Assessment, intervention, and ethical considerations of artificial intelligence for depression and suicide risk: A scoping review.

Background

Depression and suicidal behavior constitute leading causes of global disease burden. Artificial intelligence (AI) offers promise for improving detection, intervention, and prevention, yet no comprehensive review has simultaneously mapped AI-driven assessment, intervention, and ethical considerations across both conditions.

Objective

To systematically map AI applications for depression and suicide risk across three thematic axes: (1) assessment and diagnosis, (2) conversational AI and AI-driven interventions, and (3) ethical, equity, and regulatory considerations, and to evaluate translational maturity using a novel framework.

Methods

Following JBI scoping review methodology and PRISMA-ScR guidelines, six databases were searched from January 2020 through March 2026. Two independent reviewers screened 5079 records using a two-stage process, yielding 2844 included studies (Cohen's kappa = 0.66). Data were extracted using a standardized charting form with AI-assisted full-text analysis.

Results

Assessment/diagnosis dominated (n = 1785), with fewer intervention studies (n = 318). Large language model-related studies rose from <1% (2020-2022) to 11.1% (2025) of included studies annually. Translational Readiness Framework classification revealed 44.8% Laboratory, 49.2% Bridge, and 6.0% Clinical stage, with intervention studies showing the highest Clinical-stage proportion (16.7% vs. 5.9% for assessment). Five cross-cutting observations emerged, including an inverted evidence pyramid and persistent cross-axis integration gap (4.4%).

Conclusion

Despite technical advances, the evidence base remains concentrated at early translational stages. The proposed Translational Readiness Framework provides a structured lens for evaluating progress and guiding research prioritization. Closing the gap between laboratory promise and clinical impact requires a shift from model-centric to patient-centric research and from benchmark performance to real-world validation.