Artificial Intelligence-Driven Tools in Mental Health Service Delivery: A Scoping Review

Background: Artificial intelligence (AI) holds transformative potential for mental health services. However, existing reviews have predominantly focused on algorithmic accuracy, with limited attention to how these technologies are implemented and integrated into real-world service delivery. This scoping review addresses this gap by examining the contexts in which AI technologies-including large language models (LLMs) and machine learning-are implemented, as well as the factors influencing their sustainable adoption within real-world mental health service systems. Methods: Following the established methodological framework, a systematic search (2015-2026) was conducted in PubMed and Scopus. Two independent reviewers screened an initial pool of 829 records using Zotero and Rayyan to minimize selection bias. Following title, abstract, and full-text screening based on predefined eligibility criteria, 26 studies focusing on real-world AI applications (e.g., clinical settings, community services, and case management) were included in the final synthesis. Results: The findings indicate a rapid acceleration in research, with 50% of included studies (n = 13) published since 2024. AI-driven decision support systems were the most prevalent (50%, n = 13), followed by predictive machine learning models (27%) and generative AI applications (15%). Most tools were designed for clinician use (77%) and implemented in hospital-based settings (46%). Although 46% of studies reported real-world implementation, more than half remained at the pilot stage. Notably, research emphasis has shifted from technical efficacy toward feasibility, and implementation contexts (n = 17). Conclusion: AI in mental health is transitioning from laboratory validation to real-world integration. However, the current landscape remains heavily centered on clinician workflows and screening functions, with limited expansion into community-based recovery and long-term prevention. To move beyond the pilot stage, future initiatives should prioritize seamless workflow integration and the application of structured ethical and implementation frameworks that support clinician-patient relationships. This review provides an evidentiary basis for advancing sustainable, AI-enhanced mental health service delivery.