An AI Chatbot in the Reduction of Anxiety in a Community Mental Health Center

Molina-Martinez F.A.; Garcia-Martinez G.E. · Publicaciones de la Facultad de Educacion y Humanidades del Campus de Melilla (2026) · DOI: 10.30827/publicaciones.v56i1.36861

Resumen

Introduction: Anxiety is one of the fastest-growing mental health problems worldwide. In this context, digital tools may help improve access to mental health support, especially in underserved communities or peri-urban settlements with limited access to healthcare services. Method: quasi-experimental single-arm pre–post study was conducted with 30 adults presenting anxiety symptoms who used an emotional support chatbot for six weeks. Anxiety was assessed before and after the intervention using the Zung Self-Rating Anxiety Scale, and usability was measured with the System Usability Scale (SUS). Descriptive statistics (mean, standard deviation, percentages, and 95% confidence intervals) were calculated. Pre–post differences were analyzed using a paired-samples t-test with effect size estimation (Cohen’s dz). Internal consistency was examined using Cronbach’s alpha. Open-ended responses were analyzed through thematic review and word-frequency analysis. Results: Before the intervention, 36.7% of participants presented with severe anxiety, 30.0% with mild anxiety, 20.0% with moderate anxiety, and 13.3% with normal levels. After the intervention, 53.3% reached normal levels of anxiety, while severe and moderate cases decreased to 16.7% and 13.3%, respectively. The overall reduction in anxiety symptoms was 50%. Regarding usability, the system was mostly rated as acceptable to excellent. Conclusions: The Florence AI chatbot was associated with reductions in self-reported anxiety symptoms and was generally perceived as an accessible and acceptable source of emotional support among adults receiving care in a community mental health setting. However, because of the single-arm design, small sample size, and gender imbalance, these findings should be interpreted as preliminary associations rather than evidence of efficacy. Larger controlled studies with longer follow-up periods are needed to determine the effectiveness, safety, and generalizability of AI-based emotional support interventions.

Instituciones: Universidad Católica de Santa María