Source-linked AI summary
A Scoping Review of AI-Driven Digital Interventions in Mental Health Care: Mapping Applications Across Screening, Support, Monitoring, Prevention, and Clinical Education
Yang Ni, Fanli Jia
TL;DR
Access barriers and fragmented evidence motivate a broad examination of AI-driven digital interventions in mental health care. This PRISMA-ScR scoping review maps modalities and applications across five clinical phases using 36 empirical studies. It identifies expanding access, symptom monitoring, and personalized intervention support alongside persistent bias, privacy, and integration challenges.
Problem
Mental health care faces access barriers, while prior reviews often focus on single AI technologies, clinical phases, or conditions.
Method
The PRISMA-ScR scoping review maps conversational agents and predictive or monitoring models across five clinical phases and synthesizes empirical findings narratively.
Results
Thirty-six studies mapped AI contributions across screening, treatment, monitoring, education, and prevention, including access expansion, symptom monitoring, and personalized interventions.
Takeaways & Limitations
Responsible integration requires ethical design, transparent model development, human oversight, and collaboration among clinicians, developers, policymakers, and patients.
Takeaways & Limitations
The review excludes studies released after January 2024, may miss relevant literature, and is limited to English-language publications.
Abstract
from arXiv · showhide
Artificial intelligence (AI)-enabled digital interventions, including Generative AI (GenAI) and Human-Centered AI (HCAI), are increasingly used to expand access to digital psychiatry and mental health care. This PRISMA-ScR scoping review maps the landscape of AI-driven mental health (mHealth) technologies across five critical phases: pre-treatment (screening/triage), treatment (therapeutic support), post-treatment (remote patient monitoring), clinical education, and population-level prevention. We synthesized 36 empirical studies implemented through early 2024, focusing on Large Language Models (LLMs), machine learning (ML) models, and autonomous conversational agents. Key use cases involve referral triage, empathic communication enhancement, and AI-assisted psychotherapy delivered via chatbots and voice agents. While benefits include reduced wait times and increased patient engagement, we address recurring challenges like algorithmic bias, data privacy, and human-AI collaboration barriers. By introducing a novel four-pillar framework, this review provides a comprehensive roadmap for AI-augmented mental health care, offering actionable insights for researchers, clinicians, and policymakers to develop safe, effective, and equitable digital health interventions.
1 Introduction
Mental health treatment access remains constrained by stigma, cost, and professional shortages, while digital technologies and AI create opportunities to extend support. This review asks how AI modalities function across clinical phases, what impacts and barriers they present, and which technologies appear most mature or promising.
- Stigma, cost, and shortages of mental health professionals continue to hinder treatment access despite broader economic and technological progress.
- AI-driven mental health interventions include conversational agents, NLP systems, predictive ML/DL models, and transformer-based LLMs.These systems can parse input, detect sentiment, classify risks, and generate context-rich text.
- The review examines which AI modalities power interventions within each clinical phase and what evidence exists regarding efficacy and limitations.
- It also compares reported impacts, performance metrics, and barriers across AI-driven tools and clinical phases.
- The review investigates strengths, weaknesses, opportunities, threats, technological maturity, emerging research priorities, and technical, clinical, and policy challenges.
- By linking conceptual insights with empirical outcomes, the review aims to provide an integrated reference for research, practice, and policy.
2 Methods
The review maps AI technologies in mental health care using a PRISMA-ScR framework across five clinical phases, drawing on empirical literature and narrative organization. Its design emphasizes broad coverage of applications, outcomes, limitations, and practical integration considerations.
- AI technologies in mental health include chatbots, language models, prediction models, sentiment analysis, and recommender systems.
- The review aims to maximize AI benefits and minimize risks while maintaining standards of care and patient-centered integration.
- PRISMA-ScR methods covered conversational agents and predictive or monitoring models across screening, treatment, follow-up, education, and prevention.
- The search covered empirical English-language research through January 2024 using multiple databases and AI–mental health search terms.
- 1674 records yielded 146 records for further evaluation and 143 retrieved full-text reports after duplicate removal and eligibility screening.
- Thirty-six studies met inclusion criteria after excluding non-empirical, ineligible preprint, and out-of-scope publications.
- A customized charting form organized study objectives, technologies, settings, outcomes, limitations, and clinical phases for synthesis.
3 Results
The review synthesized 36 studies spanning five mental health care phases and multiple AI modalities. Results were organized by technology and clinical function, then interpreted through narrative synthesis because study designs and objectives were heterogeneous.
- Thirty-six reviewed articles covered user, student, and professional perceptions while evaluating efficacy of AI applications.
- The mapped modalities included chatbots, conversational agents, NLP, LLMs, ML, DL, and AI-based prediction systems.
- Rule-based chatbots were prominent in screening, NLP agents in empathic support, ML/DL models in post-treatment risk assessment, and LLM agents in multi-turn counseling.
- Applications spanned screening and triage, therapeutic support, monitoring and follow-up, prevention, and clinical education.
