A Winkworth estate agency in Tooting, south London, listed a property this spring with photographs that showed a spacious, well-proportioned living room. One buyer drove across the city, walked through the front door, and discovered a chimney breast the listing photos had erased. The agency pulled the images "after a customer expressed disappointment during a viewing," a spokesman told The Times. The AI disclosure, it turned out, had been buried in white text on a white background inside a PDF brochure, invisible to anyone who did not think to search the underlying text layer.
That vanishing chimney breast sits at the center of a data point that should concern anyone selling, buying, or building homes in the United States right now.
Trust Didn't Erode. It Cratered.
Cotality surveyed prospective homebuyers across the U.S. and found that confidence in AI as a home-finding tool has been nearly halved since 2025. The numbers are stark, and the demographic breakdown is worse than the headline suggests. Only 9 percent of Baby Boomers and 11 percent of Gen X say they tolerate AI errors in the homebuying process, and even among Gen Z buyers, who grew up swiping through algorithmically curated feeds, only 22 percent call AI mistakes acceptable when the stakes are a six-figure purchase with a 30-year tail.
Meanwhile, 70 percent of Realtors have already used AI tools in some capacity, according to the National Association of Realtors. The industry adopted faster than its customers trusted, and that gap is not closing but widening, because the most visible consumer-facing use of AI in real estate has been staging fraud: adding furniture to empty rooms, brightening interiors, smoothing over cracks, and occasionally deleting structural elements from photographs to make a property look like something it is not.
Sixty-four percent of buyers in the Cotality survey expressed concern that AI tools may "recycle" unverified information rather than use validated, first-party data. Nearly half said it is unacceptable for lenders or insurers to conduct automated AI valuations without prior approval. When University of Chicago researchers studied AI-generated listing content, they found it triggers measurable buyer distrust, with both sides of the transaction losing efficiency as a result.
Two States Drew Lines. Forty-Eight Did Not.
California's Assembly Bill 723, effective January 1, 2026, is 580 words long and remarkably clear for a piece of state legislation. If you are a real estate broker, salesperson, or anyone acting on their behalf, and you include a digitally altered image in an advertisement for the sale of real property, you must disclose it. The statement must be "reasonably conspicuous" and "located on or adjacent to the image." You must provide a link to the original, unaltered version. Enforcement mechanisms include regulatory discipline, civil liability, and criminal charges for willful violations.
New York's Department of State issued its own trend alert warning homebuyers about AI-generated listing photos. Secretary of State Walter T. Mosley urged "real estate agents to be mindful when listing and consumers to be vigilant when searching for their dream home." The state is also advancing Bill S09584, which would prohibit digital alterations that "add non-existent physical features or remove or conceal known material defects" without clear disclosure, with enforcement authority granted to both the Department of State and the Attorney General.
Florida, Texas, Georgia, North Carolina, Arizona, Ohio, and the other 41 states without specific AI listing disclosure rules rely on existing prohibitions against misleading advertising. Those prohibitions exist, but whether a general unfair-business-practices statute written before anyone imagined software could delete a chimney breast from a photograph provides adequate enforcement is a question no court has yet answered.
The Tools That Got Punished for the Fraud They Didn't Commit
In Denver, a home inspection company called Alpine Building Performance launched a free AI tool in March 2026 that does something genuinely useful: it predicts what a home inspector is likely to find before the inspector ever arrives. Upload an MLS property sheet, and Alpine Intelligence generates a pre-inspection summary that flags era-based risks, system-by-system concerns, and talking points an agent can share with buyers at a showing.
"A lot of what you see in an inspection is really characteristic of when it was built and where it's located," founder Andrew Sams told 5280 Magazine. The tool was born during the pandemic buying frenzy, when agents were putting in offers $60,000 above asking after five-minute walkthroughs and calling Sams to ask if he could come to a showing because they had no idea what problems they might be buying. A pre-1940s home in Denver's Congress Park neighborhood carries an 87 percent probability of containing lead-based paint. A 1950s ranch in City Park will likely have specific foundation and electrical characteristics that an experienced local inspector sees every week. That knowledge is predictable, location-specific, and exactly the kind of information a buyer needs before committing to an offer.
The tool does not stage anything, does not beautify anything, and asks no one to imagine a room with furniture it does not contain. It tries to tell you what is actually wrong with a house, not make the house look better than it is.
