AI-facilitated fraud is growing exponentially. Generative AI-enabled fraud surged 1,210% in 2025. US consumers lost $12.5 billion to fraud in 2023, with projected AI-facilitated losses reaching $40 billion by 2027. One in 10 Americans has experienced a voice clone scam. Deepfakes now account for 11% of all global fraudulent activity. The AI voice cloning market alone is projected to reach $4.06 billion in 2026.
The scale of AI-powered fraud is now documented well enough to move past anecdotes. This page compiles 30+ sourced statistics on AI scams, voice cloning, deepfakes, and related fraud — updated for 2026. All statistics are cited from government reports, academic studies, and industry surveys.
Overall AI Fraud Scale
- $12.5 billion — Total consumer fraud losses reported to the FTC in 2023 (latest full-year data). Source: FTC Annual Data Book 2023
- $40 billion — Projected US losses from AI-facilitated fraud by 2027. Source: Javelin Strategy & Research / AiPrise
- 1,210% — Increase in generative AI-enabled fraud in 2025 vs. 2024. Source: Vectra AI, March 2026
- $4.6 billion — Investment fraud losses in the US in 2025 (leading fraud category). Source: FBI IC3 Annual Report
- $2,764 — Average per-incident AI scam payment in 2025 (up 253% year-over-year). Source: Chainalysis 2026 Crypto Crime Report
- 4.5x — Revenue per operation for AI-assisted scams vs. non-AI scams. Source: Chainalysis 2026
Voice Cloning Statistics
- 1 in 10 Americans have experienced a voice clone scam. Source: 2026 survey data via ScamWatchHQ
- $4.06 billion — Global AI voice cloning market projection for 2026. Source: SQ Magazine, citing industry reports
- 400% — Increase in voice cloning scams since 2022. Source: FTC consumer complaint data
- 3-10 seconds — Audio needed to clone a voice with current AI tools. Source: Multiple security research papers, 2025-2026
- 90%+ — Accuracy of current deepfake audio in mimicking real voices. Source: SQ Magazine Voice Phishing Statistics 2026
- 32% — Rise in deepfake-related fraud attempts at financial institutions in 2025. Source: SQ Magazine
- Voice cloning is now considered the #1 AI fraud attack vector globally. Source: Multiple 2026 cybersecurity reports
Deepfake Statistics
- 11% — Share of all global fraudulent activity attributed to deepfakes. Source: ScamWatchHQ 2026
- $25 million — Single-incident deepfake fraud loss at Arup (Hong Kong, January 2024). Scammers used a deepfake video call impersonating the CFO and colleagues. Source: Financial Times
- €220,000 — Loss from a CEO voice clone scam targeting a UK energy firm (2019 — one of the first documented cases). Source: Wall Street Journal
- $17 billion — Total AI-enabled crypto scam losses in 2025. Source: Chainalysis
- 1,400% — Growth in government-impersonation deepfake scams year-over-year. Source: Chainalysis 2026
- 95.3% — Detection accuracy for image deepfakes (best-in-class model, ScamAI benchmark 2026). Source: ScamAI
- 98.5% — Detection accuracy for voice clone detection across major synthesis platforms. Source: ScamAI 2026
Phishing & Social Engineering
- AI-assisted phishing emails have a click-through rate 3x higher than traditional phishing. Source: 2026 Verizon DBIR
- $2.9 billion — Business Email Compromise (BEC) losses in 2025. Source: FBI IC3
- AI phishing tools are free, require no technical expertise, and can be used anonymously. Source: 2026 International AI Safety Report
- AI can generate thousands of unique, non-templated phishing emails per hour, bypassing traditional spam filters.
Demographic Targeting
- Adults 60+ lost the most to fraud: $3.4 billion in 2023. Source: FTC 2023
- Adults 20-29 reported fraud most frequently but lost less per incident. Source: FTC 2023
- Voice cloning scams disproportionately target elderly victims via the "grandparent scam" variant. Source: AARP Fraud Watch Network
- Romance scam victims (pig butchering) span all ages but ages 40-64 report the highest financial losses. Source: FBI IC3
Law Enforcement Response
- $1.6 billion — Funds recovered by the FBI's Recovery Asset Team from reported fraud since 2018. Source: FBI
- The FBI has conducted multiple international operations shutting down pig butchering compounds in Southeast Asia.
- The FTC Consumer Sentinel Network serves 3,000+ law enforcement agencies with aggregated fraud data.
- The AI Safety Report (2026) flagged consumer fraud as one of the top three AI safety concerns globally.
- Congress began scrutiny of AI voice cloning in April 2026, with bipartisan calls for regulation. Source: ScamWatchHQ
The Technology Behind AI Scams
- Current voice cloning models require as little as 3 seconds of audio from social media, voicemail, or video to produce a convincing clone.
- Real-time deepfake video tools now run on consumer hardware (no specialized GPU required for basic models).
- AI chatbot tools used for romance scams cost less than $20/month for unlimited conversations.
- Deepfake generation time has dropped from hours to minutes in 2026.
Methodology
Statistics are sourced from government reports (FTC, FBI IC3), peer-reviewed research, and industry surveys from established cybersecurity firms. We cite primary sources where possible and update this page as new data is released. Last reviewed: June 2026.
Prevention Guides
Related Resources
- Latest AI news and scam alerts Know what scams are making headlines.
- Legitimate AI tool reviews Compare real tools vs scam pitches.
- Safe AI tools by profession Find tools that are actually worth using.
Frequently Asked Questions
How much money is lost to AI scams each year?
US consumers lost $12.5 billion to fraud in 2023. AI-facilitated losses are projected to reach $40 billion by 2027. These figures are likely underestimates since many victims never report.
What percentage of people fall for AI scams?
One in 10 Americans has experienced a voice clone scam. AI phishing has a 3x higher click-through rate than traditional phishing. The exact percentage varies by scam type and demographic.
Can deepfakes be detected?
Best-in-class detection tools achieve 95-98% accuracy in controlled settings. However, detection in real-world conditions is harder. The most reliable defense remains behavioral — verify identity through independent channels rather than relying on detecting AI artifacts.