The three ways a listing photo lies
Sorting them first saves time, because only one of the three is a detection problem and it is the least common of the three.
- Stolen from another listing. The most common by a wide margin. A genuine photograph of a genuine item, taken by somebody else, often from a different country.
- Taken from a manufacturer. A press or catalogue image standing in for an item the seller may not own. Legal in a shop, misleading from a private seller.
- Generated. An item that does not exist as pictured, produced from a prompt. Growing, and still the minority case.
A pixel check catches the third and reports the first two as genuine photographs, because that is what they are. Running it alone and getting a clean result is how people talk themselves into a transfer they should not make.
The order that works is search first, then pixels. A reverse image search settles the two common cases in seconds, and the check is there for the case the search cannot reach.
| Reverse search | Pixel check | Ask for a new photo | |
|---|---|---|---|
| Stolen from another listing | Yes Finds the original posting | No Reads as a real photo | Yes They cannot produce one |
| Manufacturer stock image | Yes Matches the product page | Partly Studio lighting scores oddly | Yes Ends it immediately |
| Generated item | Partly No earlier copy to find | Yes This is the case it is built for | Yes Ends it immediately |
| Real item, hidden damage | No | Partly An edit shows on the map | Yes Ask for the specific angle |
The right-hand column wins every row, which is the practical conclusion of this whole article. Detection is useful; a specific, awkward, cheap request is decisive.
What the region map catches that a search cannot
There is one case where the pixel check outperforms everything else, and it is the case buyers care about most: a genuine photograph of the genuine item, with a defect removed.
Generative fill makes this trivial. A scratch on a phone screen, a dent in a panel, a stain on upholstery, a crack in a frame. Each takes seconds to erase and none of it changes the rest of the photograph, so a reverse search finds nothing and the whole-frame score stays low.
The region map is where it appears. An erased defect leaves a patch whose texture no longer matches its surroundings, and that patch reads hot against a cool frame even when the score sits in the twenties.
Eleven tiles are ordinary. The twelfth covers the part of the screen that was repaired in software.
The request that ends it
Everything above is preparation for one message. Ask the seller for a photograph of the item taken now, with a piece of paper beside it showing today's date and your name written by hand.
This is trivial for anybody who has the item and impossible for anybody who does not. It defeats stolen photographs, catalogue images and generated ones in a single step, and it does not depend on any model being right about anything.
The response is informative even when no photograph arrives. A seller who has the goods sends one in a minute. A seller who becomes irritated, offers an excuse or produces another picture from the same original set has answered the question a different way.
-
Reverse image search the main photo
A hit on an older listing, a shop page or a different country settles it before anything else runs.
-
Save the image and run a check
Read the region map, not just the score. This is where an erased defect shows.
-
Compare the photos in the listing
Different lighting, different floors, different seasons across supposedly one item are a sign of images gathered rather than taken.
-
Ask for the dated photograph
Item, handwritten note, today's date, your name. This is the step that decides it.
-
Keep the payment reversible
No bank transfer, no gift card, no cryptocurrency, no deposit before viewing. The photograph is not the last line of defence; the payment method is.
The categories where this happens most
Fraud concentrates where items are expensive, portable and hard to inspect from a photograph. Phones, games consoles, bicycles, power tools and event tickets account for a large share of it, and all five share the same profile: high value, easy to describe, impossible to verify from a picture alone.
Vehicles deserve their own note. A car listed well below market value, with photographs that look like a dealer took them, and a seller who cannot meet in person, is the oldest pattern in online selling. The photographs are almost always genuine and almost always somebody else.
Rentals are the other high-value case. A deposit paid before viewing, against photographs lifted from an agency listing in another city, is a category of fraud that no image check touches, because every picture involved is a real photograph of a real property.
The common thread is that money moves before anything is inspected. Wherever that is true, a photograph is doing more work than a photograph can do, and the checks that matter are about the payment route rather than the pixels.
One more pattern worth naming: the seller who answers every question fluently and quickly. Fabricated listings are increasingly operated with scripted replies, so responsiveness is no longer reassurance. What cannot be scripted is a photograph of the item taken to your specification.