Entities and Structured Data

Track 2 · Lesson 4 of 6 · about 15 minutes

Entities and structured data

Machines do not read words; they resolve references. “Apple” is a fruit, a company, a record label and several towns. This lesson is about making sure that when a system reads your pages, it knows exactly which things you mean — especially when one of those things is you.

Entities are the units of machine understanding

An entity is a distinct thing — a person, organisation, product, concept — that exists independently of the words used to name it. Search and answer engines maintain vast graphs of entities and their relationships, and a large part of understanding a page is mapping its words onto those graphs. When the mapping is confident, you get accurate retrieval and correct attribution. When it is not, you get confusion — or absence.

Structured data states what prose implies

schema.org markup, usually as JSON-LD, lets a page declare its entities and relationships outright: this page is an Article, written by this Person, published by this Organization, about this Thing. Models parse plain HTML perfectly well — the claim that they cannot see unmarked content is false — but explicit statements remove guesswork, and removed guesswork is the entire value proposition.

The single most useful property for disambiguation is sameAs: an identity claim linking your entity to its profiles elsewhere — a Wikipedia page, a LinkedIn profile, an official register. For a brand whose name is also a common word (ask anyone who named their company after a fruit), sameAs is one of the cheapest confidence signals available.

Consistency is the multiplier

Entity understanding is built from every mention of you across the web, not just your markup. The same name, the same description, the same facts on your site, your profiles and your citations elsewhere reinforce one coherent entity. Three slightly different company names and two conflicting founding dates fragment it. Boring consistency — identical NAP-style facts everywhere — is unglamorous and effective.

What to actually implement

One coherent graph per page, not competing fragments: know which system owns each schema type and never emit two conflicting nodes for the same thing. Organization or Person for the publisher with sameAs to your real profiles. Article with a real author. The specific types your content earns — DefinedTerm for a glossary, FAQPage where genuine questions are answered. And nothing your visible page does not support: markup that inflates beyond the content is the one reliable way to turn structured data into a penalty.

Going deeper: schema markup guides · AI knowledge graphs explained · test yourself on the schema quiz.