About Adresse.ai

Why we built this and how we think about the product.

Updated May 2026.

Why this exists

Adresse.ai was inspired by one family’s frustration and confusion with the French property-buying process. They hit the same wall every anglophone buyer hits: French property data is technically open, but practically inaccessible.

DVF, the French government’s transaction register, contains every property sale since 2014. It’s free, public, authoritative. It’s also published as French-language CSV files with a six-month lag and no analytical layer. ADEME publishes every energy-rating certificate. Same situation: open, free, and in a format no normal buyer is going to query directly.

The Zestimate-shaped tool that exists in the US and the UK simply doesn’t exist for France. MeilleursAgents is built for sellers, in French. Patrim is the official government tool but requires a French tax account most pre-purchase buyers don’t have. PriceHubble is excellent technology that’s only sold to banks. The English-language editorial sites publish guides, not numbers. So the buyer asks ChatGPT or scrolls Reddit, gets a wishy-washy answer, and ends up trusting the asking price and the agent’s word on a half-million-euro decision.

That’s a bad equilibrium. Adresse.ai is the missing tool.

How the product is designed

Three commitments shape every decision:

Buyer-side only. No agent ever pays Adresse.ai. No seller ever pays Adresse.ai. No referral fees, no sponsored placements, no “preferred partner” arrangements that affect what comes first in a report. The product exists to give buyers a defensible second opinion, and the incentives stay aligned only because no one outside the buyer relationship is paying.

Methodology is visible. Every estimate shows the comparable sales used, the adjustments applied, the math, and the assumptions. If a buyer disagrees with an adjustment, the methodology page tells them how to recalculate. The opacity that makes other tools hard to argue with is exactly what Adresse.ai avoids.

Honest about limits. When the comp pool is thin, the report says so. When the listing has features the model can’t price, the review names them. When the data lags six months, the report surfaces it. A confident number with no uncertainty band is dangerous; flagging uncertainty beats faking precision.

The buyer reading the report is not an idiot. They’re an intelligent person operating in an unfamiliar market with their normal toolkit removed. The product is built for that reader.

What it isn’t

A few things worth being explicit about:

Not a brokerage. Adresse.ai doesn’t list properties, take a cut of any transaction, or run a “buy through us” button. It helps you decide; you and your notaire execute.
Not a buyer’s agent. A chasseur immobilier is a real value for some specific cases (high budget, repeated property hunting, or buyers with complex needs). Adresse.ai is the analytical tool, not the human advocate.
Not a French lawyer or notaire. Reports inform the conversation; a notaire still closes the transaction.
Not trying to cover all of France immediately. The model needs region-specific tuning, and the South of France well beats France-wide poorly. Other regions follow as the tuning lands.

Try it on your listing

Try Adresse.ai on your listing.

Run an estimate →

See also:

Adresse.ai. Property valuations for the French market. Hosted in the EU.