Entities and structured data

Last time

You learned how Google judges content quality and how to refresh, merge or remove pages as they age. Read day 05 again.

In this lesson

• Explain the difference between a keyword and an entity, and what a knowledge graph stores.
• Describe how a system decides which entity a page means (disambiguation) and how central it is (salience).
• Write basic schema.org structured data in JSON-LD, and follow Google’s rules for it.
• Choose the schema types that matter for a site in October 2026, and know which rich results Google removed.
• Connect authors, organizations and brands to their profiles on other sites with sameAs.

Day 05 asked if a page is worth reading. This lesson asks a different question: does a search engine know what the page is about? The answer depends on entities.

From strings to things

A keyword is a string of text, for example “taj mahal”. An entity is the real thing behind the string: one specific person, place, organization, product or idea.

Google used this exact example when it launched its Knowledge Graph in May 2012. The words “taj mahal” can mean a monument in India, a Grammy-winning blues musician or a casino in Atlantic City. Google wrote that search had been “matching keywords to queries” for four decades. The new goal was “things, not strings”.

Entity SEO is the work of helping search and AI systems understand the entities on your site. It has three parts:

1. The system identifies each entity correctly.
2. The system knows how the entity relates to other entities.
3. The system knows that mentions on different sites refer to the same entity.

Day 01 showed that vector search matches meaning, not only words. Entities go one step further. They give the meaning a fixed identity that does not depend on the words.

	Keyword	Entity
What it is	A string that a person types	A real thing with an identity
Example	“apple stock price”	Apple Inc., the company
Ambiguity	One string can mean many things	One entity has one identity
Language	Different in each language	The same in all languages
Where it lives	Queries, pages, keyword tools	Knowledge graphs, structured data

Keywords still matter. People type strings, and day 04 showed how to research them. But a search engine turns the strings into entities to understand what the person wants.

Knowledge graphs, attributes and relationships

A knowledge graph is a database of entities and the connections between them. Each entity has attributes: facts about it, for example a founding year or a height. Entities also have relationships with other entities, for example “Pierre Curie was the husband of Marie Curie”.

Relationships let a system answer questions that need facts from more than one place. Google’s 2012 post gave Marie Curie as an example: the graph linked her to her husband and her children, who also won Nobel Prizes.

Google’s Knowledge Graph. At launch in 2012, it had more than 500 million objects and 3.5 billion facts. Today Google’s help page says only “billions of facts”. The facts come from three kinds of sources:

• public sources
• licensed data, for example sports scores and stock prices
• content owners, for example people who claim their knowledge panel

Day 04 showed knowledge panels in the SERP. Google says that its algorithm creates them automatically “when there is enough information available on the open web”. You cannot buy or create a panel. If a panel is about you or your organization, you can claim it and suggest changes. Businesses use their Business Profile instead (day 09).

Wikidata is a large, open knowledge graph that anyone can edit. Each item has an ID that starts with Q, for example Q42 for Douglas Adams. Each property has an ID that starts with P, for example P69 for “educated at”. Many systems use Wikidata IDs as a shared name for an entity.

Disambiguation and salience

When a page says “Taj Mahal”, the system must choose one entity. This is disambiguation. The system uses the other words on the page as clues. “Agra” and “marble” point to the monument. “Blues” and “Grammy” point to the musician.

Salience is how central an entity is to a text. A page about the Taj Mahal has high salience for it. A travel blog that mentions it in one sentence has low salience.

Google’s public Cloud Natural Language API shows both ideas. Its entity analysis returns each entity with a Knowledge Graph ID and a Wikipedia URL, if it can find them. It also gives a salience score from 0 to 1. A score near 1 means that the entity is central to the text.

Not confirmed for Search

The Natural Language API is a Google Cloud product. Google does not say that Search uses the same model or the same scores. Use the API to see how a machine can read a text. Do not treat its scores as ranking factors.

The page about "Taj Mahal" also contains

Agra marble blues Grammy Atlantic City casino

Example data in a toy model. Tick clue words and watch the bars change. Then change where the page names Taj Mahal. Notice that clues decide which entity it is, and placement decides how central it is.

What does this mean for your pages?

• Name the main entity clearly, in the title, the first paragraph and the headings.
• Give context words that only fit your entity. A page about the musician should name his albums and instruments.
• Use one name for one entity, and keep it the same across your site.
• Link to an unambiguous reference where it helps, for example a Wikipedia page. Structured data can do this too, as the next sections show.

