: embeddings, vector retrieval and how Google got there. This one covers what you do about it: a repeatable method with an entity list, a coverage matrix, relationship-first writing and schema markup, plus a worked example.

On this page

• Semantic SEO vs keyword SEO
• Why search engines think in entities
• How do you do semantic SEO step by step?
• Building an entity coverage matrix
• Schema markup for semantic SEO
• What is a semantic SEO example?
• Mistakes that pass for semantic SEO
• Making semantic SEO part of every page

Semantic SEO vs keyword SEO

Keyword SEO starts from a phrase and asks where to put it; semantic SEO starts from a topic and asks what a complete answer must contain. Both still need keyword research. They differ in what happens next.

The shift is not new. Google’s SEO Starter Guide makes the basic point with an example: some people search “charcuterie” and others “cheese board” for the same thing. A writer who understands the subject covers both because both are part of it, not because a tool flagged them.

Why search engines think in entities

Google has organized its understanding of the world around entities since at least 2012. When it launched the Knowledge Graph, Google described it as understanding “things, not strings”, with “more than 500 million objects, as well as more than 3.5 billion facts about and relationships between these different objects” at launch. Its example was “taj mahal”, a query that could mean a monument, a musician, a casino or a restaurant.

The systems that followed extended this to language. Google’s ranking systems guide describes neural matching as a system that helps Google “understand representations of concepts in queries and pages and match them to one another,” and BERT as helping it “understand how combinations of words express different meanings and intent.” How those systems work is the subject of our semantic search explainer. For the practitioner, the takeaway is simple: the page is read as a set of things and claims, not as a bag of words.

The diagram below shows the difference with Google’s own example.

Scroll the diagram sideways to see all of it.

For how this applies to your own brand as an entity, see our entity SEO guide and entity disambiguation. This post stays with topics and pages.

How do you do semantic SEO step by step?

Semantic SEO is a planning method before it is a writing style. We use six steps.

1. Define the topic and intent. One page, one intent. Write the one-sentence answer the page exists to give.

2. Build the entity list. List the things a complete answer involves: products, standards, organizations, places, people, processes and measurements. Sources:

• the pages already ranking (what do all of them name?)
• People Also Ask and related searches; see our People Also Ask guide
• the topic’s Wikipedia article and its Wikidata item, which list related entities and properties
• an entity extraction tool such as Google’s Natural Language API, which returns the entities in a text with a salience score from 0 to 1 and, for many, a Knowledge Graph ID

3. Add attributes and relationships. For each entity, note what a reader needs to know: what it is, what it is part of, what it causes, how it compares. “Sitemap index files group multiple sitemaps” is a relationship. “Sitemap index” alone is just a term.

4. Map entities to pages. Decide which entities the page explains, which it mentions and links, and which belong on another page. This is the coverage matrix in the next section.

5. Write relationship-first. State relationships in plain sentences: “X is a type of Y,” “X requires Y,” “X replaced Y in 2019.” Name things specifically instead of using pronouns across paragraphs. That also makes each passage easier to retrieve on its own.

6. Mark it up and link it. Use schema to state the main entity and the ones mentioned, and internal links to connect pages that explain related entities.

Building an entity coverage matrix

An entity coverage matrix is a grid of pages against entities that shows, at a glance, where each entity is explained, merely mentioned, or missing. It is the most useful single artifact in semantic SEO because it exposes gaps and overlaps across a whole cluster.

The matrix below uses a small cluster on XML sitemaps as an example.

Scroll the diagram sideways to see all of it.

How to use it:

• A row with no E is an entity nobody on your site explains. Decide whether it needs its own page or a section.
• A row with two E’s is a likely overlap. Two pages explaining the same entity in depth will compete.
• A column full of E’s is a page trying to explain everything. Split it or move detail to other pages and link.
• Every M should be a link to the page with the E. That is how internal links carry relationships, not just authority.

A matrix across a whole cluster is also the fastest way to see whether you are building topical authority or a pile of overlapping posts.

Schema markup for semantic SEO

Schema markup lets you state in machine-readable form which entities a page is about, which it mentions, and which known identifiers they match. It is the explicit layer on top of clear writing.

The properties that matter most here:

• about: schema.org defines it as “the subject matter of an object”. Use it for the page’s main entity or entities.
• mentions: “indicates that the CreativeWork contains a reference to, but is not necessarily about a concept”. Use it for the important secondary entities.
• sameAs on those entities: links to the Wikidata item, Wikipedia article or official site that identifies the thing unambiguously. Our sameAs schema guide covers the details.

Scroll the diagram sideways to see all of it.

Two cautions. Markup that claims entities the page does not actually discuss is misleading and gains nothing. And not every property produces a visible search feature; about and mentions are there to make meaning explicit, not to earn rich results. For implementation, see our JSON-LD guide.

What is a semantic SEO example?

Take a page targeting “xml sitemap”. A keyword-led draft repeats “XML sitemap” in every heading and paragraph and adds variants like “sitemap XML file”. A semantic draft looks different:

• It defines the entity in the first sentence: what a sitemap is and what it is for.
• It names the standard it follows, the sitemaps.org protocol, and says who maintains it.
• It explains attributes readers need: the lastmod field, sitemap index files, where the file lives.
• It states relationships: that robots.txt can point to the sitemap, that Search Console accepts submissions, how sitemaps relate to crawl budget.
• It links to the pages that explain robots.txt, Search Console and crawl budget in depth, rather than explaining them again. Our own XML sitemap guide is built this way.
• It marks up the page with about for sitemaps and mentions for the related entities.

The phrase “xml sitemap” still appears where it reads naturally. It just stops being the organizing principle.

Mistakes that pass for semantic SEO

Several popular tactics borrow the language of semantic SEO without the substance:

• Synonym sprinkling. Adding variants of the keyword is not entity coverage. Explaining related things is.
• Term-frequency tools as targets. Tools that tell you to use a word 14 times reproduce the average of the current results. Use them to spot missing concepts, not to set counts.
• Schema without content. Marking up entities the text never explains.
• One page for everything. Cramming every entity onto the pillar instead of splitting by intent and linking.
• Ignoring disambiguation. If your subject shares a name with something else, say explicitly which one you mean in the first paragraph.
• Vague writing. “It,” “this” and “the tool” across paragraphs force a machine to guess what you mean. Repeating the entity’s name is not stuffing when it removes ambiguity.
• Covering entities the reader does not need. A complete answer is not an exhaustive one. If an entity does not help this reader with this intent, link to it or leave it out.

A quick test for any draft: highlight every named thing in it. If the highlights are mostly variations of your keyword, the page is keyword-led. If they are the products, standards, organizations and processes your subject actually involves, with sentences explaining how they connect, the page is doing semantic SEO.

Making semantic SEO part of every page

Semantic SEO works best as a habit rather than a project. Add an entity list and a relationships line to every brief, keep a coverage matrix for each cluster, write relationship-first, and mark up the main entity and the important mentions. Over a cluster, that produces pages that are clearer for readers, easier for search systems to match to varied queries, and easier for AI engines to quote. For the machinery behind why this works, read our explainer on semantic search.

If you want entity research, coverage matrices and schema built into your content process, that is part of our AI SEO service.