Problem

The increasing integration of AI in vocational education necessitates an understanding of how institutional AI readiness translates into student AI literacy. Current research is fragmented, lacking empirical evidence that directly links school-level AI capacity to measurable student competencies. There's a critical need for a multilevel explanatory model that traces how institutional readiness influences instructional conditions through teacher mechanisms and ultimately shapes student AI literacy, especially given varying regional AI development contexts.

Method

This study employed a 2-2-1 cross-level mediation framework using a large-scale national survey from China. The dataset included 1,007 vocational institutions, 156,125 teachers, and 2,379,546 students. School-level AI readiness was measured across six dimensions (strategic, organizational, process, technological, data, ethical). Teacher mechanisms, such as perceived AI capability and experiential orientations, were aggregated to the school level. Student AI literacy was the individual-level outcome. Multilevel linear mixed-effects models were used, accounting for within-school dependency and assessing direct, indirect, and contextual effects, with robustness checks including Monte Carlo simulations and analyses across regional AI development levels.

Results

Overall school AI readiness was positively associated with student AI literacy. Individually, all six readiness dimensions showed positive associations, but simultaneous modeling suggested they operate as an integrated configuration. Aggregated teacher-perceived AI capability partially mediated the relationship between institutional readiness and student AI literacy, meaning that while institutional readiness directly contributes, teachers' collective belief in their AI instructional ability is a crucial transmission mechanism. This readiness-capability-literacy pathway demonstrated structural robustness across heterogeneous regional AI development contexts. Regional AI development independently influenced baseline literacy levels but did not significantly alter the strength of the readiness-literacy association.

Implications

The findings theoretically demonstrate that institutional AI readiness functions as a multilevel ecological process, influencing student outcomes through collective teacher capability, extending ecological systems theory and organizational support theory to AI in education. Practically, they highlight that institutional AI readiness must be an integrated, coordinated effort encompassing infrastructure, governance, and sustained investment in teacher professional development. Policies should prioritize building teachers' instructional AI capability over general attitudinal acceptance. The robustness across regions suggests that strengthening internal school-level organizational coherence and teacher capacity can be a scalable strategy for improving AI literacy and potentially narrowing regional disparities, even in less developed AI ecosystems.