Entities and Evidence: How Machines Figure Out What Your Website Is Actually Talking About

An entity is the person, business, place or other identifiable thing a page describes. Clear names, relationships and supporting facts help readers and machines distinguish it from something with a similar name.

For example, “Mithril” can refer to a fictional metal or a business. The organization’s name, website, services and consistent contact information establish which one the page means. For AI-search readiness, the work is practical: make the business identity explicit, align visible information with structured data and support important claims with evidence.

What is an entity?

An entity is a distinct thing.

For a service-business website, common entities include:

• the organization,
• people,
• locations,
• services,
• products,
• credentials,
• awards,
• cases,
• events,
• and sometimes named methodologies or tools.

The important part is not merely naming them. It is describing their relationships.

• Mithril is an organization.
• Lumen is a website-intelligence project associated with Mithril.
• A team member works for Mithril.
• A service is offered by Mithril.
• A location belongs to the organization.
• That office provides certain services.

Those relationships help turn text into a coherent model of the business.

The entity problem on real websites

Consider an illustrative company called Acme Services. Its homepage uses the trading name, its footer uses the legal name and an older directory lists a former address. A reader needs enough context to understand which details describe the current business.

Make the relationship between trading and legal names explicit. Keep current addresses and contact information aligned, and label historical material so it cannot be mistaken for the latest record. The goal is consistency with context. Natural abbreviations are fine when they clearly refer to the same organization.

Clear naming is underrated

Bing’s current webmaster guidance specifically recommends defining entities clearly and consistently across people, organizations, products, and locations. A company should have an obvious canonical name. A person should have one strong profile.

A location should have one authoritative page. A service should have a durable canonical URL. Variations can exist naturally. But they should point back toward the same thing.

Entity clarity is not schema

Schema can encode entity information. It is not the entity strategy itself. A website can have immaculate JSON-LD and terrible visible information. That is still a weak implementation.

The visible site should clearly communicate:

• who,
• what,
• where,
• how things relate,
• and what evidence supports the claims.

Structured data should reinforce that model. Not invent it.

Evidence changes the quality of a claim

Consider:

“We are experts in AI search.”

Okay.

Now:

“We audit raw HTML against rendered DOM output, inspect crawler policy, review Bing grounding data, and use Lumen to turn findings into a prioritized optimization plan.”

That is more informative.

Now add:

• screenshots,
• examples,
• documented methodology,
• case studies,
• or primary data.

The claim becomes evidence-backed. This matters for humans. It also creates content that machines can use more confidently because the page contains explicit support rather than promotional adjectives. Google’s 2026 generative Search guidance specifically recommends unique, non-commodity, firsthand content.

Bing recommends explicit, independently verifiable information for grounding. Those recommendations point in the same direction. Say less “we are innovative.” Show more evidence of the work.

The website should not make machines infer basic relationships

A person’s bio should not merely say:

“John has 15 years of experience.”

• Experience doing what?
• For which organization?
• In what role?
• At which location?

A location page should not say:

“Visit our Scottsdale office.”

• Which business?
• What services?
• What address?
• What phone?
• Who works there?

Humans fill gaps from context. Machines can too. That is not a reason to make them. Explicit facts are cheap. Use them.

External sources still matter

Your website is first-party evidence. It is not the only evidence.

Depending on the business, useful external sources can include:

• professional associations,
• government records,
• industry directories,
• review platforms,
• news coverage,
• academic or research citations,
• partner pages,
• conference bios,
• and authoritative business profiles.

These sources can help corroborate identity. They can also create contradictions. If the website says one office and every directory says another, the ecosystem is inconsistent. That is a maintenance problem.

Not an invitation to create twelve more citations.

Entities and local search are inseparable

A local business is a bundle of relationships:

• Organization
• Location
• Phone
• Hours
• Services
• People
• Service area
• Reviews
• Categories

The website is one representation. Google Business Profile is another. Bing Places is another. Directories and professional profiles add more. This is why local SEO is such a natural foundation for AI search.

It already requires maintaining distributed entity information. Our local SEO guide goes deeper into that distributed entity problem.

People deserve real pages when they matter

For professional services, person entities can be commercially important. Lawyers. Doctors. Consultants. Financial professionals. Architects. Subject-matter experts.

A strong person page can establish:

• full name,
• role,
• organization,
• credentials,
• education,
• licensure,
• practice areas,
• experience,
• publications,
• awards,
• and related content.

That gives machines and humans one canonical place to understand the person. A four-line team card is not always enough.

Location pages should represent real locations

A real office page can provide:

• business name,
• address,
• phone,
• hours,
• map context,
• services,
• people,
• parking/access information,
• and local details.

A fake location page created only to insert a city name does not represent a real entity. That distinction matters. Do not confuse local landing-page scale with local entity clarity.

Services are entities too, in a practical sense

You do not need to philosophize about whether “technical SEO” is ontologically real. You do need a clear canonical page explaining the service.

That page should connect to:

• the provider,
• relevant industries,
• supporting expertise,
• case studies,
• locations where applicable,
• and educational resources.

The web of relationships makes the offering easier to understand.

Structured data should mirror the visible model

If a page visibly describes:

• a person,
• their role,
• their organization,
• and their credentials,

structured data can help formalize those relationships. If the page does not contain that information, injecting a giant schema graph does not make the site clearer. Google requires structured data to accurately represent visible page content for supported features.

Bing similarly says structured data can support clarity but does not guarantee grounding or traffic. That is the right perspective. Schema is an encoding layer. The website still needs meaning.

Evidence should be close to the claim

Do not force the reader to trust marketing adjectives.

Instead of:

“Industry-leading performance”

give:

• the metric,
• date,
• scope,
• method,
• and source.

Instead of:

“Trusted by businesses nationwide”

show:

• real client evidence,
• case studies,
• reviews,
• or specific experience.

Not every claim needs a footnote. But strong claims benefit from support. This creates better human content and better source material.