AI-powered methods for psychological assessment in adolescence psychological disorders: A systematic review and...

Artificial intelligence (AI)-powered assessment, with its ability to process multimodal data and support real-time evaluation, is transforming traditional psychological assessment. However, the predictive performance of AI-powered assessments, such as the area under the curves (AUCs), and their clinical applicability remain unclear. Adolescence, marked by heightened neuroplasticity and psychological vulnerability, demands more precise approaches to assessment and intervention to promote mental health. This review developed a multimodal assessment framework that integrates both modality type and disorder type, and conducted a large language model (LLM) assisted meta-analysis in adolescent populations. Eight databases were systematically searched from inception to December 12, 2025. Studies that used AI-powered assessments of adolescent mental outcomes were included. A total of 188 studies involving 4,794,001 participants (aged 10-19 years) met the inclusion criteria. In the meta-analysis, we included studies that determined mental health outcomes using gold-standard assessments (e.g., clinical interviews) or self-reported behavioral indices (k = 154). The mean AUCs ranged from 0.75 to 0.85 under internal validation (performance tested within the same dataset used for model development) and 0.76 to 0.96 under external validation (performance tested in a new, independent dataset) across various mental health outcomes, indicating good discrimination. Subgroup analyses showed that multi-modal achieved higher AUCs than single-modality models for depression and suicide. AI holds promise for adolescent mental health assessment, with performance varying by data modality and mental health outcome. These findings may inform the development of AI-driven clinical decision support systems. However, integration into routine care requires staged, externally validated, and reciprocal clinical transition approaches that align with real-world clinical workflows. It is also essential to enhance clinical interpretability, deployment, ethical governance, and cultural adaptation to ensure equitable access to mental health resources.