Self-reported AI literacy and gendered self-efficacy: Evidence from Chinese adolescents

Artificial intelligence (AI) literacy has emerged as a key dimension of youth development, with profound implications for education, labor markets, and social equity. Yet, disparities in AI literacy risk amplifying existing inequalities. This study examines gender differences in self-reported AI literacy among Chinese adolescents, conceptualizing self-reported AI literacy as a reflection of self-efficacy. Drawing on Bandura's self-efficacy theory and a developmental-contextual perspective, we construct a multidimensional AI literacy index and employ counterfactual analysis together with Blinder–Oaxaca decomposition to validate and decompose the gender gap. Using a stratified sample of adolescents from diverse backgrounds in China, the study reveals a gender gap in self-reported AI literacy, with males rating themselves higher than females. This gap is heterogeneous across different dimensions of AI literacy and is most pronounced in the dimensions of Understanding, Ethics, and Society. The results further indicate that these differences are attributable both to perception biases and to actual disparities in observed background endowments. The findings highlight the risks of neglecting gendered self-efficacy gaps, which may perpetuate female exclusion from STEM opportunities, and underscore the importance of interventions that strengthen girls' AI self-efficacy. This research makes both theoretical and empirical contributions: it provides insights into the conceptualization and measurement of AI literacy, investigates the connection between AI literacy and self-efficacy, and offers policy-relevant implications for fostering more inclusive and equitable AI education.