AI chatbots and youth suicide risk: current evidence, critical gaps, and a clinical research agenda

Key Findings

• — Existing studies of chatbot responses to suicide-risk disclosure rely almost entirely on single-turn, researcher-generated prompts that were not designed with youth in mind
• — Chatbot performance degrades substantially with indirect, ambiguous or gradually disclosed risk — the disclosure patterns characteristic of adolescent communication
• — The real-world impact of chatbot interactions on youth suicide risk remains unknown; no prospective outcome research exists
• — The proposed four-priority agenda is: integrating youth perspectives through qualitative and participatory research; youth-specific risk detection and evaluation methods; prospective real-world outcome research; and improving connection of at-risk youth to crisis resources and trusted adults

Methodology Notes

Review article (npj Digital Medicine classes it as a Review, in the 'AI-Enabled Therapies in Mental Health' collection), published online 2026-08-01 under CC BY 4.0. It presents no original data — its load-bearing contribution is the methodological critique and the research agenda, and it should be graded as a position piece rather than as evidence. All four authors are at Research & Impact, Crisis Text Line, New York, which is what makes it a citable institutional position from a crisis-line operator rather than an ordinary literature summary; that affiliation is also a standpoint worth naming when citing it. Verified by fetching nature.com directly with a browser user agent for the citation metadata, with the full abstract read from the Crossref record; not yet indexed in PubMed or Europe PMC at time of verification.

Sources