Who Is Represented in Evaluations of AI Mental-Health Tools for Young People? A Structured Audit Informed by Pandemic-Era Youth Research

Claims that an AI mental-health tool is safe or useful “for youth” depend on who participated and who was excluded. We audited eight peer-reviewed evaluations published from 2022 through August 2026, alongside two population studies of youth use and one related participatory-design study. Five evaluations involved young users; three relied on researcher-written personas or prompts. Six reported gender or sex, and two reported sexual orientation. None reported disability or neurodivergence identity beyond the target health condition. None analyzed tool outcomes by queer, disabled, or neurodivergent identity. Three of the five direct-participant evaluations excluded youth based on severe symptoms, suicidality, psychosis, concurrent care, English fluency, or device access. Five studies evaluated generative systems, but three of them simulated youth. Three of the five direct-participant studies evaluated constrained agents. Meanwhile, 2026 surveys found affective or mental-health chatbot use among 19.2% of U.S. youth aged 12–21 and 21.1% of Ontario students; the Ontario study found higher use among gender-diverse students. We propose a short reporting standard covering evidence source, identities measured, recruitment and access, exclusions, subgroup analysis, and youth involvement. This bounded pilot does not estimate prevalence across the field.