Why this matters now: Generative AI is being folded into everyday work, education, and personal life faster than almost any prior technology. As that happens, it’s inevitable, and appropriate, that we start asking not just whether these tools are useful, but whether they might have unintended effects on mental health.
A recent study published in JAMA Network Open1 tackles this question using data from more than 20,000 U.S. adults. The results have already been summarized as showing a link between AI use and depression. That’s not wrong, but it’s incomplete.
What the study examined
The researchers analyzed responses from a large online survey conducted in spring 2025. Participants reported how often they used generative AI tools, what they used them for (personal, work, or school), and completed the PHQ-9, a commonly used measure of depressive symptoms.
The core question was straightforward: Are people who use AI more frequently more likely to report depressive symptoms?
How they examined it
This was a cross-sectional, observational study. Everything was measured at a single point in time, and participants were not randomly assigned to use or avoid AI. This design is well-suited for describing patterns of association at scale, but it cannot examine causation.
The authors used regression models to adjust for key demographic differences—age, gender, education, income, race and ethnicity, and urban versus rural residence—and explored whether the association between AI use and depressive symptoms differed by age group.
What they found
People who reported using AI daily or more often had slightly higher depression scores than people who did not use AI, about a one-point difference on the PHQ-9. In a sample this large, that difference was statistically significant, but from a clinical perspective, the difference may be too small to be very meaningful.
Much larger differences in depressive symptoms were associated with age, income, and education. In fact, one of the most striking findings in the paper had nothing to do with AI use: adults aged 65 and older reported dramatically fewer depressive symptoms than younger adults, by several points on the PHQ-9.
The AI–depression association appeared mainly among adults aged 25 to 64, and not among younger adults or those 65 and older. That age pattern is interesting, but it should be read as descriptive rather than definitive.
How to think about these results
This study does not show that AI use causes depression. It is just as consistent with other explanations, for example, that people experiencing more distress are more likely to turn to AI tools, or that both AI use and mood are shaped by work stress, isolation, or other unmeasured factors.
For clinicians, evaluators, and policymakers, the takeaway is not that AI use is dangerous, but that context matters: who is using these tools, how often, and for what purposes. The effects detected here are real but modest, and far smaller than many well-established drivers of mental health.
Bottom line
This is a careful, well-executed descriptive study. Its value lies in mapping patterns and raising better questions, not in sounding an alarm. As AI use continues to expand, the next step will be studies that can move beyond association to better understand when, for whom, and under what conditions these tools might help, or harm, mental well-being.
References
Perlis RH, Gunning FM, Uslu AA, Santillana M, Baum MA, Druckman JN, Ognyanova K, Lazer D. Generative AI Use and Depressive Symptoms Among US Adults. JAMA Netw Open. 2026 Jan 2;9(1):e2554820. doi: 10.1001/jamanetworkopen.2025.54820. Erratum in: JAMA Netw Open. 2026 Feb 2;9(2):e262242. doi: 10.1001/jamanetworkopen.2026.2242. PMID: 41563755; PMCID: PMC12824790.
