Refund fraud with AI photos: how to spot fake damage claims
A customer says the item arrived broken and attaches a photo. That photo used to be proof. Today it can be made in seconds with an AI app. Here is how the fraud works, and how to check the photo before you refund.
Updated 2026-11-03 · Veriflied
How it works
The classic claim is simple: the customer says the item was defective, sends a photo, and the shop refunds or sends a replacement. Many stores never ask for the item back, because return shipping costs more than the product. That makes the photo the only evidence.
Free AI tools now let anyone take a photo of an intact product and add a crack, a stain or a tear that looks natural. You can also generate a whole picture of a broken item from a text prompt. The customer keeps the product and gets the money back.
Signs in the claim itself
Same customer, repeated damage
The same customer files damage claims again and again for issues no other customer reports on the same product.
The damage does not fit the product
The crack sits where the item cannot break in transit, or the packaging in the photo is perfectly intact.
One photo, no video
The customer cannot send a second angle or a short video when asked.
Screenshot instead of a photo
The image is a screenshot or was sent through an app that strips the file's information. That can be innocent, but it also removes the traces.
Pressure for a fast refund
The customer wants the money now and refuses to return the item, even when you pay the shipping.
Signs in the image
An AI-edited damage photo is usually a real photo where only the damage was added. Everything else in the picture is genuine, which is exactly why it is hard to spot. Look for sharpness and noise that do not match: the damaged area can be smoother or sharper than the rest, and its edges can look too soft.
The eye is no longer enough, though. The newest models produce edits that cannot be seen at normal viewing size. A detector that analyses the pixels helps here, and a heatmap shows which region was altered. If the whole image was generated, it assesses the image as a whole.
In practice
Ask for a second photo from another angle when the amount is high. Genuine customers send it without fuss. Run the photo through an AI detector before approving the refund, and keep the original image in the ticket.
If you run Shopify, the Veriflied app checks customer photos automatically, and in Gorgias or Zendesk the result is posted as an internal note on the ticket, so your support team sees it before they reply. You can try it free for seven days.
Frequently asked questions
How common is refund fraud with AI photos?
There are no reliable figures yet. But the tools are free and need no technical skill, and stores that refund without a return are the obvious target.
Can I reject a claim because the photo is flagged as AI?
Use the result as a reason to look closer, not as the only evidence. Ask for the item back or for more photos. If the customer cannot provide them, you have strong grounds.
What if the customer sent a screenshot?
Then the camera information is gone, but the pixels can still be analysed. Ask for the original photo if the case is large.
Does it hit honest customers?
A real photo of real damage is not flagged as AI-generated. A phone's own image processing can cause a rare false flag, which is why a person should always make the final call.
How do I try it?
Install Veriflied from the Shopify App Store and test it on your own claim photos free for seven days.
Try it on your own document
Create an account and analyse your first 10 images or documents for free. You get a report with the concrete findings, not just a score. All processing in the EU.