Ethical evaluation of AI-supported mental health applications

Key Findings

• — Four shipped applications were analysed across fourteen categories: Wysa, Youper, Amaha and Yuna.
• — The app analysis identified gaps in transparency, accountability, crisis detection and clinical accuracy.
• — Five practising psychiatrists were interviewed, analysed by grounded theory, producing four themes.
• — The themes were ontological mismatch with psychiatry, clinical role displacement, limits of digital empathy, and clinicians' own emotional responses to AI.
• — Findings were interpreted through psychiatric-ethics norms together with the Beauchamp and Childress principles.

Methodology Notes

Mixed design in which the app-audit arm carries the weight; the interview arm has only five participants and is exploratory. Category-level assessment of four products at a point in time, with no response-level testing and no denominators. Published 2026-08-28 in Frontiers in Psychiatry, CC BY 4.0, DOI 10.3389/fpsyt.2026.1870948; metadata and full abstract verified at Crossref.

Sources