Dynamic pathways in the development of university students’ AI literacy: integrating quantitative and qualitative evidence

Introduction This study investigates the formative mechanisms underlying university students’ artificial intelligence (AI) literacy, focusing on the complex interrelationships among external support, AI self-efficacy, AI anxiety, and AI literacy. Methods An integrated framework was developed and tested using a mixed-methods approach, combining structural equation modeling (SEM) with qualitative interviews of 15 high-anxiety, high-literacy (HAHL) students. Results SEM results indicated that external support positively predicted AI literacy ( β = 0.163, p < 0.001) and AI self-efficacy ( β = 0.454, p < 0.001), while also increasing AI anxiety ( β = 0.505, p < 0.001). Both AI self-efficacy ( β = 0.728, p < 0.001) and AI anxiety ( β = 0.527, p < 0.001) significantly and positively influenced AI literacy, highlighting that moderate anxiety may serve as vigilance-driven motivation to foster learning. Qualitative analyses further revealed that HAHL students translate anxiety into structured engagement through cognitive appraisal, resource accumulation, and professional identity reflection. Conclusion These results demonstrate the dynamic interplay of environmental, cognitive, and affective factors in shaping AI literacy, and show how qualitative insights complement quantitative SEM analysis by revealing mechanisms behind anxiety-driven engagement. The study provides evidence-based guidance for higher education institutions to enhance AI literacy through optimized support systems, fostering technological confidence, and strategically managing AI-related anxiety.