Effects of AI Conversational Agents on Stress Reduction: A Meta-Analysis (2015- 2025)

Abstract

Between 2015 and 2025, researchers conducted randomized and quasi-experimental trials to investigate whether AI chatbots can reduce perceived stress in adults. These interventions typically delivered CBT, mindfulness, or positive psychology techniques through mobile or web-based platforms. Stress was most commonly assessed using the Perceived Stress Scale (PSS); few studies included physiological measures such as HRV. A review of 34 studies (29 RCTs, 5 quasi-experiments) across diverse populations, such as students, healthcare workers, older adults, and individuals with chronic illnesses, revealed that most interventions lasted between 1 and 16 weeks. About half of the trials reported significantly greater stress reductions in chatbot users compared to controls, generally with small to moderate effect sizes. The remaining studies showed no significant differences. Several studies also reported meaningful improvements in anxiety, depression, or coping self-efficacy, even when stress effects were minimal. The findings suggest that the efficacy of chatbots for stress reduction is mixed and context-dependent, likely moderated by factors such as population, intervention design, and engagement level. Chatbots are seen as promising, scalable tools for stress management; however, future trials should include standardized outcomes, objective stress markers, and longer follow-up periods to better understand their sustained impact.

Keywords: Artificial Intelligence, Stress, Chatbots

DOI: 10.54941/ahfe1007093