Real estate valuation, long a monopoly of professionals, has entered a new era. Thanks to tools like ChatGPT, individuals can now obtain an initial evaluation of their property in just a few minutes, drawing on considerable volumes of data. An evolution which profoundly changes the market balanceé, since the information, formerly held by agents, becomes accessible to everyone.
A recent example illustrates this shift: in Florida, a seller explained to the People.com site that he managed to sell his house in five days by relying largely on artificial intelligence, without an agent on the seller’s side. The good would even have been sold for almost 100,000 dollars (around 85,000 euros) above initial estimates, a spectacular case which shows that certain steps can now be taken care of without an intermediary.
In this context, the estimation differences can reach significant levels: up to €85,000 difference between an estimate from AI and those proposed by several agencies. Something to question, especially when the fees represent 3% to 6% of the sale price, or several thousand euros.
Real estate valuation with ChatGPT: why AI can be more precise
One of the main advantages of artificial intelligence lies in its ability to analyze a massive volume of data in a very short time. It notably crosses recent transactions from the DVF database, online advertisements, the characteristics of the property (surface area, floor, exposure), as well as territorial indicators.
In the most liquid markets, the margin of error of the most advanced models would be less than 3%. A level of precision made possible by an analysis capacity inaccessible to a human in real time.
But in the field, this approach quickly shows its limits. “It’s very mathematical”observes Ana Moreno, real estate agent at Orpi in Fleurance (Gers). “The AI is based a lot on the price per square meter, but it does not sufficiently take into account essential elements such as the layout, the real condition of the property or the neighborhood” she insists.
A more objective estimate than that of a real estate agent?
Unlike a professional, ChatGPT does not pursue any commercial objectives. He is neither seeking to obtain a mandate nor to speed up a sale. Artificial intelligence produces an estimate based only on available data, which limits certain biases. An agent may be tempted to overvalue a property to attract a selleror on the contrary to offer a lower price to secure a quick transaction. AI compares and calculates without intention.
But this neutrality remains entirely relative. “ We can also make him say whatever we want »underlines Ana Moreno, who continues: “If an owner presents his property as perfect, the AI will rely on these elements, while the reality may be very different on site. » Result ? The differences can be significant. “This is not a representation of reality at all. It can give a basis, butif it is necessary to refine behind »insists the professional.
Real estate agency fees: are they still justified compared to ChatGPT?
Faced with these performances, the question of fees becomes central. In France, they generally vary between 3% and 10% of the price of the propertydepending on the networks, markets and services offered. Traditional networks like Orpi, Century 21 or Laforêt are most often located in the middle part of this range. The annual study by the National Union of Real Estate Professionals reports an average drop of 11% in fee rates between 2020 and 2025. But in absolute value, the cost remains high: a commission of 7% on a property sold for €400,000 represents €28,000. However, agents do not yet sense a massive revolution on the ground. “For the moment, I have never had a client who arrives with an estimate made by ChatGPT”confides Ana Moreno.
What ChatGPT AI can do alone today in real estate
The impact of ChatGPT and other AI is not limited to estimation. He is now involved in a large part of the real estate process. AIs know how to write optimized ads, analyze technical documents and synthesize complex information in seconds. According to several sectoral analyses, artificial intelligence could automate a significant part of real estate agents’ tasks. Firms like McKinsey & Company or PwC estimate that up to 60% to 70% of activities in service professions can be partially automated, in particular those linked to data analysis, writing or processing of information. And in real estate, this transformation is already visible. Studies carried out by
What ChatGPT does not yet replace: negotiation and the field
Despite these advances, certain steps remain profoundly human. Real estate negotiation is based on understanding emotions, managing objections and adapting in real time. “ A sale is not just a price », recalls Ana Moreno. “You can’t compare two houses just on data. A recent house and a stone house, for example, have nothing to do with each other. » Market context also plays a key role. “Today, the market is tight. We can no longer be satisfied with a price per square meter »she insists.
Legal security is also a key point: drafting the compromise, suspensive clauses, coordination with the notary… so many steps which imply a responsibility that the AI cannot yet assume.
Finally, atypical goods or poorly documented markets remain difficult to model. AI, which works by comparison, shows its limits as soon as references become rare.
