Self-Efficacy over Trust: Rethinking Sustainable AI Use among University Students

Purpose – Artificial intelligence (AI) is increasingly embedded in higher education, yet sustained use remains insufficiently explained by adoption-centered models that primarily emphasize direct effects.This study examines continuous AI use among university students by integrating perceived enjoyment, effort expectancy, trust in artificial intelligence, and AI self-efficacy within a parallel-mediation framework. Methods – A quantitative cross-sectional design was employed involving 247 undergraduate students who reported varying levels of AI use in academic contexts. Data were collected through an online questionnaire and analyzed using partial least squares structural equation modeling (PLS-SEM) to assess direct and indirect relationships. Findings – Perceived enjoyment (β = 0.386) and effort expectancy (β = 0.363) significantly predicted continuous AI use, with perceived enjoyment showing the stronger direct effect. AI self-efficacy significantly mediated both relationships, particularly the effect of effort expectancy (β = 0.297). By contrast, trust significantly mediated only the perceived enjoyment pathway (β = 0.196), while its mediating role in the effort expectancy pathway was not supported under the conventional significance criterion. Research Implications -The study suggests that higher education institutions should strengthen students’ AI self-efficacy while designing AI-supported learning environments that are both intuitive and engaging. Originality - The study contributes to post-adoption AI research by clarifying the relative mediating roles of self-efficacy and trust, highlighting that AI self-efficacy is a more robust mechanism than trust in sustaining AI use among university students.