AI & Real Estate: 5 Signals to Watch This Week (Week of September 28, 2026)

A buyer can now ask an AI about an energy rating, a neighborhood or a price per square meter at any hour. That doesn't mean they decide faster. This week, several stories show text and voice losing their value as proof, while what can be seen and checked gains it. The agent's role shifts accordingly.

In an opinion piece for Journal de l'Agence, consultant Peter Jørgensen describes a paradox many agents already see. Getting one more piece of information now costs nothing. A buyer wondering about the energy rating asks AI about renovation costs, then their effect on value, then available subsidies. Each answer opens a new question, and the decision slips further away. He also notes a shift in trust: agents increasingly hear "the AI tells me something different." His recommendation is to move from information provider to decision facilitator. That means showing your reasoning rather than asserting a conclusion, and explaining why two comparable properties don't carry the same value. It also means helping clients tell real risks from theoretical ones, and knowing when to say they have enough to decide. It's an argument for visual material too: a clear floor plan answers in a second questions that three pages of text leave open. Source: Journal de l'Agence, "AI answers everything, but buyers hesitate: how to help them decide?", 2026.

AMD announced it is acquiring World Labs for about $8.2 billion in stock. The lab founded by Fei-Fei Li builds so-called spatial intelligence models. They generate, reconstruct and simulate 3D environments from text, images and video. Fei-Fei Li will join AMD as chief scientist. The deal is expected to close by the end of 2026. It's the second-largest acquisition in AMD's history. The price shows where the industry places AI's next step: after text and images, understanding physical space. For real estate, the medium-term consequence is concrete. Reconstructing a home in 3D from a few photos will become increasingly common. The difference will then come down to fidelity to the actual property, not the ability to produce a nice scene. Sources: The Verge, "AMD is acquiring AI company World Labs in a deal worth more than $8 billion" and Ars Technica, "AMD acquires World Labs AI startup, upping the ante against Nvidia", 2026.

Since October 1, ChatGPT offers a "Try on" button: users upload a photo of themselves, and the AI shows how a piece of clothing or an accessory they're considering would look on them. The product doesn't even need to be sold in ChatGPT: a screenshot from any store works. The feature targets fashion, but the habit it builds goes further. Millions of people will get used to visualizing a purchase on their own reality before deciding. For a buyer, the next question comes naturally: what would this living room look like with my furniture, or with a new floor? Agencies already offering a projection faithful to the property are answering an expectation that is about to become standard. Source: TechCrunch, "ChatGPT can now virtually try on clothes for you", 2026.

A study by Graphite, reported by TechCrunch, catalogs 13,000 phrases at least twice as common in AI-generated text as in human writing. Some are far more frequent: in Claude Opus 5.5, "this matters" appears 116 times more often than in human text. According to the researchers, models fix the best-known tics, but each new version introduces others. For a property listing, the lesson is practical. A description generated and published without editing can be recognized, especially when every listing in an area relies on the same tools. Readers won't always know why, but they'll sense interchangeable copy. The right approach hasn't changed: let AI suggest a structure, then rewrite it with what only you saw during the viewing. Source: TechCrunch, "Opus 5.5 loves to tell you 'this matters' (and other AI writing tells)", 2026.

Details of a fraud that took place in Italy in February were made public this week. The chairman of Fideuram, the private bank of the Intesa Sanpaolo group, received a WhatsApp message apparently from the group's CEO. He then received a call using an AI imitation of the voice of a lawyer he knew. The result: a series of transfers abroad totaling €95 million, of which around €36 million is still missing. No computer system was hacked. Real estate is particularly exposed to this kind of scenario. A transaction involves large sums, several parties who barely know each other, and the exchange of bank details. A fake notary calling about a "change of bank details" no longer needs to forge a signature: a credible voice is enough. The rule is simple and belongs in every agency's procedures. Any payment instruction or change of bank details is verified by calling the person back on a number you already have, never the one given in the message. Source: Numerama, "A fake AI-generated lawyer's voice costs a major Italian bank €95 million", 2026.

Text can be generated, a voice can be cloned, an AI answer can confidently contradict the agent. What keeps its value is what can be verified: a measured space, a projection faithful to the property, a callback to the right number. For an agent, the job is no longer to supply information, but to help clients find their way through it and decide. So: in your last sale, how many of the buyer's questions came from an AI answer rather than from the viewing?