Insights/SEO/GEO and agentic commerce/27 October 2026/Jean-Baptiste Duquesne

Entities and schema.org: describing your brand with facts AI can verify

Search engines and AI assistants don't reason in keywords but in entities: a brand, a product, a person, a place, linked by facts. The more precise and consistent those facts, the easier you are to place. Here is how to build that description, and what schema.org can and cannot do.

Entities and schema.org: describing your brand with facts AI can verify

From keywords to entities

For a long time, SEO meant putting the right words in the right places. Search engines and AI assistants now work differently: they try to identify entities (a brand, a product, a person, a place, a concept) and the relationships between them.

When an assistant answers "which Bordeaux estates grow forgotten grape varieties?", it is not looking for a page that repeats that sentence. It is looking for estates, located in Bordeaux, associated with specific grape varieties, themselves described as rare or old. Your brand gets recommended if those links are established and confirmed by several sources.

À retenir

An entity is an identifiable "what", not a word.
What matters are the facts linking it to other entities.
Those facts must be consistent on your site and elsewhere. :::

Precise facts rather than vague labels

The difference lies in the details. Take a wine's product page. A description by labels gives:

Vague label	Precise, verifiable fact
Soil: gravel	Soil type: deep gravel, named plot
Grape: Mancin	Grape variety: Mancin des Palus, red
Winemaking: natural	Fermentation: indigenous yeasts
—	Producing estate: Château Cazebonne
—	Region: Bordeaux vineyard
—	Farming: [organic or biodynamic, according to the actual certification]

The right-hand column links the wine to an estate, a terroir, a grape variety, a practice. It only holds information that is relevant and verifiable: you don't add a certification you don't have.

The same reasoning applies to every entity of a brand: the company itself (legal name, address, executives, activities), its products or services, its experts (background, areas of expertise, publications), its projects.

What schema.org brings, and what it doesn't

Schema.org is a shared vocabulary for describing these entities in a machine-readable format: Organization, Person, Product, Offer, Article, FAQPage, and so on. In particular it lets you:

link a person to the organisation that employs them, and an article to its author;
point, with the sameAs property, to the entity's official profiles (LinkedIn, Wikidata, company registry…) to resolve name clashes;
describe product characteristics beyond the standard fields, with additionalProperty.

It is worth staying clear-eyed, though. Google states that no special markup is needed to appear in its AI answers, and reminds that structured data must match the visible text of the page. Schema.org does not create authority that isn't there: it describes more clearly what is already true and already displayed.

:::encadre The consistency rule A fact declared in the markup must appear on the page and must not be contradicted elsewhere: directories, marketplaces, social networks, press. An inconsistency is often worth less than an absence.

Organising the entities on your site

Before marking anything up, decide which entities your site describes and how they connect. It is the same logic as pillar pages and internal linking.

Inventory: list the key entities (organisation, offers, experts, products, projects, topics).
Define one reference page per entity: a single authoritative URL that other pages link to.
Write the facts: for each entity, the precise and verifiable attributes, with their source.
Mark up: translate those facts into schema.org, linking entities together (author, publisher, manufacturer, place).
Align the outside world: fix directory listings, profiles and marketplaces that contradict those facts.
Measure: check how assistants describe the brand before and after, for example with GEO Radar.

The trap to avoid

The temptation is to mark up everything, including marketing claims ("leader", "the best") or unverifiable data. It backfires: such claims bring no usable fact, and markup that doesn't match the page may be ignored. Ten accurate facts beat fifty approximate properties.

Start by seeing how AI describes your brand today: run a GEO Radar scan, then let's discuss your entities in an audit.

Written by

Jean-Baptiste Duquesne

Partner — SEO, GEO & CRM

A web pioneer since 1995 and founder of 750g.com (sold to Webedia). At Good Morning AI he leads SEO, visibility inside generative engines (GEO) and customer lifecycle work.

All his articles LinkedIn

Published on 27 October 2026

The Good Morning AI newsletter

Daily, or weekly on Tuesday: your choice.

Want to discuss this with us?

30 minutes, a quick call, and we'll see together how we can help you move forward.

Discuss your project