Retailers are seeing a rapid rise in return fraud that uses generative AI to produce fake damage photos and doctored receipts. Retail Gazette reports data from digital trust platform Forter showing AI-generated damage claims are the fastest-growing form of return abuse, with 53% of merchants reporting "wardrobing" and 44% of UK businesses saying returns abuse affects them. Retail Gazette also reports some fraud rings offer "returns-as-a-service." PYMNTS documents merchant cases where images contained AI watermarks and cites industry data that U.S. consumers returned nearly $1 trillion in merchandise in 2024, with retailers incurring an estimated $200 billion annually to recover returned goods, per PYMNTS reporting and Riskified analysis. Editorial analysis: this trend raises detection and operational trade-offs for fraud teams and retail operations.
What happened
Retail Gazette reports data from global digital trust and fraud prevention platform Forter showing that AI-generated damage claims are now the fastest-growing form of return abuse. Retail Gazette reports that 53% of merchants are seeing "wardrobing," 30% of consumers admit buying extra items to qualify for free shipping, and 44% of UK businesses report being affected by returns and refund abuse. Retail Gazette also reports some fraud rings have scaled by offering "returns-as-a-service," and that almost half of retail leaders surveyed considered scaling back operations or shutting down this year because of returns pressure.
PYMNTS documents merchant incidents in which submitted damage photos carried AI watermarks and cites industry-wide figures, reporting that U.S. consumers returned nearly $1 trillion in merchandise in 2024 and that retailers spend an estimated $200 billion annually to recover value from returned goods. PYMNTS cites a Riskified analysis finding refunds account for 1% to 2% of total sales dollars in the dataset analysed, with nearly 1 in 4 refund dollars linked to abuse.
Editorial analysis - technical context
Generative image models and widely available image-editing tools lower the technical barrier for fabricating visual evidence. Industry-pattern observations: digitally generated damage images can be high-quality, stripped of metadata, and produced at scale, which undermines heuristics that rely on low-resolution or obviously edited images. Fraud detection pipelines that depend on legacy rules or isolated signals (receipt OCR, simple image checks) typically lag when adversaries can iterate new synthetic outputs rapidly.
Industry context
Industry-pattern observations: retail verticals with high return rates, notably fashion, face amplified margin pressure because returns already erode profitability. Reporting highlights new operational vectors such as coordinated "returns-as-a-service" and fabricated "item not received" narratives that combine social engineering with synthetic media. For fraud teams, the tension between minimizing false positives (customer experience) and flagging sophisticated synthetic evidence is widening across merchants and logistics partners.
What to watch
- •Adoption of image-forensics and multi-signal verification (photos, timestamps, delivery telemetry) across returns workflows.
- •Uptick in AI-watermarked or machine-generated indicators in submitted tickets, as reported by merchants in PYMNTS coverage.
- •Emergence of third-party "returns-as-a-service" marketplaces and whether regulators or industry groups publish guidance on synthetic-evidence fraud.
Editorial analysis: practitioners should monitor model-driven synthetic-evidence patterns and invest in multidisciplinary detection that combines visual forensics, device/identity signals, and logistic telemetry, while recognising trade-offs with customer friction.
Scoring Rationale #
This story is notable for practitioners in fraud prevention and retail operations because generative AI materially lowers the cost and scale of synthetic-evidence attacks, creating immediate operational and detection challenges. It is not frontier-model news but has direct operational impact across merchants and logistics.
Practice with real Retail & eCommerce data
90 SQL & Python problems · 15 industry datasets
250 free problems · No credit card
See all Retail & eCommerce problems