Structured data and schema.org

Structured data is a standard format for information about a page. Google calls it “explicit clues about the meaning of a page”. The system does not have to guess from the text. It reads labelled facts: this is a recipe, its cook time is 2 minutes, its author is Mary Stone.

Most structured data uses the schema.org vocabulary. Google, Microsoft, Yahoo and Yandex founded schema.org, and anyone can take part in its work. In October 2026 it has 826 types and 1,540 properties. The types form a tree. The root type is Thing. Below it are types like CreativeWork, Event, Organization, Person, Place and Product. Each type has more specific types, for example Article → NewsArticle.

Three formats

Google supports three formats, and says that all three are “equally fine”:

Format	How it works
JSON-LD	A JSON block in a <script type="application/ld+json"> tag, separate from the visible HTML. Google recommends it.
Microdata	HTML attributes on the visible elements, for example itemprop="name".
RDFa	Different HTML attributes on the visible elements, for example property="name".

Use JSON-LD if your site allows it. Google says that it is the easiest to maintain at scale and causes fewer errors. Google can also read JSON-LD that JavaScript adds to the page. But for product prices and stock, Google warns that JavaScript markup can make Shopping crawls less frequent.

This is the JSON-LD on the Wikipedia page about the monument, in October 2026 (shortened):

{
  "@context": "https://schema.org",
  "@type": "Article",
  "name": "Taj Mahal",
  "sameAs": "http://www.wikidata.org/entity/Q9141",
  "mainEntity": "http://www.wikidata.org/entity/Q9141",
  "headline": "marble mausoleum in Agra, India"
}

The mainEntity value removes all doubt. The page is about Wikidata item Q9141, the monument. The musician’s page points to a different item.

Why it matters, and what it does not do

Structured data has two jobs. It can make a page eligible for a rich result: a search result with extra parts, for example stars, a price or a video thumbnail. It also helps Google understand the page and “the world in general”, for example which people and companies exist.

Google’s case studies report large gains from rich results. For example, Rotten Tomatoes measured a 25% higher click-through rate on 100,000 marked-up pages. But know the limits:

• No guarantee. Valid markup makes a page eligible. Google decides if it shows a rich result.
• Rich results, not rankings. A structured data manual action removes rich result eligibility, but Google says it “doesn’t affect how the page ranks”. Do not add markup to rank higher.
• Not needed for AI answers. Google’s AI optimization guide (May 2026) says that structured data “isn’t required for generative AI search”. There is no special markup for AI Overviews or AI Mode. Day 07 covers AI search.

The rules that matter most

Google’s general structured data guidelines have many rules. These cause the most problems:

1. Mark up only visible content. If the JSON-LD names a performer, the page text must name the same performer.
2. Tell the truth. No fake reviews and no markup about things that are not the focus of the page. This can lead to a manual action, as in day 02.
3. Include all required properties. Google says that fewer complete properties are better than many inaccurate ones.
4. Use the most specific type, for example Restaurant, not only LocalBusiness.
5. Do not block the page with robots.txt or noindex.

Test your markup with Google’s Rich Results Test before you publish. It shows which rich results the page can get. The Schema Markup Validator at validator.schema.org checks all schema.org types, not only Google’s. After launch, Search Console reports show errors across the site (day 08).

The schema types that matter

Google’s documentation, not schema.org, decides what Google uses. Many schema.org types and properties have no visible effect in Google. Google changes this list often, so check the dates.

Type	What Google does with it (October 2026)
Organization	No required properties. Put it on the home page or about page. It helps Google tell your organization apart from others, and can set the logo in the knowledge panel.
Person	Mainly used inside other types, for example as the author of an article.
Article	Types Article, NewsArticle, BlogPosting. No required properties. Helps Google show the right title, images, dates and authors.
Product	Two features. Product snippets are for pages where people cannot buy, for example reviews. Merchant listings are for pages where people can buy, and need name, image and offer price.
Place / LocalBusiness	LocalBusiness is both a Place and an Organization. It needs a name and an address. It feeds business details such as opening hours. Day 09 covers local SEO.
Breadcrumb	BreadcrumbList shows the page’s place in the site. See the next section.
Video	VideoObject needs name, thumbnailUrl and uploadDate. Clip and SeekToAction can mark “key moments”; BroadcastEvent gives a LIVE badge.
Review	Review and AggregateRating can show stars. A business cannot get stars from reviews of itself on its own site. A July 2026 rule bans fake and undisclosed incentivized reviews.
FAQ	Removed. In August 2023 Google limited FAQ rich results to well-known government and health sites. It stopped them completely on 7 May 2026.
HowTo	Removed. Google stopped HowTo rich results on 13 September 2023.

