Generative AI and Tertiary EFL Teachers’ Pedagogical Shifts: The Dual Mediating Roles of Cognitive Regulation Scaffolds and Personalized Learning Design

Traditional tertiary English Language Teaching (ELT) has long faced an intractable dilemma: standardized, one-size-fits-all instruction fails to address the heterogeneous linguistic and cognitive needs of second language learners, while teachers’ limited instructional capacity severely restricts the delivery of fine-grained cognitive scaffolding and individualized differentiated feedback. Generative Artificial Intelligence (Generative AI) offers unprecedented technical affordances to resolve these long-standing pain points, yet extant research remains predominantly student-centric, with no empirical studies unpacking the causal mediating mechanisms between Generative AI integration and sustained pedagogical change in tertiary ELT. To address this critical gap, this study employs a fully integrated sequential exploratory mixed-methods design with 205 in-service tertiary EFL teachers across 28 Chinese higher education institutions. Findings reveal that Generative AI drives systematic, multi-dimensional shifts in teachers’ pedagogical practices, primarily channeled through two synergistic mediating pathways: Generative AI-enabled Cognitive Regulation Scaffolds Design (CRSD), and Generative AI-facilitated Personalized Learning Design (PLD). Quantitative analyses confirm significant independent, serial mediating, and synergistic interaction effects of the two pathways, with effects moderated by teachers’ AI-specific Technological Pedagogical Content Knowledge (AI-TPACK), AI literacy, and resolution of core implementation dilemmas. Full-chain endogeneity treatment, reverse model validation, and exhaustive competitive model testing further confirm the causal order and optimality of the dual-pathway framework. This study constructs and empirically validates a revised, theory-grounded framework for higher education ELT, advances scholarship in Generative AI-enabled language education by refining foundational second language acquisition (SLA) theories, and informs evidence-based teacher professional development in the AI era.