Trust in LLM-based Conversational Agents for Mental Health Support: A Systematic Review

This project is a systematic review examining user trust in large language model (LLM)-based conversational agents used for mental health, psychological, or emotional support. These systems are increasingly being used in contexts involving sensitive disclosure, uncertainty, and vulnerability, yet trust remains inconsistently conceptualized and operationalized across the existing literature. The purpose of this review is to synthesize empirical studies examining how trust in LLM-based conversational agents for mental health support is conceptualized, operationalized, and investigated; which factors are associated with the formation and development of trust; and which outcomes are associated with trust. The review includes quantitative, qualitative, and mixed-methods empirical studies involving human participants and LLM-based conversational agents in which a large language model substantively contributes to the generation of user-facing, free-form, context-sensitive conversational responses. A systematic search of multiple electronic databases will be conducted using predefined eligibility criteria. Findings will be synthesized using narrative, descriptive, and thematic approaches, with exploratory analyses examining variation across AI roles characterized by different levels of autonomy and emotional engagement. The review is intended to provide an integrated overview of the current evidence on trust in LLM-based conversational agents for mental health support and to inform future research, system design, evaluation, and governance in this emerging field.