This week, AI ran into the real world. First the law: in New Zealand, a law firm points out that advertisers remain liable for a misleading photo even when a provider edited it, and Pennsylvania now requires a label on AI-altered ads. Then buyers, with a Canadian agent who doubts virtual staging is the future. Agencies, where AI is spreading faster than working practices. Property management, with one lease in thirty in arrears. And neighbors, whom xAI is offering to buy out if they drop their lawsuit against its data center.
A week where AI runs into the real world
For several weeks, this watch has followed the same question: how far can AI go in a property listing? This week, the answer no longer comes only from buyers or consumer groups. It comes from lawyers, lawmakers, agents themselves and, more unexpectedly, from people living next to data centers. AI remains a powerful tool, but everywhere it meets the same realities: liability, organization and location.
Housefishing: the law takes on edited photos
English speakers have coined a word for photos that make a home look better than it is:
housefishing. New Zealand law firm Simpson Grierson analyzed the practice under the country's Fair Trading Act. Its conclusion is clear. Fixing the lighting or removing a trash can is fine. Adding a renovated kitchen, hiding a defect or shrinking furniture to make a room look bigger can be misleading, even without any intent to deceive. Courts look at the overall impression of the listing. The advertiser remains liable
even if the images were produced by a provider, and a generic disclaimer at the bottom of the page doesn't fix a misleading impression.
The same week, Pennsylvania passed a law requiring a label on ads generated or significantly modified by AI, when they could create a false perception of the product. According to law firm Holland & Knight, it is set to apply from December 21, 2026.
Last week we covered the same rules in Canada and Belgium. The New Zealand firm adds a list of good practices that applies just as well elsewhere. Keep the original photos and check the final images against the actual state of the property. Place the disclosure on the image or right next to it, and say what was changed. That's the principle behind our
AI charter for property listings.
Sources: Simpson Grierson, "'Housefishing' - is using AI in a property listing misleading under the Fair Trading Act?", October 2026; The Verge, "Pennsylvania will require labels on AI-generated ads", October 2026; Holland & Knight, "Did AI Touch Your Ad? Pennsylvania May Require You to Say So", October 2026.Virtual staging doesn't convince everyone
In an opinion piece for Canada's Real Estate Magazine, Toronto agent Nik Lunin explains why he doesn't believe virtual staging will replace real furniture. Worth noting: he also runs a furniture rental company for staging. His main argument still deserves a hearing. Disclosing that a photo was virtually staged protects buyers, but it doesn't remove the gap between the image seen online and the empty room found at the viewing. In his view, the easier it becomes to produce perfect images, the more a verifiable physical experience is worth.
He doesn't reject the technology, though. He finds virtual staging useful to show possible layouts or to make the most of a difficult space. He also likes the idea of glasses that would furnish an empty room at its real dimensions. For agents, the lesson is practical. A furnished image must respect the real volumes, with furniture at the right scale. Otherwise the viewing becomes a letdown, and that letdown turns against the listing and the agency alike.
Source: Real Estate Magazine, "Everyone says virtual staging is the future. I am not convinced.", October 2026.
AI in agencies: the appetite is there, the organization isn't
In a column for French outlet Journal du Net, consultant Florian Berthoud cites figures from France's statistics office, Insee: 26% of real estate companies with 10 or more employees reported using at least one AI technology in 2025, up from 14% in 2024 and 7% in 2023. The main obstacle cited is lack of expertise, by more than half of companies. In the agencies he works with, AI remains mostly an individual habit, sometimes on free accounts, which raises a real problem for client data.
That point stands out this week. Google changed its privacy rules: when users connect Gmail or Google Photos to its search engine's AI features, that data can now be used to personalize ads. An agent who pastes an exchange with a seller into a consumer assistant doesn't always know where that information ends up.
The column's recommendations are simple. Start from a task repeated every week, like the viewing report, rather than from a tool. Then write a short usage rule listing the approved tools, on business accounts, and the client data that's off-limits. Finally, appoint a point person and measure the results before rolling the method out to the whole team.
Sources: Journal du Net, "AI in real estate agencies: the appetite is there, the change is still to come", October 2026; The Verge, "Now Google says it might use your Gmail and Google Photos data to target ads", October 2026.
Property management: one lease in thirty in arrears
A property management barometer published by Rivage and reported by Journal de l'Agence shows that more than one lease in thirty has at least one month of unpaid rent. The median arrears reach €1,998, about 2.9 months of rent including charges. The same barometer finds nearly 10% of managed homes rated in France's two worst energy classes, and 21% in Paris, including 4.3% in the lowest class. France's national bailiffs' chamber recorded 175,000 payment orders and 30,500 evictions in 2025.
These figures show where property managers' real workload lies: reminders, arrears files, energy renovation works to schedule before rental bans kick in. These repetitive, administrative tasks are where AI can save time. Applicant screening, on the other hand, remains bound by anti-discrimination rules, as the French government reminded agents in September.
Source: Journal de l'Agence, "Property management: 1 lease in 30 in arrears, nearly 10% of energy sieves", October 2026.
xAI offers to buy homes in exchange for dropped lawsuits
In Southaven, Mississippi, a subsidiary of Elon Musk's xAI is offering to buy homes in the neighborhood next to its data center. The offers are conditional on dropping claims linked to a lawsuit filed in June by residents. They complain about near-constant noise and vibration from the gas turbines powering the site, and seek compensation for lost property value among other things. A recording cited in the press mentions 418 homes targeted, a figure xAI hasn't confirmed. The company denies the nuisance claims.
The story is far from France, but it says something concrete about our trade. AI infrastructure isn't immaterial. It takes up land, consumes energy and water, and makes noise. Wherever it settles, it ends up weighing on the value of nearby homes. Data center projects are multiplying in France too: the day one of them lands near a neighborhood, the question will come up at viewings.
Sources: The Verge, "xAI is buying out homeowners who promise not to sue over data center noise", October 2026; Forbes, "Elon Musk's xAI Offers Buyouts to 400 Mississippi Homes to Quash Data Center Lawsuit", October 28, 2026.
The thread this week: AI runs into the real world
A law firm reminds us that liability for a photo can't be delegated. An agent reminds us that a disclosure doesn't replace a successful viewing. A column reminds us that a tool without a method stays an individual habit. And residents remind us that AI has an address. Technology moves fast, but the realities it meets stay the same: the law, the organization and the ground.
So: does your agency have a written rule on which AI tools to use, and with which data?