Which matter most? For most sites: Organization on the home page, Article on content pages, Product for shops and LocalBusiness for local businesses. Add Video and Review where the page really has them.

Many old SEO guides still recommend FAQ and HowTo markup. That advice is out of date for Google. Google says unused markup “does not cause problems”, so you do not have to remove it. But do not add it to get rich results. Google also retired course info, estimated salary, learning video, vehicle listing and other types in 2025. This shows a trend: Google keeps fewer rich result types.

The breadcrumb rich result

A breadcrumb is a trail like Books › Science Fiction › Award Winners. It shows where a page sits in the site hierarchy from day 03. With BreadcrumbList markup, Google can show the trail in the result instead of the full URL.

Since January 2025, Google shows breadcrumbs only on desktop. Google said that they get cut off on small screens, so they did not help mobile users. Google’s advice:

• Show a typical path that a user takes to the page. Do not only copy the URL folders.
• You do not need an item for the home page or for the page itself.
• A page can have more than one trail, if users can reach it in more than one way.

sameAs, authors and brands

Structured data can describe more than one entity on a page. It can also connect them. Two tools do this.

@id gives an entity a fixed identifier inside your markup, usually a URL with a fragment, for example https://example.com/#org. Other items can then refer to the same entity. Google’s guidelines give an example: link a recipe and its video with @id. If you do not, Google may not know that the video belongs to the recipe.

sameAs links an entity to its other pages on the web. Schema.org defines it as the “URL of a reference Web page that unambiguously indicates the item’s identity”, for example a Wikipedia page, a Wikidata item or an official site. Google’s Organization docs suggest profile pages on social media or review sites. Schema.org reports sameAs on more than 10 million domains. This is entity linking: it tells a system that many pages describe one entity.

The article's JSON-LD includes

author as a Person with a url publisher as an Organization Person sameAs a profile Organization sameAs Wikidata Organization logo

Example site. Turn on parts of an article's markup one by one. Watch the author change from a text string to a Person entity. Notice that each sameAs adds a link outside the site that confirms an identity.

Authors

Google’s Article docs give author best practices:

• Use the type Person for people and Organization for organizations. Do not use Thing.
• List every author that the page shows, each in their own author item. Do not put two names in one field.
• Put only the name in author.name. Put the job title and the publisher in their own properties.
• Give an author.url to a page that “uniquely identifies the author”, for example a bio page or a social profile.

This connects to E-E-A-T from day 05. Markup does not create trust. But it helps a system connect a real byline to a real person with a real history.

Brands

For products, the brand property names the brand. Google’s merchant listing docs ask for brand.name, with at most one brand name. Global identifiers such as gtin also help, because they identify one exact product.

For your own brand, the Organization markup on the home page is the centre. Give it @id, name, url, logo and sameAs links to your official profiles. Then use the same @id as the publisher of your articles and the brand of your products. Every page then points to one brand entity.

The markup must agree with the rest of the web. Your name, logo and profiles must be the same everywhere. Day 06 covers this as entity consistency.

Key terms

• Keyword: a string of text that a person types or a page contains.
• Entity: one specific real thing, such as a person, place, organization or product, with its own identity.
• Entity SEO: work that helps search systems identify your entities, their relationships and their mentions across the web.
• Knowledge graph: a database of entities, their attributes and their relationships.
• Attribute: a fact about one entity, for example its founding year.
• Relationship: a connection between two entities, for example “founded by”.
• Knowledge panel: a SERP box with facts about one entity from Google’s Knowledge Graph.
• Wikidata: an open, editable knowledge graph with Q IDs for items and P IDs for properties.
• Disambiguation: deciding which entity a name refers to when several entities share it.
• Salience: how central an entity is to a text.
• Structured data: labelled facts about a page in a standard format.
• schema.org: the shared vocabulary of types and properties for structured data.
• JSON-LD: a structured data format in a script tag; Google recommends it.
• Rich result: a search result with extra parts, such as stars, a price or a breadcrumb.
• Breadcrumb: a trail that shows a page’s place in the site hierarchy.
• @id: an identifier that lets items in markup refer to the same entity.
• sameAs: a property that links an entity to other pages that identify it.
• Entity linking: connecting mentions on different pages to one entity.

Exercise

Exercise · about 15 min

Read and write entity markup

Part 1: read markup on a real page.

