About evaluatorAI
Who runs it
evaluatorAI is built and run by 1590987 B.C. LTD., a company registered in Vancouver, British Columbia.
We are not a portal, an agency or a dealer. We do not sell property or vehicles, and we do not take a cut when you buy one. Nobody pays to appear in an evaluation, there are no referral fees from agents, dealers, brokers, surveyors or lenders, and we carry no advertising. The only money we take is from people who buy credits.
Reach out to us for support, questions or partnerships at: [email protected].
Where we use AI and the rules and guardrails we apply
The AI reads one thing: the listing data we send it. Every field the site discloses, and every word the seller wrote.
An AI model writes the evaluation. What makes it worth reading is not the model. It is the list of things the model is not permitted to do.
It may only use fields the listing actually contains. Designations and statuses are the sharpest case: whether a property sits in a conservation area, whether a building is listed, its flood zone, its school catchment, a vehicle's inspection status. If the listing does not state it, the evaluation does not state it either, however obvious it might look from the address.
The same holds for numbers. No comparable sales, no invented running costs, no commute times, no figure presented as live market data. Any comparison comes from a figure the listing itself carries, such as the site's own price rating, and it says whose figure it is. It may not tell you from memory that an area holds its value, or that a price sits below what similar ones fetch. Those are the same unverifiable comparison, without the number.
The description is the seller's marketing copy and we treat it as exactly that.
We do not hold a database of sold prices, crime figures or flood maps.
The rules are written into the prompts, real evaluations are replayed against them every time we build, and every assessment we generate is checked in code against the rules its own prompt states.
We've built our prompts to inspire trust
What usually goes wrong when an AI reads a listing is not that it fails. It is that it produces something confident and plausible in the space where it should have said nothing at all.
An evaluation is built to leave that space empty. Where the basis for an observation is not verifiable, it is omitted rather than hedged with a disclaimer.
Where something matters and the listing does not answer it, you get a question to put to the agent or the seller instead. Same information, in a form you can use.
What evaluatorAI is not
We are not a surveyor, a solicitor, a mortgage adviser or a mechanic, and an evaluation is not regulated financial, legal or tax advice.
We are not affiliated with any of the sites we support, and none of them endorse us.