Return and claim fraud

How to spot return fraud: an empty box, a different item or an AI damage photo

A return that does not add up is expensive and hard to prove. The Dutch Disputes Committee for online retail handled a case where a customer returned two headphones and the shop said it received an empty box and a different headphone — and ruled for the customer, because the shop had insufficient evidence: a photo of the wrong headphone that was not clear enough, and no photo of the empty box at all (Thuiswinkel.org, 10 April 2024). New technology makes it harder, not easier: brands report a surge of damage claims with AI-generated photos, and at US brand Boll & Branch a submitted "damage photo" still carried an AI watermark (Modern Retail, 2 March 2026).

The rule that breaks the pattern: record every return the moment you open or ship it, and treat a photo as a claim — not as evidence.

Direct answer: five checks before you refund or return

1. Open and film the return as soon as it arrives. Use one continuous recording showing the shipping label, the box and the full contents, and record what was actually inside. That is exactly what the committee expects a shop to be able to do: explain and evidence what it received (Thuiswinkel.org). At Boll & Branch the support team asked to verify the claimed damage over a video call; the customer never responded (Modern Retail).
2. Compare the contents against your own shipping record. Photos of the parcel with its label before dispatch, the weight and the tracking reference are the starting point. If something different comes back, that belongs in your system before the discussion starts — not after the customer asks for a refund (Thuiswinkel.org).
3. Treat a damage photo as a claim, not as evidence. Ask for the original file instead of a forwarded screenshot, for several angles with a reference object, or for a fresh photo or video call showing a detail you choose. Someone who really has the damaged item can usually supply that within minutes (Modern Retail).
4. Watch the pattern across returns. Return fraud concentrates in expensive goods — electronics, designer clothing, jewellery — and recently in trading cards; retailers respond with mandatory pre-registration, per-item return fees and monitoring of abnormal return behaviour (RTL Z, 24 January 2026). This is not anecdotal: an estimated 14 percent of all retail returns is fraudulent, totalling more than 100 billion dollars a year (NRF and Happy Returns, cited by Modern Retail).
5. Escalate proportionally, without accusing anyone. Record the five checks in the file, stick to your own published return terms, and let a dispute or suspicion be judged through the platform, the disputes committee or, where fraud is suspected, the police. Never accuse on a single signal — and certainly not on a detector score, which says nothing about the damage or the intent (Thuiswinkel.org).

Two patterns, one structure

Pattern A — the contents do not match. The box comes back empty, with an older or cheaper item, or with stones instead of the product. In the Amazon case, return labels were scanned with a cloned terminal and cheaper products or empty packaging were sent back (RTL Z). Without your own unboxing recording, this pattern is almost impossible to prove (Thuiswinkel.org).

Pattern B — the damage does not match. The customer claims transit damage or a defect and attaches a photo. That photo can be genuine, but it can also be cropped, edited or entirely generated — a convincing damage photo takes about a minute with image generators (Modern Retail). Assessing it is the same problem as an insurance claim: the five checks on a damage photo work here too. A payment proof someone attaches is judged like a fake payment proof: your own environment is the source of truth.

What a detector does and does not see

A damage photo is a file. You can run it through the image detector to see whether it carries patterns common in generated or edited imagery. That is triage: it tells you which claim to verify first, and which question to ask. The score says nothing about whether the damage exists or about the sender's intent — never evidence, never an accusation, only a reason for your own check.

If you are the one returning something

• Film the packing or photograph the item, the box and the shipping label before you hand it in; that is your evidence if the arrival is disputed.
• Use a tracked route and keep proof of dispatch and delivery.
• Register the return through the platform or the shop itself, as retailers increasingly require (RTL Z).

Frequently asked questions

What is return fraud?

Return fraud is abuse of the right of withdrawal: sending back an empty box or a different, cheaper item while claiming a refund, or claiming damage that does not exist — nowadays sometimes with an AI-generated damage photo (Thuiswinkel.org; Modern Retail).

Does a detector score prove a damage photo is fake?

No. The score describes patterns in a file and says nothing about the damage or the sender. A genuine photo can draw a signal from compression, and a generated one can stay quiet; use the outcome only to decide which claim to verify first.

What may a shop ask when a return looks wrong?

For the original file or fresh imagery with a detail the shop chooses, and for a concrete substantiation of the claim. The disputes committee expects the shop to evidence what it received; the customer keeps the right to a substantive assessment of the claim (Thuiswinkel.org).