US Study Finds No Link Between AI Use and Depression
Key Clinical Summary:
- A target trial emulation involving 19,099 US adults found no significant association between frequent generative artificial intelligence (AI) use and depressive symptoms measured by Patient Health Questionnaire 9-item (PHQ-9) scores.
- High-frequency AI use, defined as multiple times per week or more, was not associated with worsening mental health outcomes at follow-up.
- Researchers noted that while findings do not support a causal link between AI use and depression in most adults, potential risks among vulnerable populations require continued monitoring.
Generative artificial intelligence (AI) has raised concerns about potential mental health harms, particularly as adoption increases across work, education, and personal settings. In a target trial emulation published in BMJ Mental Health, investigators found no evidence of a causal relationship between frequent AI use and increased depressive symptoms among adults.
Study Findings
Researchers conducted a target trial emulation using 3 waves of non-probability survey data from a nationally representative US sample collected between June 2024 and January 2025. Adults aged 18 years or older reported how often they used generative AI at baseline, with high-frequency use defined as using AI multiple times per week or more.
Among 19,099 participants assessed at baseline, 2862 individuals (15.0%) reported high-frequency AI use. Follow-up data were available for 3109 participants (16.3%).
The primary outcome was depressive symptom severity measured with the Patient Health Questionnaire 9-item (PHQ-9) at follow-up. In the weighted primary analysis, high-frequency AI use was not significantly associated with changes in depressive symptoms compared with lower-frequency or no use. The mean difference in PHQ-9 score was −0.18 (95% CI, −0.94 to 0.59; p = 0.65).
Investigators also conducted multiple sensitivity analyses using alternative outcome definitions, which found no significant causal relationship between generative AI use and depressive symptoms.
To evaluate whether certain subgroups experienced different effects, researchers applied generalized causal forests to assess heterogeneity of treatment effects. No significant heterogeneity was identified (p = 0.81).
Clinical Implications
For clinicians, this study suggests that generative AI use alone should not currently be viewed as a major population-level risk factor for depression. This may help contextualize patient concerns about technology-related mental health effects and inform discussions about digital tool use.
However, the investigators emphasized that the absence of a significant overall effect does not eliminate the possibility of harms in specific vulnerable groups. Individuals with pre-existing psychiatric conditions, social isolation, or problematic technology use patterns may still warrant closer evaluation.
Expert Commentary
“Our emulated clinical trial, drawing on large-scale longitudinal survey data, does not identify evidence of a causal link between regular AI use and depressive symptoms,” wrote Roy H. Perlis, MD, MSc, Department of Psychiatry, Massachusetts General Hospital, Boston, Massachusetts, and study coauthors.
“While this approach cannot rule out substantial effects in a very small subset of individuals, it suggests that prior cross-sectional findings may have arisen from confounding and that population-wide effects are likely to be modest on average,” they concluded.
