Skip to content
ceeme

For AI agents

How does an AI agent know that what it says about your company is correct?

Today it does not: it reconstructs your company from scattered fragments and presents the result with as much certainty as a fact. The verified entity record turns that around. It is one machine-queryable file about your company in which every claim — name, address, services, opening hours, rating — carries its provenance: which source said it, by what method it was established, and how heavily that weighs. An agent then has nothing left to infer. It reads what is proven and sees what is not.

Measure your own company → Rather talk it through first

How heavily does it weigh who claims something?

Four methods, with the weight the code genuinely uses. The order is not a taste but a rule: if more than one source fills the same field, the highest method wins and the rest does not vanish — it stays on as a contradiction.

methodconfidencewhat it means
owner confirmed · human0.95You confirmed it at intake. A human who signs off is the strongest evidence there is.
your own JSON-LD · structured0.80Your own JSON-LD on your own site. Machine-readable and yours, but nobody else confirms it.
third party · thirdparty0.70Reviews, your Google profile, a directory, Wikidata. Independent, and therefore heavier than what you write yourself — but it can be out of date.
what AI said · inferred0.40What an AI engine said. Counts as UNPROVEN and stays that way until one of the three above confirms it.

Why does what AI itself says count least?

Because an engine saying something about you is not a source but a reader. It repeats what it found somewhere, often without saying where, and sometimes without it ever having been there. Recording that as fact builds a graph that confirms itself: the engine reads the record, the record confirms the engine, and nobody ever checked anything.

How does an agent query that record?

As JSON-LD, with standard schema.org for the fields that exist there and a namespace of our own for the evidence around it. That combination is deliberate: an agent that only knows schema.org simply reads your company; an agent that looks further sees per field who said it and how heavily it weighs. Nothing is hidden behind a key or a login.

What does ceeme not put in your record?

Nothing you have not confirmed and nothing we cannot point at. No estimated headcount, no inferred revenue, no sector we guessed from your texts. A record that guesses is exactly the problem it was meant to solve.

Frequently asked questions

Is this the same as putting schema.org on my site?

No, and that difference is exactly why this exists. Schema.org on your own site is one method of four, and it weighs 0.80 — machine-readable and yours, but nobody else confirms it. The record lays the other three alongside and says per field which one won.

Who can read my record?

Anyone, and that is the point. A record behind a login is invisible to precisely the agents it was made for. What goes in it is your call — nothing enters without your confirmation.

What happens when a source changes?

Then the claim changes with it and the record holds on to both values, with the date. A record that silently overwrites loses exactly the information you keep it for: that something changed, and when.

Read on

Free, no account and no card.