ProductivityBy Serchai ·
How to value properties with AI: comparables, range and honesty
Guide to property valuation with AI: building the comparables base, calculating the reasoned range and presenting it without promising magic numbers.
Tools you will use
Stack: From $15/moKagi
Free trial · from $5The paid search engine with no ads and no tracking.
Read the reviewGamma
From $10Turns a text or a topic into presentations and documents that look good.
Read the reviewTLDR: Orientative valuation with AI is a well-run comparables process: gather the area’s real sales (closings, not listings), adjust for differences with explicit criteria, calculate a reasoned range instead of a magic number and present it in a dossier that shows the reasoning. Kagi is where those closings get found without portal noise, the adjustments live in your spreadsheet and Gamma builds the dossier. The limit gets declared from the title: this orients sale prices, and official appraisal (mortgages, inheritances, litigation) is certified appraisers’ territory.
This guide is for agents preparing prospecting valuations and for owners wanting to understand their home’s worth before talking to anyone.
The limits note, first: this guide’s valuation is the market-orientative one, the one that sets sale prices. Official appraisal with legal effects (mortgage, inheritance, divorce, litigation) gets signed by a certified appraiser, and no AI replaces it.
1. Gather real comparables, not asking prices
The error inflating half of all valuations: using asking prices as reference. The listing is what someone wishes to collect. The comparable that counts is what someone paid: recent closings of similar properties in the area, gathered by combining your own operations (every sale of yours is gold), official statistics by area (registries and portals publish aggregated closing data) and disciplined market watching: the listing that disappears after two price cuts tells a complete story.
The searching part is where a tool genuinely helps. Kagi returns those sources with no ads and no content farms, which in a housing-price query is exactly the noise you do not need: official statistics and the portals publishing aggregated closing data surface ahead of the pages repackaging listings. Its domain controls also let you pin your market’s good sources so you stop hunting for them every quarter. From $5 a month with a free trial.
The minimum standard for serious valuation: five or six comparables from recent months, same area and similar typology, with their data noted (surface, floor, condition, initial price, final or estimated closing price, time on market).
Time on market is the datum almost nobody checks and that informs most: the home “sold at asking price” after eleven months tells the opposite of what it seems.
2. Adjust for differences with explicit criteria
No comparable is identical, and there the adjustments enter: the comparable’s price per meter gets corrected for what the valued property has more or less of. The ones that truly move price: condition (renovated versus to-renovate is the big jump), floor and elevator, orientation and light, exterior versus interior, terrace, parking and the high community fees that subtract.
This step belongs in your spreadsheet, not in a chat, for two reasons worth saying out loud. A valuation’s numbers get checked one by one, and an assistant that writes well and calculates unevenly is the worst possible tool for a table you will defend in front of the owner. And how much a fourth floor without elevator subtracts in that specific area is your local knowledge, not general statistics you can ask for.
The step’s result is a defensible table: each comparable, its reasoned adjustments and the adjusted value it contributes to the valuation.
3. Calculate the range, not the magic number
Markets do not give exact figures: they give ranges, and the honest valuation reflects it. From the adjusted comparables table comes the range: the floor (at this price it sells fast), the center (the reasonable starting price) and the ceiling (defensible only if the market cooperates and there is no rush).
The range gets completed with the time variable, which is the conversation the owner needs: starting at the ceiling usually costs months and cuts that end below the center, while starting at the center with good marketing generates the demand tension that sometimes beats the starting price. Your area’s data (average cut, time per price band) turns that conversation into numbers.
What the range is not: a promise. Markets move, every property has its story and valuations expire: the six-month-old one is history, not reference.
4. Present the valuation so it gets understood and believed
The valuation competes against the inflated figures of promise-based prospectors, and its weapon is visible reasoning: the dossier showing the comparables, the adjustments and the range with its logic convinces where the bare figure only competes upward.
Gamma (free credits to start) builds the dossier in half an hour: cover with the property, the comparables table, the reasoned range, the sales plan and the expected times per price band. The document does double work: it convinces the serious owner and filters out the one who will only sign with whoever promises most, which is exactly the client not worth having.
The presentation conversation has its script: first the method (where the data comes from), then the comparables (their street, real cases), then the range, and the starting price as a shared decision on the table. The whole sector lives in AI for real estate.
Common mistakes
Valuing with asking prices. What is listed is desire, what closed is market: the listings-based valuation inherits the inflation of every optimistic seller in the area.
The exact figure without a range. “Your home is worth 237,000” is false precision: markets give ranges, and the reasoned range is more honest and more defensible.
Adjustments from memory. The no-elevator discount or the renovation premium gets noted with its why: adjustments without explicit criteria cannot be defended to the owner nor repeated on the next home.
Promising to prospect. The inflated valuation wins the signature and loses the sale: the overpriced home burns in the market and the cut arrives late and worn. The honest range loses some signatures and sells the homes it signs.
Frequently asked questions
Are portals’ automatic valuations any good?
As a first rough reference, yes: as a valuation, no. They work on listing data and area averages, without seeing the condition or the details that move price. They are the starting point your comparables work corrects.
How many comparables do I need?
Five or six from recent months in similar area and typology is the practical standard. With fewer, the range comes out wide and that gets said. In low-turnover areas the radius or period gets widened, declaring it.
Does this valuation work for the bank or an inheritance?
No: mortgages, inheritances and litigation require an official appraisal signed by a certified appraiser. This valuation sets sale prices and expectations: different documents with different jobs.
How fast does a valuation expire?
In a normal market, within a few months. In a moving market, sooner. The practical rule: if the comparables are over six months old, the valuation needs new comparables, not an eyeballed adjustment.
The steps, in short
Gather real comparables, not asking prices
What sold and what it closed at counts: what is listed is what someone wishes, not what the market pays.
Adjust for differences with explicit criteria
Floor, condition, orientation and extras move the price: every adjustment with its written why.
Calculate the range, not the magic number
Markets give ranges: the honest valuation is a range with reasoning.
Present the valuation so it gets understood and believed
The dossier with comparables and reasoning convinces where the bare figure only competes.
Related guides
How to work with your accountant in the AI era (and pay for judgment, not typing)
Guide to the new division with your accountant: ordered data with Holded and Dext, documented…
Updated July 25, 2026How to do bank reconciliation with AI in one sitting a month
Guide to reconciling the bank with AI: connected accounts, assisted matching with Holded and the…
Updated July 25, 2026How to build your SMB's annual budget with AI (and keep it useful all year)
Guide to the annual budget with AI: start from real data, build by scenarios and the monthly…