2026-08-03 · Jack Stovell

How to become an entity AI models recognise

An entity, to an AI system, is a thing it can identify consistently across sources: the same person, with the same facts, wherever it looks. You become one by supplying that consistency deliberately — one canonical URL, the same one-line bio everywhere, structured identity markup, and corroboration from pages you don't control. There is no application form and no registry to join; there is only evidence, accumulated until the machine stops treating your mentions as strangers.

What does it mean to be an entity to an AI model?

It means the system can tell that a mention of your name on one page and a mention on another refer to the same thing, and can attach facts to that thing with some confidence. Search engineers call this entity resolution. When it works, an assistant asked about you retrieves several pages, recognises they describe one person, and merges them into a coherent answer. When it fails, you get the familiar symptoms: the model blends you with a namesake, attributes someone else's job to you, or answers "I don't have information about this person" while your website sits unread. If you're being merged with the wrong person, that failure mode has its own fix — see when AI confuses you with someone else.

How do you become a recognised entity?

By giving machines consistent, corroborated evidence that all your mentions are one thing. In practice that comes down to five layers:

One canonical URL. A single page that is unambiguously the page about you, which everything else links back to. Five half-maintained profiles dilute; one maintained page concentrates.
The same one-line bio and facts everywhere. Same name form, same role, same location, same links — on your site, your GitHub, your directories, your profiles.
Structured identity. Schema.org Person JSON-LD on your canonical page, with a sameAs array naming your other profiles. This is the machine-readable statement "these URLs are all me".
Presence in structured sources. Databases that machines already parse: Wikidata (which has a notability bar — see how to add yourself to Wikidata), GitHub, ORCID for researchers, professional registers where your field has them. Some are open to anyone; some are not, and pretending otherwise is how bad advice gets written.
Corroboration you don't control. Pages by other people — an event listing, a client's site, a podcast page, a published paper — that state the same facts. This is the strongest layer precisely because you can't fabricate it.

The sameAs pattern is the one piece worth copying exactly. This is the fragment that declares one entity across every surface:

<script type="application/ld+json">
{
  "@context": "https://schema.org",
  "@type": "Person",
  "@id": "https://janeexample.com/#person",
  "name": "Jane Example",
  "jobTitle": "Freelance data engineer",
  "url": "https://janeexample.com/",
  "sameAs": [
    "https://github.com/janeexample",
    "https://www.wikidata.org/wiki/Q00000000",
    "https://i.hilyt.it/janeexample",
    "https://www.linkedin.com/in/janeexample"
  ]
}
</script>

Every URL in that array should carry the same name and the same basic facts. The full property set, and what models actually consume, is covered in our Person schema walkthrough.

Why do consistent bio details matter for AI recognition?

Because consistency is the machine's evidence that two mentions are the same person. Entity resolution runs on overlapping facts: same name plus same employer plus same city is a strong match; same name alone is a weak one. Local SEO people have known this for years as NAP consistency — name, address, phone — for businesses. The personal equivalent is name form, role, location and links. Every profile where your title is two jobs out of date, or your name appears in a different form, weakens the match and adds noise to the merge. Auditing your surfaces into agreement is unglamorous and it is genuinely the highest-leverage hour in this whole exercise.

What sources do LLMs trust about people?

Honestly: nobody outside the AI labs knows the weightings, and anyone who claims a ranked list of trusted sources is selling something. What we can say from observation is that widely-crawled structured sources — Wikipedia, Wikidata, GitHub, major news — surface disproportionately often in model answers about people, and that facts corroborated across several independent pages surface more reliably than facts stated once. We publish our own citation results, including the misses, at hilyt.it/geo, and the pattern there matches: readability and corroboration are necessary, and nothing is sufficient. The mechanics of source selection are explored further in how AI decides which sources to cite.

Does Crunchbase or Wikidata affect what AI says about you?

They can, because both are structured, heavily crawled, and reused downstream by other databases — which multiplies the corroboration. But be honest about access. Crunchbase accepts self-submitted profiles, though it skews towards startup and company roles. Wikidata is not "add yourself and done": items need serious public references or a structural need, and self-created items about people who don't qualify get deleted. GitHub is open to anyone and is one of the most heavily crawled identity surfaces on the web. None of these is a magic entry; each is one more independent, machine-parseable statement of the same facts.

Can you get into Google's Knowledge Graph?

Not by applying — there is no application, and Google's inclusion criteria are opaque. Knowledge Panels appear when Google's systems are confident an entity exists and is being searched for; you can claim an existing panel, but you cannot order one into existence. The influenceable part is exactly the work above: a canonical page, consistent facts, structured markup, corroborating sources. That is also the honest limit of this whole guide — entity-building removes the reasons systems can't recognise you, and after that, recognition depends on retrieval, ranking and model behaviour that nobody outside the labs controls. You can check how much of the groundwork is in place with our free AI visibility checker.

Create your AI-readable profile