How Scammers Use AI-Generated Images to Get Refunds (And How to Stop It)
From dead crabs to shredded sheets, a new wave of 'Synthetic Fraud' is costing merchants billions. Here is how the scam works and why manual teams cannot catch it.
The New Fraud Landscape
Refund scammers use AI-generated damage photos instead of actually damaging products. A Midjourney prompt creates refund-worthy images in seconds.
The Scale
Veto data shows 400% month-over-month increase in synthetic refund fraud across:
- Apparel (AI-generated "shredded" sheets)
- Electronics ("cracked" screens on intact devices)
- Beauty ("broken" palettes, "leaking" products)
Case Study: Dead Crabs
Wired reported Chinese merchants receiving thousands of fraudulent claims with AI-generated videos of dead crabs. Detection: impossible leg counts, sex changes between frames.
Why Manual Review Fails
Experienced fraud agents misidentify AI-generated images 35% of the time. Modern AI (Midjourney v6) eliminated traditional tells:
- Perfect lighting
- Accurate text rendering
- Screenshot metadata stripping
At 50 tickets/hour, agents lack time for deep analysis.
Veto's Detection Method
Statistical artifact analysis:
- Noise pattern signatures from diffusion models
- Compression history (raw camera vs. synthetic file)
- Physics violation detection
Human eyes vs. AI generation: humans lose. Use algorithmic detection.