A Review of Large Language Models in Mental Health Chatbots for Emotional Support

Abstract

The integration of artificial intelligence (AI) into mental health care has sparked significant advancements, with large language models (LLMs) powering modern chatbots that offer scalable and accessible emotional support globally. This review provides a synthesis of recent literature on LLM-driven mental health chatbots, emphasizing their potential to address the growing burden of mental health challenges, especially in underserved populations where stigma, geographical barriers, and limited resources prevent access to traditional services. LLM-powered chatbots are designed to engage users in empathetic, context-aware, and supportive dialogues, aiding in mood tracking, anxiety management, crisis escalation, and psychoeducation, while simultaneously promoting privacy and user empowerment. Despite their promise in democratizing mental health interventions, several challenges persist, such as the opaque “black box” nature of these models that raises concerns regarding interpretability, bias, data privacy, and ethical application. Ensuring transparency in model governance, establishing clear boundaries of chatbot functionality, and mechanisms for prompt escalation to human professionals remain critical imperatives for responsible deployment. This review underscores the importance of multidisciplinary collaboration and systematic evaluation to inform best practices and regulatory frameworks for AI-driven mental health solutions. Ultimately, LLM-based chatbots should be considered complementary supports that bridge the gap between individuals struggling with emotional difficulties and professional care, rather than substitutes for trained clinicians. Advances in model explainability, stakeholder engagement, and inclusive design are key to fostering trustworthy, safe, and effective AI applications in mental health care. The future of AI in this field hinges on proactive efforts to mitigate risks and maximize benefits, ensuring that these digital tools contribute meaningfully to reliable and equitable mental health support for diverse populations worldwide.