Abstract
Mental health is a critical determinant of individual well-being, societal contribution, and economic development. Despite its importance, access to effective treatment is constrained by stigma, cost, availability of professionals, and geographic disparities. Recently, arti-ficial intelligence (AI)-based tools have emerged as a complementary approach, offering scalable, personalized, and accessible interventions. Yet, limited empirical evidence exists on how traditional and AI-based approaches can be integrated into a unified framework of care. This study addresses this gap by empirically examining the extent to which traditional and AI-based methods contribute to an integrated mental healthcare model. A quantitative, cross-sectional survey was conducted among 152 participants in Gauteng Province, South Africa, chosen for its demographic diversity. Partial least squares structural equation mod-eling (PLS-SEM) was applied to assess the measurement and structural models. Construct reliability was established with Cronbach’s alpha values above 0.70 and composite reliability values exceeding 0.716. Convergent validity was supported for most constructs, with average variance extracted (AVE) values above 0.50, and discriminant validity was confirmed with heterotrait-monotrait (HTMT) ratios below 0.90. Results show that AI-based methods exert a substantially stronger effect on integration (β = 0.795) relative to traditional methods (β = 0.159). The model explained 72.7% of the variance (R² = 0.727), with satisfac-tory fit indices (SRMR = 0.067; d_ULS = 0.767; d_G = 0.476; NFI = 0.792). This study advances theoretical and practical understanding of integrated mental healthcare by offering a validated model that high-lights the complementary, though asymmetrical, contributions of tradi-tional and AI-based approaches.