Generative AI Use and the Verification Gap Follow Opposing Social Gradients in Türkiye

Abstract

Generative artificial intelligence (AI) use and online information checking may be socially patterned in different ways. This cross-sectional study analysed the 2025 Turkish Statistical Institute ICT survey of 22,502 internet users aged 16–74 using weighted modified Poisson regression with household-clustered standard errors. Overall, 19.2% reported generative AI use, and 48.3% of respondents who noticed doubtful content did not check it. Conditional non-checking (the verification gap) increased with age and decreased with education and reported digital activity, whereas AI use followed the opposite pattern. Compared with ages 16–24, the adjusted prevalence ratio at ages 65–74 was 1.48 (95% confidence interval 1.30–1.68) for non-checking and 0.10 (0.07–0.16) for AI use. Health information seekers and AI users reported less non-checking among those who noticed doubtful content, but neither group had a lower prevalence of noticing without checking among all internet users. Among non-checkers, 87.3% reported already judging the information or source unreliable and 15.2% reported insufficient skills or knowledge; responses could overlap. These findings distinguish participation in AI use, reported noticing of doubtful content, and subsequent checking. Verification support should address reported difficulty separately from decisions not to investigate content already considered unreliable.