A couple in a suburban area of Phoenix corresponded via email for three weeks with a real estate agent, or someone they thought was one, regarding a three-bedroom home in a peaceful neighborhood that seemed to be everything they were looking for. The listing photos were clear and comprehensive. The virtual tour went smoothly. On the phone, the agent’s voice was kind and competent. The cost was fair. They changed the escrow account at the last minute and complied with the wire transfer instructions. The funds were taken out of their account. The home had never existed as they had witnessed it. The agent who had led them through the entire process with her voice and face was unaware that someone was posing as her.
Compared to the amateur phishing scams that consumers have been warned about for years, real estate fraud is much more sophisticated in 2025. A fraud pipeline that can trick buyers, evade remote notarization checks, and reroute closing wire transfers has been created by combining AI-generated listing imagery, cloned voices, and real-time video face-swapping. All of this is possible without the perpetrator ever physically being close to the transaction. In 2024, damages from real estate and rental fraud totaled $145 million, according to the FBI’s Internet Crime Complaint Center. In the same time frame, victims in all company categories lost $446 million due to wire fraud diversions, the category that deepfake scams most directly target. After monies are transferred from a US bank to an international account, recovery rates for wired funds typically range from three to five percent. The majority of victims receive nothing in return.
The technology used in these systems is more complicated than the mechanics of how they operate. A fraudster locates a target property, usually an empty lot, an absentee-owned rental, or an inherited house whose out-of-state heirs won’t discover it right away, and then uses public records to obtain the listing agent’s name. Next, they gather content, including photos, audio from podcast appearances or social media reels, and video snippets from YouTube, Instagram, or LinkedIn. To create a realistic voice clone, thirty to ninety seconds of original audio are sufficient. Real-time face-swapping software that operates on consumer-grade hardware may obtain sufficient facial data from a few minutes of footage. AI-generated images, which the industry is beginning to refer to as “AI slop,” are used by the fraudster to create a listing for the target property. These images can be identified in retrospect by floating furniture, rooms with improbable geometry, window lights that don’t match the exterior, or grass that extends in the wrong direction. They wait for purchasers to contact them after posting it on aggregator platforms that don’t regularly confirm the legitimacy of listings.
The initial exchanges take place over text or email, utilizing a fictitious number or address that is sufficiently similar to the actual agent’s contact details for a buyer who is preoccupied to miss it. The face-swap software operates in real time, superimposing the stolen agent’s appearance on the fraudster’s face and translating their voice through the clone when a video chat is requested, which many purchasers now view as a due diligence measure. A live video check is usually used to confirm identity on remote notarization systems, which are becoming more and more popular for facilitating out-of-state transactions. Many have admitted that under normal protocols, modern real-time deepfake technology can evade these tests. The deal is completed. The wire instruction is modified. The buyer transfers their down payment or closing monies to an account controlled by the fraudster, thinking they are adhering to valid closing advice from a reputable expert.
Because it eliminates one of the conventional safeguards that buyers depend on, the vacant lot angle merits special consideration. The perfect target for this type of fraud is a property that actually exists but has no occupants, no active owner management, and no close neighbors who could detect unusual activities. Before anyone detects that the underlying land record and lawful ownership don’t match the transaction, the scammer can market it, display it using AI-generated graphics and street-level photos taken from mapping services, and collect money from a buyer who wires money. The money has gone through several accounts by the time the true owner finds out their property was sold without their knowledge, which is frequently only when documentation or a buyer shows up.
Since 2024, the National Association of REALTORS® and a number of state real estate commissions have issued guidelines advising buyers to independently confirm all wiring instructions by contacting the escrow agent or title company using the number listed in the original engagement documentation, rather than a number given in an instruction-change email or call. AI-generated listing imagery is now subject to disclosure rules in Texas and California. The MLS platforms that disseminate listings differ greatly in how thoroughly they search for fake photos, and there is currently no federal standard in place. The technology that creates fake listings is advancing more quickly than the detecting tools that keep an eye on it.
