A new survey from Malwarebytes finds that 88% of adults across the US, UK, and Europe say they can no longer reliably distinguish real content from AI-generated material. The figure is up sharply from 66% in 2025, reflecting rapid improvements in generative AI quality across text, images, audio, and video.
The survey, covering 1,500 adults aged 18 and older across five countries, also found that half of respondents had encountered an AI-driven scam in the past year. Common experiences included personalized scam messages, manipulated product reviews, AI-generated images used in fraud, and voice impersonation attacks.
The findings align with a separate Veriff report from May 2026 showing that Americans consistently fail deepfake detection tests even when they believe they can spot fakes. Together, the research paints a picture of a widening gap between AI generation capabilities and human detection abilities.
Security researchers warn that traditional advice like 'look for visual artifacts' is becoming obsolete as generation quality improves. The reports suggest that technical detection tools and platform-level authentication will need to replace human judgment as the primary defense against AI-generated misinformation and fraud.