Abstract
BACKGROUND Artificial intelligence (AI) is increasingly integrated into digital mental health interventions, including mobile health (mHealth) applications, telehealth platforms, and AI-augmented cognitive behavioral therapy (CBT). Despite rapid development, comprehensive large-scale evidence evaluating both clinical effectiveness and algorithmic performance remains limited OBJECTIVE This study systematically reviews and meta-analyzes AI-driven digital mental health research indexed in the Scopus database between 2013 and 2025, assessing clinical outcomes and machine learning performance metrics. METHODS A systematic search of Scopus was conducted using predefined TITLE-ABS-KEY search strings related to digital mental health, artificial intelligence, and machine learning. Following PRISMA 2020 guidelines, 2,042 records were identified, and 1,008 duplicates were removed. After screening 1,034 articles and assessing 425 full-text studies, 385 met inclusion criteria for qualitative synthesis and meta-analysis. Effect sizes were calculated using Hedges’ g under a random-effects model. Algorithmic performance was synthesized using pooled area under the curve (AUC), sensitivity, and specificity. RESULTS AI-driven interventions demonstrated statistically significant improvements in depression, anxiety, and mood outcomes, with moderate pooled effect sizes. Machine learning models showed strong predictive performance, with pooled AUC values generally above 0.80. However, substantial heterogeneity and reporting variability were observed. CONCLUSIONS AI-enabled digital mental health interventions provide clinically meaningful benefits and robust predictive capabilities, though greater methodological standardization and algorithmic transparency are required. CLINICALTRIAL -

This publication has 33 references indexed in Scilit: