Survey Finds AI Used in 12% of 2025 Scams
A Gallup and Stop Scams Alliance survey of 5,173 U.S. adults, conducted January 8 to February 18, 2026, found that 12% of successful 2025 scams involved AI or deepfakes, part of an estimated $68 billion lost by Americans to scams last year, or roughly $186 million per day. About 6% of adults, or 15 million people, said they were scammed out of money in 2025, and researchers noted the 12% AI figure likely undercounts actual AI use, since many victims cannot tell an AI-generated scam from a human one. Stop Scams Alliance CEO Ken Westbrook compared the scale to "the annual revenues of Delta Airlines," calling it "like a Fortune 500 company." The survey also found 73% of victims reported harm to their mental health or well-being, and 24% of Americans said they have been scammed at least once as an adult.
The number worth sitting with is not 12%, it is that researchers themselves flagged it as a floor, not a ceiling: self-reported victim surveys cannot reliably detect AI involvement in a scam the victim never suspected was AI-generated in the first place. For anyone building fraud detection or content-provenance systems, that gap between reported and actual AI usage is the harder problem, not the headline percentage.
What happened
Gallup and the Stop Scams Alliance published "United States of Scams: The Financial and Emotional Fallout," based on a survey of 5,173 U.S. adults conducted January 8 to February 18, 2026. About 6% of respondents, roughly 15 million people, said they were scammed out of money in 2025, and 12% of those successful scams involved AI or deepfakes. Total US losses to scams were estimated at $68 billion for the year, about $186 million per day. Researchers noted AI-driven fraud may be undercounted because many victims cannot tell whether AI tools were used against them. Stop Scams Alliance founder and CEO Ken Westbrook said: "These guys aren't called organized crime for nothing. They're actually organized, and they're using their organization to start attacking us with scale now to a tune of $68 billion, which is like the annual revenues of Delta Airlines. It's like a Fortune 500 company. It's huge." Beyond financial loss, 21% of victims reported severe financial hardship and another 25% moderate hardship, while 73% said the scam hurt their mental health or well-being; 24% of all Americans said they have been scammed at least once as an adult.
Industry context
The survey follows earlier public reporting, including incident summaries from OpenAI, documenting attempts to use generative AI tools for fraud, and fraud-prevention experts warned earlier in 2026 that deepfake scams were becoming an "industrialized fraud channel" using voice cloning, fake endorsements, and AI-generated impersonations at scale. That framing, industrialized rather than novel, matches what this survey measures: AI is one capability among several that organized fraud operations are folding into existing playbooks rather than a wholly new threat category.
For practitioners
For teams building detection or trust-and-safety systems, the self-reported nature of the 12% figure argues for content-provenance and behavioral-anomaly signals that do not depend on a victim recognizing AI involvement after the fact. Detection systems built only around user reports of "this felt like a deepfake" will systematically miss the more effective attacks, the ones convincing enough that victims never suspect AI was involved at all.
What to watch
Watch for follow-up survey waves from Gallup and the Stop Scams Alliance to see whether the AI share of scams grows, and for vendor abuse-report disclosures (model providers, payment networks) that can offer harder, non-self-reported evidence of AI-enabled fraud volume and technique.
Key Points
- 1A Gallup survey found 12 percent of successful 2025 scams involved AI or deepfakes, part of 68 billion dollars in US losses.
- 2Researchers flagged the 12 percent figure as a likely undercount, since victims often cannot detect AI involvement themselves.
- 3Beyond financial loss, 73 percent of scam victims reported harm to their mental health or overall well-being.
Scoring Rationale
A credible, well-corroborated survey (Gallup methodology, cross-reported by NBC News and Benzinga) quantifying AI's role in a large-scale, economically significant fraud problem. Notable baseline data for practitioners building detection systems; kept at the same level since it is one self-reported survey with an acknowledged undercount limitation, not a definitive measurement.
Sources
Public references used for this report.
Practice interview problems based on real data
1,625 SQL & Python problems across 15 industry datasets — the exact type of data you work with.
Try 250 free problems
