Introduction.
Academic anxiety is a widespread concern among university students, negatively affecting motivation, academic performance, and psychological well-being. Recent advances in Artificial Intelligence (AI) — including generative models such as ChatGPT, adaptive tutoring systems, and predictive algorithms — have opened new opportunities for emotional regulation and cognitive support in higher education. However, despite the growing number of empirical studies, no previous meta-analysis has quantitatively assessed the effectiveness of AI-based interventions in reducing academic anxiety and enhancing self-efficacy.
Purpose.
This meta-analysis aimed to evaluate the overall efficacy of AI-based interventions in improving emotional and cognitive outcomes among university students, with particular attention to the reduction of academic anxiety, the enhancement of motivation, and perceived self-efficacy.
Method.
Following PRISMA guidelines, 20 quantitative studies published between 2022 and 2025 were systematically identified and analyzed (total N ≈ ). A random-effects model (DerSimonian-Laird method) was applied, using Hedges' g as the standardized effect size. Heterogeneity was assessed through Cochran's Q and the I² index, and meta-regressions explored the moderating roles of AI type, study design, and intervention duration.
Results.
The overall effect size was g = 0.64 (95% CI [0.55, 0.74], p < .001), indicating a moderate-to-high positive effect of AI interventions on learning-related and emotional outcomes. Heterogeneity was moderate (I² = 61%), and publication bias appeared limited. Predictive machine learning models and integrated AI platforms yielded the strongest effects, while conversational AI tools showed smaller but consistent improvements.
Conclusions.
AI-based tools demonstrate significant potential to reduce academic anxiety and enhance students' motivation, engagement, and self-efficacy. These findings highlight the relevance of AI as a complementary instrument for promoting psychological well-being and adaptive learning in higher education settings. Future research should investigate longitudinal effects, differential impacts across disciplines, and ethical dimensions of AI use in psychological and educational interventions.