Answers are grounded in this paper and cite their sources.
Background: Artificial intelligence-based chatbot interventions have been identified as potential solutions to address disparities in the increasing global prevalence of mental disorders, including anxiety and depression. This systematic review and meta-analysis aimed to evaluate the effectiveness of AI-based interventions on well-being, with well-being measured through a combination of depression and anxiety symptom scales and affective instruments. These measures captured both negative and positive aspects of mental health, including tools such as the Patient Health Questionnaire-9 (PHQ-9), Generalized Anxiety Disorder-7 (GAD-7), and the Positive and Negative Affect Schedule (PANAS). Methods: We searched PubMed, ProQuest, Web of Science, PsycINFO, IEEE Xplore, and Scopus databases. Randomized controlled trials and reported outcomes using validated psychometric instruments (PANAS, GAD-7, or PHQ-9) were included. Data extraction and risk of bias assessment (RoB 2) were independently completed by two reviewers. Standardized mean differences (SMDs) with Hedges’ g correction to adjust for small sample bias were implemented to execute meta-analyses. Results: Sixteen studies were incorporated. A small but statistically significant reduction in depressive symptoms was observed in AI-based interventions, predominantly chatbot-based (Hedges’ g = −0.38; 95% CI: −0.60, −0.16), with substantial heterogeneity (71.2%; p < 0.001). Anxiety (GAD-7: Hedges’ g = −0.25, 95% CI: −0.63 to 0.14) and affect (PANAS-P: SMD = 0.13; PANAS-N: SMD = −0.41) did not exhibit any statistically significant effects. Conclusions: AI-based chatbots showed a small effect on depressive symptoms, while effects on anxiety and affective outcomes were not statistically significant. Given the low to very low certainty of evidence, substantial heterogeneity, and short follow-up, these findings remain uncertain and require confirmation in high-quality RCTs with longer follow-up.
No takes yet. Share an insight, caveat, or question.
Al-Garni et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: