Randomized Trial of AI-Powered Mental Health Support Embedded in University Academic Infrastructure

Key Findings

• — 1,137 University of Manitoba undergraduates were randomized to app access (n = 556) or care as usual (n = 581); retention was 88 to 89% at week 2, 82% at week 4 and 75% at week 6 in both arms.
• — Treatment-arm students used the app on 18.9 days on average (SD 10.2), about 3.2 days per week, and 95% activated within two weeks.
• — App access protected against declines in positive affect, resilience and flourishing over the semester and accelerated reductions in loneliness; belonging and closeness to the institution did not differ.
• — The intervention was described as including crisis detection and routing, guardrails on the scope of support, and clinician review, and it extends an earlier three-site US trial to a resource-constrained setting.
• — One author is affiliated with Flourish Science, the app's developer.

Methodology Notes

Preregistered parallel-arm RCT, Fall 2025 semester, University of Manitoba, n = 1,137 undergraduates, linear mixed-effects condition-by-time analysis on weekly self-report measures; the PDF reports the CONSORT flow, an average of 18.9 use days and week-6 effect sizes around d = 0.2. Posted to PsyArXiv 2026-09-10 (OSF API date_published); no journal version as of 2026-09-15. Verified through the OSF preprint API and the downloaded 29-page PDF; the osf.io HTML page is a client-side shell. Vendor co-authorship and the fact that the app was introduced through academic infrastructure (with implications for perceived voluntariness) are the two caveats to carry.

Sources