Researchers at Beth Israel Deaconess Medical Center, Harvard University, and the University of Wisconsin reviewed published studies, surveys, and public reports to estimate how many AI users turn to AI for mental health support. Based on the available evidence, they estimate that approximately 27% of AI users have used AI for mental health support.
The authors caution that this estimate should be interpreted carefully. Across the studies they reviewed, reported rates ranged from 3% to 70%, largely because researchers used different definitions of mental health support, surveyed different populations, and applied inconsistent methodologies. As a result, the authors conclude that current evidence cannot produce a precise prevalence estimate.
Rather than focusing solely on the percentage, the paper concludes that the field needs standardized definitions and more consistent research methods before AI use for mental health can be measured reliably across studies.
As AI becomes more common in mental health care, clinicians and practice leaders will encounter studies reporting widely different estimates of AI use. This review explains that those differences often reflect inconsistent definitions and research methods rather than conflicting evidence.
For Therapy AllyTM, the findings reinforce the importance of clearly defining AI's role in clinical care. The paper distinguishes between therapy, emotional support, and wellness support as separate categories and calls for greater consistency across the field. Therapy Ally's model of clinician-guided between-session support fits within this broader effort to clearly define how AI is used alongside, rather than in place of, therapy.