Fake listing photos: how to check rental and marketplace images before you pay

The flat is cheap, the photos are lovely, the landlord is abroad, and there is another applicant deciding tonight. Everything about that situation is designed to make you check less carefully than usual, so it helps to know in advance which checks are worth the ten minutes.

Start with the awkward part: most fake listings do not use AI-generated pictures. They use real photographs taken from a genuine listing somewhere else, often a real property that really is for rent, in a different city or a different year. An AI detector will look at those and correctly report real photographs, because that is what they are. The photos are stolen, not invented, and the two problems need different tools.

Reverse image search, first and always

Search two or three of the listing photos (Google Lens, TinEye, Bing Images, Yandex). What you are hoping to find is the same room on another site, an estate agent's archive, a hotel page, or a listing from two years ago in another country.

For rentals specifically, add a text search: paste a distinctive sentence from the description in quotes. Scam listings are copied wholesale and reposted, and the text is lifted along with the photos, so identical descriptions across several cities show up quickly.

If a photo appears in an older listing under a different name, you are finished. That is the majority case, and no further analysis is needed.

Ask for something that cannot already exist

The single most useful request: ask for a new photo of a specific thing, taken now. The inside of the kitchen cupboards. The view out of the back window. The serial number on the appliance. The item next to today's newspaper or a handwritten note with your name on it.

Someone reposting another listing's photos cannot produce that, and what you get instead is a reason: the tenant is still there, the agent has the keys, the photos are all they were given. For a used item on a marketplace, the equivalent is asking for a photo from an unusual angle, or of the item powered on and showing a specific screen.

This is a test of whether they possess the thing, which is a more useful question than whether the pictures were generated.

When the photos really are AI, and the honest complication

AI-generated interiors have a recognizable failure pattern, though it fades with each generator release. Look for geometry that does not resolve: skirting boards that change height, a doorway leading somewhere impossible, tiles that shift pattern mid-floor, radiators merging into walls, window views that do not match the light in the room. Reflections in mirrors and glass remain among the harder things to get right.

Now the complication, and it is the reason a detector result on a listing photo needs care. Virtual staging is a legitimate and widespread practice. Agents routinely use software, increasingly AI-based, to furnish an empty room or replace dated furniture in an otherwise genuine photograph of a genuine property. Some jurisdictions require it to be disclosed, and many listings do disclose it. A detector can flag that image, and the flag would not be wrong: part of the picture really was generated. What it would not mean is that the property does not exist or that anyone is trying to defraud you.

So a flag on a listing photo is a reason to ask a direct question, "has this photo been virtually staged", rather than a reason to conclude anything.

Where detection fits, with real numbers

A calibrated detector answers one narrow question with a measured error rate: does this image carry statistical traces of generation. Two caveats specific to listings.

Condition matters. Listing photos have usually been compressed by the platform, and if you screenshot them rather than saving the file, that is another pass. Compression removes the fine detail detection depends on. On our benchmark, at a threshold calibrated so at most 5% of real photographs are flagged, detection runs at roughly 60% on screenshotted images, against about 64% on clean original files. Save the original image rather than screenshotting it where you can.

Direction matters more. A flag is meaningful evidence, because the threshold is set so false alarms stay under a published cap. A clean result is not evidence of anything: it means nothing was found. Since the common scam uses real stolen photographs, clean is precisely what you should expect a fraudulent listing to produce.

The rules that do not depend on the photos

These are the ones that actually protect the money, and they hold no matter how the image checks come out.

• Never pay a deposit, holding fee, or first month before seeing the property in person, or having someone you trust see it. A request to pay before viewing is the scam, more or less regardless of everything else.
• Never send money by bank transfer, crypto, gift cards, or a "friends and family" payment to someone you have not met. Use a method with a reversal path.
• Verify the person independently. For rentals, check the agency exists, then call the number listed on its own website rather than one you were sent.
• Be suspicious of a price well under the local market. That is the hook, and it is usually the first thing that was fake.
• Refuse the deadline. Manufactured urgency exists to stop exactly the checks in this article.

The limits, before you rely on any of it

• No method proves a listing is genuine. Detection can find evidence of generation, and nothing certifies authenticity.
• A clean result is not reassurance, especially here, where stolen real photos are the dominant pattern.
• A flag can have an innocent cause, virtual staging being the common one.
• Heavily compressed listing images carry less signal, so the numbers above are the realistic ones rather than the flattering ones.
• Photos were never the safeguard. Viewing the property and using a reversible payment method are.