It launched into a market where 84 percent of buyers do not trust AI to help them find a home.
The Carrying Cost of Distrust
Cotality's own data tells a story the company may not have intended to emphasize. Their report estimates that AI-powered workflows could shorten time to close by one to three months. Over 7 million mortgages are originated in the United States annually. If even a modest fraction of those transactions are delayed because buyers distrust AI-powered affordability calculations, risk assessments, or property analysis tools, the cumulative cost is not trivial.
Consider the arithmetic on a single transaction: a buyer carrying a $350,000 mortgage at 7 percent pays roughly $67 per day in interest. A one-month delay in closing, attributable to a buyer manually verifying outputs that an AI system could have provided reliably, costs that buyer approximately $2,000 in additional carrying costs, not counting the rent they continue to pay on their current housing, the inspection fees for redundant checks, and the opportunity cost of a slower search process.
Scale that across even 10 percent of the 7 million annual originations and a one-month average delay: $1.4 billion in excess carrying costs absorbed by buyers who no longer trust the tools designed to help them. That is the approximate annual cost of vanishing chimney breasts and AI-smoothed drywall cracks in listing photos, paid not by the agents who used the tools fraudulently but by the buyers who now refuse to trust any AI tool at all.
The calculation has assumptions that deserve scrutiny. We do not know what percentage of delayed closings are attributable specifically to AI distrust versus the dozen other friction points in a real estate transaction, including appraisal delays, title issues, and rate lock expirations. We cannot isolate AI distrust from general market sluggishness. But the directional point stands: the staging fraud has imposed a tax on the entire market, payable by buyers, and collected by nobody.
What 68 Percent of Buyers Are Actually Asking For
The Cotality survey contains a signal that regulators and platform operators should read carefully. Sixty-eight percent of respondents said clear AI labeling for property listings and mortgage recommendations is important or essential, not optional, not aspirational, but a baseline expectation from more than two-thirds of the buying public. Thirty-seven percent went further and said labeling should be mandatory, a figure that jumped to 61 percent among Baby Boomers who have the least patience for AI's learning curve and the most equity at stake.
Buyers are not asking for AI to go away. Fifty percent of Gen Z respondents said AI involvement would actually make them more confident about buying a home, provided the technology speeds up the process and they can see where it is being used. The demand is not for less technology but for honest technology, clearly labeled, with the original data available, and a human checkpoint at the moments that carry financial consequence.
California got this approximately right: AB 723 does not ban AI staging but requires disclosure and access to originals. That is a regulatory structure built for a market that wants AI but does not yet trust it, because the first generation of consumer-facing AI in real estate was deployed by agents trying to sell houses, not by systems designed to help buyers evaluate them.
What to Do Before Your State Catches Up
If you are buying a home in 2026, in any state, three steps protect you against AI-altered listings regardless of whether your jurisdiction has specific disclosure requirements.
Request original photographs. In California, sellers must provide them by statute, and everywhere else a buyer's agent can request unaltered images as a condition of making an offer. If the seller refuses, that refusal tells you something worth knowing before you drive 45 minutes to a showing.
Use predictive tools proactively. Free tools like Alpine Intelligence exist precisely to arm you with knowledge before the inspection, not after. If you are looking at a 1960s split-level in a suburb with clay soil, the probable inspection findings are statistically predictable. Knowing them before you write an offer changes your negotiating position and your expectations.
Demand labeling. Ask your agent whether any images in the listing were generated or altered by AI, and if they do not know, they should find out before you commit to a showing. NAR has not issued standardized guidelines, but individual agents who proactively disclose AI usage are, at minimum, telling you they take the accuracy of their representations seriously.
Limitations
The 14-point trust decline reported by Cotality is based on survey methodology that the company has not fully disclosed. We do not know sample size, geographic distribution, or margin of error for the 2025 and 2026 surveys, which limits our ability to assess statistical significance. The $1.4 billion carrying-cost estimate is an illustrative calculation, not an econometric finding; it assumes a 10 percent exposure rate and a one-month average delay, both of which are unverified. Alpine Intelligence's predictive accuracy has not been independently audited, and the tool's own marketing materials note it "does not replace a professional property inspection." California's AB 723 has been in effect for five months; no enforcement actions have been publicly reported, making it impossible to evaluate the law's deterrent effect on actual agent behavior.