How Adresse.ai works

Updated May 2026.

If you’ve ever pasted a French property listing into a search bar and felt the absence of a Zestimate, this page is for you. We’ve built the tool that fills the gap. The data sources are public, the methodology is transparent on every report, and the assumptions are named.

What Adresse.ai does

You paste a French property listing. We extract its details, find recent comparable sales from the French government’s official transaction register, adjust for the things that move price, and produce a defensible price range with a qualitative review. Around forty-five seconds end to end.

The single sentence version: we make French government property data readable, and we apply comparable-sales analysis in English, in under a minute.

Step 1: paste your listing

You paste a URL from SeLoger, LeBonCoin, Bien’ici, an agency site, or any of the major portals. You can paste a PDF or plain text from a printed brochure. We extract the property type, surface, location, energy rating, and standout features the listing mentions.

If the listing is missing something important (like the DPE rating or habitable surface), we tell you and ask for it before running the analysis.

Step 2: comparable sales from DVF

Comparable sales come from DVF (Demande de Valeurs Foncieres), the French government’s mandatory register of every property transaction recorded by a notaire. The dataset covers transactions from 2014 forward across metropolitan France and is refreshed twice a year. Source: data.gouv.fr.

We filter DVF to find recent sales of similar properties near your target. The output is a comp pool of actual transactions, each with a verified sale price and date.

If the comp pool is thin, we tell you. We don’t pretend a handful of transactions is a fair test.

Step 3: adjustments

A raw price-per-square-metre median is misleading. Two houses in the same village can sell for prices 40% apart based on condition, energy rating, whether one has a pool, and whether the stone facade is original or rendered.

Before we produce a number, we adjust each comparable to account for the differences between it and your target property. The adjustments cover condition, energy performance, and property features. Each adjustment is named and visible in your report. If you disagree with one, the report shows how the number changes without it.

The adjustment factors are tuned per region against actual sale data and updated quarterly.

Step 4: the qualitative review

The comp engine produces a number. The qualitative review checks whether that number deserves your confidence.

A second AI pass re-reads the listing, the comp pool, and the result, then flags anything that needs attention. Common flags include thin comp pools, outlier comparables pulling the median, features the model couldn’t fully price (like a large outbuilding or guest house), and patterns that suggest renovation cost or pricing issues.

The verdict is a one-paragraph summary in plain English: whether the asking price is fair, stretched, or a bargain, and why. That verdict is your negotiation ammunition.

Step 5: the report

Every report shows:

The adjusted price range with the median estimate
A confidence indicator based on the strength of the comp pool
The comparable sales, each linked to the public DVF record so you can verify
The adjustments applied and their rationale
The qualitative review verdict
A shareable read-only link you can send to your partner, accountant, or agent

When confidence is low, we say so in plain language. If your target is a rare property with few good comparables nearby, that is a real limit, and we surface it instead of pretending around it.

What we don’t do

We’re a decision-support tool. We don’t:

Sign anything for you. A French notaire is still legally required to close any property purchase.
Replace a private survey. The mandatory diagnostic file (DDT) is not a UK-style survey, and we recommend commissioning one for any property you’re seriously considering.
Know about the neighbours. We can’t tell you about planned developments or disputes next door.
Price the lifestyle. Whether the village is right for you is not a number we can produce.
Take a fee from any side of the transaction. No agent or seller pays us anything to influence your estimate.

Where the data comes from

Source	What we use it for	Update cadence
DVF (data.gouv.fr)	Recent property sale prices	Twice a year (April, October)
ADEME DPE register	Energy ratings for individual properties	Continuous
Notaires de France	Regional price benchmarks and trends	Quarterly

DVF lags real time by roughly three to six months because notaires record transactions retrospectively. That lag is structural. We compensate by weighting recent sales more heavily and by surfacing the date range of your comp pool in the report.

Questions

Is Adresse.ai accurate?

A comparable-sales analysis on adjusted DVF data with a sufficient comp pool produces a price range that experienced local agents typically agree with. On thin comp pools or unusual properties, the range widens, and we tell you. We don’t claim a single accuracy number because doing so on rural French property would be misleading.

Can I trust DVF when it lags?

For evaluating a specific property against recent evidence, yes. For catching the very latest market shift, no tool gets you faster data than what the notaires publish. We show the lag explicitly.

What if my listing is in an area you don’t cover well?

We currently cover the South of France with region-specific tuning. If your target sits outside well-covered areas, the qualitative review will say so, and the confidence band will widen. Regions are added one at a time as coverage and tuning land.

Who built this?

A small team that bought French property and walked into exactly the data gap the product addresses. We use the tool ourselves.

Try it on your listing

Run an estimate →

See also: