The effectiveness of AI chatbots in reducing anxiety and depression symptoms, and feelings of loneliness: A systematic review and meta-analyses (Preprint)
BACKGROUND Globally, there is a high prevalence of mental health conditions, but access to timely and appropriate care is limited by accessibility, stigma, wait times, and worker shortages. Artificial intelligence (AI) chatbots are an emerging technology offering an alternative to traditional forms of mental health support. OBJECTIVE The aim of this paper is to assess the effectiveness of AI chatbot interventions in reducing symptoms of anxiety and depression, and feelings of loneliness. METHODS PubMed, CINAHL and Embase databases were searched in April 2025 for studies published in peer-reviewed journals in English language since 2022. The search strategy included terms for mental health and AI. Eligible studies included a text-based synchronous AI chatbot as the main intervention for participant screening, diagnosis, or treating symptoms of anxiety or depression, or feelings of loneliness. Additionally, studies needed to: use a validated psychometric scale to measure effectiveness; include participants aged 16 years and over; and have a focus other than prenatal or perinatal mental health. Screening and selection were conducted by two of the authors. Eligibility disagreements were resolved by a third author. Data extraction was performed by one author and checked for data integrity by a second author. Meta-analyses of mean changes were performed for studies reporting results using the same psychometric scale, study method, and comparators. The primary outcomes were results of validated psychometric scales measuring symptoms of anxiety or depression, or feelings of loneliness. Results were compared at baseline and post-intervention. Study details and demographics were also reported. RESULTS Thirty studies met eligibility criteria and were included for review. Statistically significant improvement was found in 67% of studies measuring depression and 56% of studies measuring anxiety, but only 25% of studies measuring loneliness. There was a small mean reduction in symptoms for all conditions but these are not considered clinically significant. CONCLUSIONS When measured using psychometric scales, AI chatbots may be effective in reducing anxiety and depression symptoms in some cases, and have the potential to be used in conjunction with other forms of mental health care for these conditions. They are less effective in reducing feelings of loneliness. According to present findings, chatbots are not consistently effective enough to be recommended as a stand-alone intervention for symptoms of anxiety and depression, or feelings of loneliness. CLINICALTRIAL PROSPERO ID: CRD420250651616