The influence of generative artificial intelligence support on learning anxiety and academic expectation stress: the mediating role of self-efficacy

With the rapid integration of Generative Artificial Intelligence (GAI) into higher education, growing attention has been paid to its influence on students' psychological well-being. While prior studies have focused mainly on cognitive benefits, less is known about GAI's emotional regulation effects. Grounded in Social Cognitive Theory, this study constructed a structural model to examine how GAI support affects learning anxiety and academic expectation stress, with self-efficacy as a potential mediator. Survey data from 1,462 university students were analyzed using PLS-SEM. Results revealed that GAI support significantly and negatively predicted both learning anxiety and academic expectation stress. Self-efficacy partially mediated these relationships, indicating that students who engaged more frequently with GAI tools experienced lower stress and anxiety through enhanced confidence and perceived competence. The findings highlight a “psychological benefit” pattern associated with GAI use. Theoretically, this study extends Social Cognitive Theory to AI-supported learning contexts by demonstrating how environmental affordances (GAI support) influence affective outcomes via cognitive mechanisms (self-efficacy). Practically, it suggests integrating GAI responsibly into educational design to strengthen students' academic resilience, emotional regulation, and well-being.