Most people rarely use AI, and dark personality traits predict who uses it more (www.psypost.org)

Researchers analyzed more than 14 million browser visits from two cohorts (499 university students and 455 members of the public) who shared up to 90 days of Google Chrome history to objectively measure AI use. Using a curated list of AI sites (e.g., ChatGPT, Microsoft Copilot) and an LLM-powered content classifier, they found AI-related browsing comprised under 1% of visits for most people (≈1.0% for students, 0.44% for the general public), with ChatGPT accounting for over 85% of AI traffic. AI sessions were often embedded in workflows—typically preceded by searches or logins and followed by education or professional sites—suggesting productivity-focused use rather than entertainment. Self-reports correlated only moderately with actual use, highlighting limits of survey-based estimates. The study also linked heavier AI use to higher scores on aversive personality measures (Machiavellianism, narcissism, psychopathy), especially among students and among “prolific users” whose AI visits exceeded 4% of browsing. Findings offer a rare objective baseline for real-world AI adoption, call attention to who is actually using generative tools, and suggest design, policy, and research priorities: integrate AI into workflow-support features, monitor effects on outcomes like academic integrity, and extend measurement beyond web-only, Chrome-based traces (e.g., mobile apps, interaction content) to better understand motivations and consequences.