What Shapes Learners’ Trust in AI? A Meta-Analytic Review of Its Antecedents and Consequences

With the growing integration of artificial intelligence (AI) into higher education, learners’ trust in AI systems has attracted increasing attention. This meta-analysis synthesized 78 independent effect sizes from 42 empirical studies (N = 28 970) to examine key antecedents and consequences of AI trust. Four categories of antecedents were assessed: technological performance, user experience, ethics and safety, and sociodemographic and individual differences. Trust in AI was also analyzed in relation to behavioral intention, learning outcomes, and AI acceptance. Results showed that technological performance was the strongest predictor of trust, followed by user experience and individual differences, while ethics and safety had a relatively weaker impact. Trust in AI significantly predicted all three outcomes, underscoring its mediating role. Moderator analyses revealed that high-level feedback enhanced the effects of technological performance and ethics on trust, and that trust was generally higher in learning-support tasks than in academic writing tasks. These findings offer theoretical and practical insights for optimizing educational AI design and promoting adaptive, trustworthy systems.