1. Open https://en.wikipedia.org/wiki/Taj_Mahal in your browser.
2. View the page source (Ctrl+U, or Cmd+Option+U on a Mac). Search for ld+json.
3. Write down the @type, the sameAs value and the mainEntity value.
4. Do the same for https://en.wikipedia.org/wiki/Taj_Mahal_(musician). Compare the two sameAs values.
5. Open the Wikidata URL from step 3. Write down two attributes of the entity.

Part 2: write markup for a site.

6. Pick a site: your own, or an invented business.
7. Write Organization JSON-LD for its home page. Include @id, name, url, logo and two sameAs links.
8. Paste your code into the Rich Results Test (https://search.google.com/test/rich-results) with the “Code” option, or into https://validator.schema.org. Fix any errors.

Show a possible solution

Part 1, checked in October 2026:

• Monument page: @type is Article. Both sameAs and mainEntity are http://www.wikidata.org/entity/Q9141.
• Musician page: also Article, but sameAs is http://www.wikidata.org/entity/Q353023. The same string “Taj Mahal” leads to two different entities. The markup disambiguates them.
• Wikidata Q9141 has the description “marble mausoleum in Agra, India”. Attributes include its country (India) and its coordinates, among about 100 properties.
• Note that the Wikipedia markup uses Article with an Organization as author (“Contributors to Wikimedia projects”). This is correct: no single person wrote the page.

Part 2, a possible answer for an invented coffee roaster:

{
  "@context": "https://schema.org",
  "@type": "Organization",
  "@id": "https://www.example-roasters.com/#org",
  "name": "Example Roasters",
  "url": "https://www.example-roasters.com/",
  "logo": "https://www.example-roasters.com/logo.png",
  "sameAs": [
    "https://www.linkedin.com/company/example-roasters",
    "https://www.instagram.com/exampleroasters"
  ]
}

If the roaster has a shop that people visit, use CafeOrCoffeeShop (a LocalBusiness type) with an address. That is more specific. Then use the same @id as the publisher in the blog articles of the site.

Check yourself

A page about the blues musician Taj Mahal ranks for searches about the monument. What is the most likely cause?

The page has too few keywordsThe page gives weak clues, so the system cannot disambiguate the name wellThe page has no FAQ markupThe page has too much salience for the musician

A site adds valid Product markup and passes the Rich Results Test. No stars or prices appear in Google. Which statement is correct?

The markup must be broken, because valid markup always shows a rich resultGoogle has applied a ranking penaltyValid markup makes a page eligible, but Google decides if it shows a rich resultGoogle only shows rich results for JSON-LD in the page body

In October 2026, a client asks you to add FAQPage markup to every service page to get bigger search results. What do you tell them?

Good idea. FAQ rich results are the easiest rich result to getGoogle stopped showing FAQ rich results on 7 May 2026, so the markup will not give that resultFAQ markup is now a spam violation and causes a manual actionIt works only if the FAQ is hidden from users

Two blog posts list the author as plain text 'Willow Lane, Regula Felix'. What is the best change?

Add both names to one author.name field with a job titleUse two author items, each a Person with a name and a url to a bio pageRemove the authors, because Google ignores themUse the type Thing for each author

Your company has the same name as a famous company in another country. Which markup helps Google tell them apart most directly?

Organization markup on the home page with sameAs links to your official profilesMore keywords with the company name on every pageReview markup with five-star ratings of your own companyA BreadcrumbList on the home page

Sources

• Introducing the Knowledge Graph: things, not strings (Google, May 2012)
• How Google’s Knowledge Graph works (Knowledge Panel Help)
• Google Knowledge Graph Search API (Google for Developers)
• Wikidata: Introduction (Wikidata)
• Natural Language API basics and Entity reference (Google Cloud)
• Introduction to structured data markup in Google Search (Google Search Central)
• General structured data guidelines (Google Search Central)
• Structured data markup that Google Search supports (Google Search Central)
• Latest Google Search documentation updates (Google Search Central, read October 2026)
• Changes to HowTo and FAQ rich results (Google Search Central Blog, August 2023, updated September 2023)
• Optimizing for generative AI features (Google Search Central)
• Google feature guides: Organization, Article, Breadcrumb, Video, Review snippet, Product, Product snippet, Merchant listing, Local business
• Generate structured data with JavaScript (Google Search Central)
• Schemas, Thing, sameAs and LocalBusiness (schema.org)
• Taj Mahal and Taj Mahal (musician) (Wikipedia, page source read October 2026)

Next

Day 06 moves from your own pages to the rest of the web: backlinks, link building, authority metrics, digital PR and brand signals.