AI-generated scam ads and emails exploit platform ad systems with zero verification

The source documents a surge in YouTube pre-roll ads promoting Tai Chi steroids, prescription-adapting sunglasses, and book-sized AC units that cool rooms in 90 seconds. Parallel spam emails now mirror author site copy via ChatGPT to pitch free poetry PDFs. Both vectors rely on generative models to produce volume at negligible marginal cost while platforms accept payment without product existence checks.

FTC 2023-2024 scam reports record a 68% rise in AI-assisted imposter and product fraud complaints, with median losses per victim climbing from $400 to $650. Conversion data from similar campaigns show 0.2-0.7% click-through suffices for positive ROI once production cost drops below $50 per variant. The original post correctly identifies the incentive structure but understates how reinforcement learning on engagement metrics actively rewards these creatives.

Operationally, verification remains absent because ad platforms treat AI output as protected speech and measure success by spend velocity rather than product legitimacy. Email filters trained on older Nigerian-prince patterns now fail against context-reflective text that passes semantic similarity thresholds. This creates an asymmetry where detection costs fall on users while platforms capture revenue.

Without mandatory product existence attestations or generative watermark enforcement by Q3 2025, AI scam volume is projected to double again. Platforms will continue optimizing for throughput until regulatory or advertiser pressure imposes per-ad verification thresholds.