Chainalysis — the blockchain analytics firm whose data is used by law enforcement agencies and financial institutions worldwide to trace cryptocurrency crime — has released its 2026 Crypto Crime Report, and the scam chapter contains figures that mark a clear inflection point in the history of digital fraud.

$17 billion in cryptocurrency scam losses in 2025. 1,400% growth in impersonation scams year-over-year. 4.5x higher revenue per campaign for AI-enabled operations compared to non-AI operations. An average scam payment of $2,764 — up from $782 the prior year, a 253% increase.

These are not incremental changes. They reflect a fundamental shift in how cryptocurrency fraud operates, who runs it, and how much damage each operation can inflict.

The $17 Billion Figure

The Chainalysis estimate of $17 billion in cryptocurrency scam losses for 2025 encompasses the full range of crypto fraud categories: pig butchering (romance-to-investment fraud), high-yield investment programs (HYIP), fake trading platforms, impersonation scams, recovery scams, and approval phishing.

This figure is higher than the FBI IC3’s own cryptocurrency fraud estimate of $11.36 billion for the same year. The difference reflects methodology: IC3 data captures only reported complaints from US-based victims. Chainalysis tracks on-chain fund flows globally, capturing fraud losses regardless of whether victims filed reports, regardless of their country of origin, and including cases where victims transferred funds denominated in cryptocurrencies that don’t map cleanly to fiat loss figures in FBI reporting.

Both figures are almost certainly undercounts — IC3 because of underreporting, and Chainalysis because not all fraud moves through trackable on-chain flows. The true global figure for cryptocurrency fraud losses in 2025 is probably substantially higher than either estimate.

Impersonation Scams: 1,400% Growth

The most striking finding in the Chainalysis report is the growth rate for impersonation scams: a 1,400% year-over-year increase in 2025.

Impersonation scams in the cryptocurrency context include:

  • Fake exchange support: criminals impersonating customer service for Coinbase, Binance, Kraken, and other legitimate exchanges to steal credentials or prompt “security transfers”
  • Celebrity endorsement fraud: AI-generated video of public figures promoting fake investment platforms
  • Government/agency impersonation: fake SEC, FTC, FBI, or DOJ agents claiming a victim’s assets are under investigation and must be “transferred to a secure wallet” during the investigation
  • Familiar entity impersonation: scammers pretending to be known organizations — Apple, Amazon, Microsoft — initiating tech support scams that escalate to crypto transfers

The 1,400% growth figure is partly attributable to better tracking methodology — Chainalysis now identifies more impersonation fraud than in prior years. But the firm is clear that the scale of actual impersonation fraud growth is real and significant, driven by AI tools that dramatically reduced the cost and skill required to run convincing impersonation campaigns at scale.

AI Changes the Unit Economics of Fraud

The central analytical finding of the Chainalysis report is what AI tools have done to the economics of fraud at the campaign level.

Non-AI scam operations in 2025 averaged approximately $719,000 in revenue per campaign across the measurement period. AI-enabled scam operations averaged approximately $3.2 million per campaign — a 4.5x difference.

This premium breaks down into several components:

Higher-quality initial targeting: AI tools analyze social media data to identify potential victims with characteristics that predict susceptibility — recent financial setbacks, life transitions, expressions of financial goal-setting — allowing operators to focus human attention on the highest-potential targets.

Greater convincingness: AI-generated content (fake profiles, personalized messages, deepfake endorsements) passes authenticity checks that manual content frequently fails. Fewer victims detect the scam early, which means more victims reach the high-value extraction phase.

Faster victim throughput: AI handles the early-stage relationship maintenance that previously required significant human time, allowing operators to manage more victims simultaneously without increasing the human labor investment.

Higher per-victim extraction: AI-maintained relationships are more consistent and emotionally calibrated over time, which research suggests produces higher final transfer amounts before victims become suspicious.

The 76% figure — 76% of AI-enabled scams falling into the highest-value loss category — reflects this dynamic. AI does not just make scams more numerous; it makes them more effective at extracting the maximum possible amount from each victim.

The Average Payment Increase: What It Means

The jump in average scam payment from $782 to $2,764 — a 253% increase in a single year — is one of the report’s most practically significant figures for consumers.

A $782 average implies that many successful scam extractions involved relatively small amounts: a single “investment” transfer, a one-time “fee” payment, a modest emergency request. At this level, victims lose money but not necessarily life-changing amounts.

A $2,764 average implies something different: victims are being extracted multiple times, in growing amounts, over longer engagement periods. This is consistent with the pig butchering model, where victims are managed through weeks or months of relationship-building and incrementally larger deposits before the final extraction.

The payment increase also reflects the shift toward AI-assisted operations that can maintain longer victim relationships without the labor costs that previously limited how long a manual operation could sustain engagement.

The Pig Butchering Adaptation

Chainalysis’s 2026 data shows that pig butchering and high-yield investment programs remain the dominant categories by total volume — but the specific execution has evolved.

Early pig butchering operations followed a consistent formula: contact on dating or social media apps, relationship building over weeks or months, introduction to a fake cryptocurrency trading platform, incremental deposits, fabricated profits, failed withdrawal, final extraction.

In 2025 and 2026, the Chainalysis data shows increasing convergence between categories. Operators are mixing relationship-building approaches with impersonation elements (the “contact” claims to be from a legitimate financial institution). They are using deepfake video alongside text relationships. They are incorporating task-scam elements (small initial “jobs” that appear to pay, building trust before larger requests) alongside investment fraud pitches.

This convergence is being driven by AI tools that make it easier to blend multiple scam approaches within a single victim engagement, and by criminal networks that are sharing techniques across previously distinct operation types.

What Chainalysis Says Happens Next

The Chainalysis report includes a forward-looking assessment that frames 2025 as a turning point rather than a peak:

“We are moving toward a future in which virtually all scams will incorporate AI into their operations to some degree,” the report states. “Deepfake videos, voice cloning, and AI-generated messages have made impersonation far more convincing, allowing criminals to exploit trust on an unprecedented scale.”

The firm identifies several near-term developments that will likely drive continued loss growth:

  • Agentic AI systems capable of autonomously managing full victim relationships from initial contact through extraction, without human intervention
  • More convincing synthetic identity documents that allow scam platforms to pass Know Your Customer checks at legitimate exchanges, improving the apparent legitimacy of the fake platforms
  • Cross-chain laundering automation that makes it harder to trace and freeze funds after a successful extraction

Against these developments, Chainalysis cites the growing investment in on-chain detection systems by financial institutions, the improving coordination between blockchain analytics firms and law enforcement, and the expansion of regulatory frameworks that require exchanges to implement more robust fraud detection.

The 2026 enforcement actions — Operation Atlantic ($12M frozen), the DOJ Strike Force seizure of 503 domains and $700M in crypto, Florida’s record $5.4M recovery — are evidence that the defensive tools are getting more effective.

The question is whether they are getting more effective fast enough.

The full Chainalysis 2026 Crypto Crime Report is available at chainalysis.com.