CBT-Based Conversational Agents for Depression: Effects on Symptoms and Engagement Outcomes

Background and Purpose

Depression is a leading cause of disability worldwide, yet access to evidence-based treatment remains limited due to cost, stigma, and workforce shortages. CBT-based conversational agents (CAs) offer a scalable, low-intensity approach to expanding access to mental health support. Although early studies suggest these interventions may reduce depressive symptoms, the evidence base remains fragmented, particularly regarding how engagement is conceptualized and measured. This systematic review synthesizes empirical evidence on the effectiveness of CBT-based conversational agents in reducing depressive symptoms and examines engagement patterns to inform implementation in social work practice.

Methods

A systematic review was conducted in accordance with PRISMA guidelines. Searches of PubMed, APA PsycINFO, and Scopus identified 405 records using terms related to conversational agents, cognitive behavioral therapy, and depression. After duplicate removal and screening, 333 titles and abstracts were reviewed, followed by 89 full-text articles. Eleven studies met the inclusion criteria. Eligible studies evaluated CBT-based conversational agents for adults with depressive symptoms and reported outcomes using validated measures. The final sample included randomized controlled trials (n = 8), systematic reviews/meta-analyses (n = 2), and one mixed-methods study. Data were extracted on study design, sample characteristics, intervention features, outcomes, and engagement indicators. A narrative synthesis was conducted to identify cross-study patterns.

Results

Across studies, CBT-based conversational agents consistently produced short-term reductions in depressive symptoms, with effect sizes ranging from small to moderate. Interventions demonstrated comparable or stronger outcomes relative to waitlist and informational controls, with mixed findings when compared to active self-help conditions. The majority of studies reported engagement-related constructs, including adherence, message frequency, satisfaction, usability, and therapeutic alliance; however, inconsistent operationalization across studies limited comparability. Evidence suggests that higher engagement and stronger therapeutic alliance may be associated with improved outcomes, although this relationship was not systematically tested. Design features such as personalization, empathic response, social cues, and cultural adaptation were associated with improved user experience and, in some cases, stronger outcomes. However, the evidence base is constrained by short intervention durations, homogenous samples, and limited use of standardized engagement measures.

Conclusions and Implications

CBT-based conversational agents show promise as accessible tools for reducing depressive symptoms and expanding access to care. However, inconsistent measurement of engagement represents a key barrier to understanding mechanisms of change and to scaling these interventions effectively. For social work, these findings highlight the potential of digital tools to reduce disparities in access while emphasizing the need for ethical implementation, integration within existing service delivery systems, and critical evaluation of digital interventions. Future research should prioritize standardized engagement metrics, longitudinal designs, and more diverse samples to strengthen the evidence base and support equitable, system-level integration of conversational agents into mental health care systems.