Shuhua Hou, Xianhe Liu, Xiaoqing Shen, Chenyun Zhang, Dali Liu, Yang Wu
This study investigates whether English proficiency moderates core Technology Acceptance Model (TAM) pathways in the context of AI-enhanced English MOOCs for vocational students. Drawing on an extended TAM that links Perceived Ease of Use (PEOU), Perceived Usefulness (PU), Behavioral Intention (BI), and Perceived Learning Outcomes (PLO), we surveyed 516 learners from a provincial AI-powered MOOC. Confirmatory factor analysis confirmed strong measurement properties (all factor loadings > 0.74, AVE > 0.57, CR > 0.80). Structural analysis revealed robust direct effects: PEOU → PU ( β = 0.756), PU → BI ( β = 0.696), and BI → PLO ( β = 0.814). Hierarchical regression showed no significant moderating by English proficiency on any TAM path, though a small positive direct effect on BI was observed ( β = 0.064, p = 0.042). Results suggest that well-designed AI personalization can mitigate language-related barriers, allowing core TAM mechanisms to operate consistently across proficiency levels. The findings highlight the potential of adaptive AI tools to foster equitable engagement in vocational language learning. Future research should employ multi-item or objective proficiency measures and incorporate actual usage data to further validate these insights.