How Common Are AI-Powered Crypto Scams Becoming?
Chainalysis, FBI, and TRM Labs data show AI-powered crypto scams growing fast and earning far more per operation. Here's what the numbers really say.
It's easy to assume AI scams are mostly hype. The data from the past year suggests otherwise: AI isn't just present in crypto fraud, it's associated with the most profitable operations.
The Overall Scale of Crypto Fraud
Blockchain analytics firm Chainalysis estimated in its 2026 Crypto Crime Report that around $17 billion was lost to crypto scams and fraud in 2025, the highest figure it has recorded. That estimate is expected to be revised as more scam addresses are identified over time.
AI-Linked Operations Earn Far More
The most striking finding: scam operations with on-chain links to AI tool vendors took in roughly $3.2 million on average, about 4.5 times more than operations without such links. AI doesn't just make scams cheaper to run, it appears to make them considerably more effective.
Impersonation Is Growing Fastest
The same report found impersonation scams, such as fake exchange support, fake government notices, and fake trusted insiders, grew by around 1,400% year over year. These are exactly the formats AI makes easiest to produce: convincing support chatbots, cloned voices, and deepfaked officials.
Fewer Victims, Bigger Losses
The average scam payment rose from roughly $780 in 2024 to about $2,760 in 2025. That pattern suggests a shift away from mass, low-value spam toward more targeted, personalized approaches aimed at people with more to lose.
Other Data Points Tell a Similar Story
In the United States, the FBI reported more than 22,000 complaints referencing AI in 2025, with close to $900 million in reported losses. Separately, TRM Labs' scam-reporting data showed reports of generative-AI-enabled scams rising by over 450% between mid-2024 and mid-2025.
Why These Numbers Can't Simply Be Added Together
Each source measures something different: blockchain flows, formal complaints, or user reports. They overlap and use different time windows, so they shouldn't be combined into one total. What they share is direction: every dataset points upward.
Why the Real Figures Are Likely Higher
Many victims never report losses, and many scams aren't identifiable as AI-assisted from the outside. Published figures are best read as lower bounds, not complete counts.
What This Means for You
AI-assisted scams are no longer an edge case. They're a mainstream, growing, and unusually profitable part of crypto fraud, which makes consistent verification habits more valuable than ever.
Check any platform, token, or request independently before acting. The trend lines are all pointing in the same direction.