Generative AI Use and Depressive Symptoms Among US Adults

Key Findings

• — 10.3% of respondents (2,152 of 20,847) reported using generative AI at least daily: 5.1% daily and 5.3% multiple times per day; among daily-or-more users 87.1% used it for personal applications, 48.0% for work and 11.4% for school.
• — Daily use was associated with higher PHQ-9 scores than non-use (beta 1.08, 95% CI 0.55 to 1.62; multiple times per day beta 0.86, 95% CI 0.35 to 1.37) in sociodemographic-adjusted survey-weighted models.
• — Daily-or-more use was associated with greater odds of at least moderate depressive symptoms (OR 1.29, 95% CI 1.15 to 1.46); similar patterns were observed for anxiety and irritability.
• — The largest estimates were for personal (non-work, non-school) use (beta 0.31, 95% CI 0.10 to 0.52) and for adults aged 25 to 44 (beta 1.22) and 45 to 64 (beta 1.38); daily-or-more use was more common among men, younger adults, and those with higher education and income and in urban settings.
• — The authors compare the pattern with the cross-sectional social-media literature and note that longitudinal and experimental designs are needed to test direction of effect.

Methodology Notes

Survey-weighted regression on wave 35 of the Civic Health and Institutions Project (CHIP-50, formerly the COVID States Project), a nonprobability internet survey through a commercial panel aggregator (PureSpectrum) with state-level quotas and reweighting; fieldwork 20 April to 27 May 2025; adults 18 and over in all 50 states; data analysed August 2025. Outcome PHQ-9; exposure self-reported generative AI and social media use. Limitations stated by the authors: opt-in panel so response rates and biases cannot be estimated; cross-sectional and correlative. Open access (PMC12824790, PMID 41563755). The JAMA landing page returns HTTP 403 to automated fetchers; verified from the Europe PMC full-text record.

Sources