AI-Powered Early Intervention for Depression and Anxiety: A Randomized Clinical Trial

Importance: Depression and anxiety are prevalent and disabling, yet most individuals receive inadequate or no care. Purpose-built generative AI platforms may offer scalable first-line support, though large-trial evidence remains limited.Objective: To determine whether Flourish, an AI-powered mental health platform, prevents and reduces depression and anxiety symptoms compared with care-as-usual.Design: Preregistered, parallel-group, superiority randomized clinical trial conducted in 2025, with assessments at weeks 0, 2, 4, and 6.Setting: A large North American university.Participants: A total of 1,137 college students aged 18 years or older was recruited through a research participation system without symptom-based screening and randomized to treatment or care-as-usual. 850 participants (74.6%) completed the final Week 6 follow-up.Intervention: Participants assigned to treatment received Flourish, which offered self-guided support informed by cognitive behavioral therapy, dialectical behavior therapy, acceptance and commitment therapy, positive psychology, and motivational interviewing, and were instructed to use it at least twice weekly. Participants assigned to care-as-usual retained access to usual mental health resources.Main Outcomes and Measures: Preregistered primary outcomes were depression (Patient Health Questionnaire-2) and anxiety symptoms (Generalized Anxiety Disorder-2), assessed biweekly.Results: Among 1,137 participants, mean (SD) age was 20.77 (5.34) years, and 853 (75.0%) were female. Symptom trajectories were more favorable in treatment (vs control) for depression (condition × time: β = −0.03; 95% CI, −0.05 to −0.01; P = .006) and anxiety (β = −0.04; 95% CI, −0.06 to −0.01; P = .001). Among participants with elevated baseline symptom severity, treatment increased rates of meaningful change for depressive symptoms (52.3% vs 31.7%; risk ratio [RR], 1.65 [95% CI, 1.16-2.34]; number needed to treat [NNT], 4.9 [95% CI, 2.9-14.8]) and anxiety symptoms (50.6% vs 38.1%; RR, 1.33 [95% CI, 1.05-1.68]; NNT, 8.0 [95% CI, 4.4-44.5]). Participants with low baseline symptoms showed less symptom worsening at 6 weeks. No adverse events attributable to the intervention were identified.Conclusions and Relevance: This trial demonstrates that a purpose-built generative AI platform can serve as a scalable first-line mental health tool, effective for prevention and early intervention of depression and anxiety symptoms in an unscreened general population.Trial Registration: ClinicalTrials.gov identifier: NCT06668532 (https://clinicaltrials.gov/study/NCT06668532).