AI fake damage photos in refund claims: how merchants can respond

A return claim rests on a single photo the buyer uploads, and nothing about that file is anchored. Its date and location can be edited, and the image itself can be generated or retouched, so a reviewer cannot tell a real defect from a fabricated one by looking. Manual review does not scale to the volume of returns, so most claims are approved on trust.

The honest customer suffers too. Once staff suspect AI, a genuine photo of real damage can be waved away as a possible fake. The whole channel loses credibility, and the dispute drags on for a loss that actually happened.

An AI media detector is useful as a first screen: it scans an uploaded file and returns a probability that the image was generated or edited. Danaya offers exactly this, and it is worth running on incoming claims. But a probability is not proof. A detector can flag a suspicious image and still be wrong in either direction, so it triages claims rather than deciding them.

The stronger move is to stop relying on a file the buyer could have made freely, and instead require the proof to be created at capture. That is a different question: not was this image generated, but was this photo taken live, now, of the real item.

Danaya lets a merchant send a capture link to the buyer. The buyer opens it, captures the damaged item live in the browser with no account to create, and the sealed result lands in the merchant's dashboard with a server timestamp, the GPS position and a SHA-256 fingerprint of the image. Capture is live only, so an imported file or an AI-generated image cannot earn a valid seal.

A practical policy is to screen every claim with the detector, and require a sealed live capture for high-value or already-flagged returns before a refund is approved. Danaya is a layer that makes evidence verifiable, not a guarantee that every fraud is caught, and human review stays part of the process. The GPS position is declarative, taken from the capturing device.