- Functions included assessment, diagnosis, monitoring, outcome prediction, counseling, therapy, clinical decision support, and general mental health assistance.
- Narrative synthesis identified overarching themes, patterns, discrepancies, and future directions across heterogeneous studies.
3.3 Key Findings in AI Technologies in Mental Health Care
This section catalogs the main AI modality classes and summarizes the empirical outcomes and limitations reported for each algorithm class.
- The review catalogs rule-based chatbots, traditional NLP agents, ML/DL predictive models, and LLM-based agents.
3.3.1 Applying Natural Language Processing
AI-driven mental health interventions use NLP, ML, and deep learning across screening, treatment, monitoring, and broader service delivery. Evidence highlights conversational support, predictive personalization, operational improvements, and persistent concerns about bias and human-AI preferences.
- NLP-powered chatbots and agents support mental health care through interactive text and voice conversations, including emotion detection and empathy training.Examples include HAILEY, which analyzes user text to detect emotions and train empathic responses.
- LLM-based conversational AI may expand access by understanding language, identifying emotions, tailoring interventions, and offering guideline-aligned advice.The review also notes that these capabilities remain preliminary and require attention to model limitations.
- ML and deep learning models support prediction, diagnosis, treatment personalization, emotional-state monitoring, and operational decisions such as caller-counselor matching.Deep learning extends these applications to real-time monitoring and individualized online CBT and art psychotherapy.
- Across the four-pillar framework, AI supports assessment and referral, therapeutic intervention, remote monitoring, and proactive population-level mental health support.The framework also describes benefits for patients, clinicians, health organizations, and the general public.
- Limbic Access automates symptom and demographic intake using PHQ-9 and GAD-7, reducing assessment and treatment waits while lowering dropout and increasing recovery.The tool was evaluated against traditional assessment methods in NHS Talking Therapy referrals.
- AI agents can provide moderate-to-high-efficacy therapy and counseling, but users still prefer human responses over AI empathy.The evidence links emotional disclosure by agents with higher satisfaction while preserving a preference for human interaction.
4 Discussions
AI-driven mental health interventions show broad potential across support, clinical assistance, prevention, and population-level care, but their integration remains constrained by privacy, bias, oversight, and limited real-world evidence. The review therefore emphasizes human-AI collaboration, adaptive safeguards, and stronger implementation research.
- AI Modalities Applied Across Phases: AI commonly provides emotional support and clinical assistance through chatbots, while also supporting personalized treatment, recommendations, diagnosis, and outcome prediction.The review describes direct therapy and counselling automation as a potential route to improving accessibility, but frames current applications primarily as support for clinical work.
- AI-Clinician Collaboration: AI applications are currently viewed as complements to clinicians because accuracy risks make standalone diagnosis or therapy unsuitable at present.Effective deployment depends on dividing responsibilities, integrating tools into workflows, and defining human oversight obligations.
- Strengths, Weaknesses, Opportunities, and Threats: AI can expand access through first-line support, personalize counseling, uncover patterns in large datasets, and scale mental health assistance across many users.Examples include Limbic Access, HAILEY, machine-learning treatment prediction, and the LLM-based ChatCounselor.
- Strengths, Weaknesses, Opportunities, and Threats: Privacy and security risks, model misinterpretation, uncertain diagnostic precision, and limited recognition of mental disorders remain major weaknesses requiring human oversight.The review also questions whether conversational agents consistently foster empathy or engage users effectively.
- Advancing Mental Health Prevention and Improvement: Population-level prevention may benefit from emotional-support chatbots, behavioral apps, predictive screening, wearable monitoring, and personalized early interventions.These tools could support services tailored to specific populations, including students, older adults, and people with substance use concerns.
- Policy Implications: Policy should combine secure data handling, transparency, clinical validation, human oversight, innovation pathways, and demographic performance checks.The proposed approach pairs safeguards with regulatory sandboxes, adaptive licensing, rapid ethics review, and inclusive training data.
- Limitations: The review’s conclusions are limited by its January 2024 cutoff, English-only search, heterogeneous evidence, and many prototype or single-site studies with uncertain routine-practice generalizability.These constraints prevented quantitative synthesis and leave implementation factors, sustainability, and user diversity insufficiently characterized.
5 Conclusions
This scoping review synthesized 36 studies mapping AI-driven mental health interventions across five clinical phases. It broadens prior work by integrating multiple AI modalities and identifies priorities for safe, equitable future evaluation.
- 36 studies were mapped across screening, treatment, monitoring, clinical education, and prevention in mental health care.
- The review identified contributions of chatbots, natural language processing, machine learning, deep learning, and large language models to access, symptom monitoring, and personalized interventions.
- Unlike reviews focused on single technologies, clinical phases, or conditions, this review synthesizes multiple AI modalities across five clinical phases.
- The review connects conceptual insights with empirical outcomes to support research, practice, and policy across the mental health care continuum.
- Future research should use multi-site and longitudinal evaluations in diverse populations and assess safety, privacy, and equity impacts.
Funding
The research received no external funding.
- The research received no external funding.