Malwarebytes survey finds 9 in 10 misidentify AI content

Malwarebytes published a research report, "Face Value: How AI is reshaping trust, identity and scams," in June 2026, based on a survey of 1,500 adults in the United States, the United Kingdom, Austria, Germany, and Switzerland, commissioned via independent research consultant Forsta, according to PR Newswire and the report page on malwarebytes.com. The report finds 85% of respondents say they can no longer tell AI-generated content from real content, up from 66% in 2025, and 88% say it is becoming harder to tell if online content is genuinely human, per Malwarebytes. The report also documents that 50% of respondents encountered some form of AI-driven scam in the past 12 months, with exposure rates of 24% for personalized scam messages, 22% for AI-generated photos or product reviews, 19% for identity manipulation, and 16% for voice-cloned calls, as shown in the Malwarebytes findings and summarized in the PR Newswire release.

Advances in generative models, text-to-speech, and image synthesis have reduced the signal-to-noise ratio that humans use to judge authenticity. The Malwarebytes report documents specific modalities being abused - images, audio, video, and text - and gives modality-level exposure rates. Multi-modal detection requires cross-checking metadata, provenance, and behavioral signals across text, audio, and visual channels rather than relying on a single indicator.

The PR Newswire release quoted Mark Beare, Head of Consumer at Malwarebytes: "AI's deepest impact isn't on our devices; it's on us. When people can no longer trust what they see, hear, or who they're talking to, the damage reaches far beyond any single scam and into the building blocks of our society." The report also notes higher exposure among younger cohorts (Gen Z 67% exposure reported) and geographic variation (US exposure 56% vs UK 48% and DACH 47%), per Malwarebytes and PR Newswire. Many respondents reported posting less personal content and removing older posts; relatively few reported adopting defensive tactics such as family code words or data removal requests.

Companies and teams confronting widespread synthetic content typically accelerate investment in three areas: robust provenance and metadata standards, multi-modal detection models that combine audio/text/image signals, and user-facing indicators that surface risk (for example, caller ID protections). The Malwarebytes findings provide quantified consumer exposure data useful for risk modeling, communication strategies, and training dataset labeling.