Generative AI and Differentiated Instruction in Private Undergraduate English Education: Implementation Pathways and a Conceptual Framework

The rapid diffusion of generative artificial intelligence (GenAI) has reopened an old question in language pedagogy: how to teach heterogeneous classrooms without leaving slower or more advanced learners behind. In the Chinese private undergraduate sector, non-English majors often enter with uneven proficiency while large class sizes constrain individual attention, so the traditional ideal of teaching according to aptitude has remained more aspiration than routine practice. This paper argues that GenAI, used deliberately, can operationalize differentiated instruction at scale. Drawing on Tomlinson's differentiation framework, Vygotsky's zone of proximal development, and technology-acceptance theory, the paper develops a conceptual model and proposes six implementation pathways: diagnostic profiling, AI-generated tiered materials, personalized feedback, learner choice in products, teacher role reconfiguration, and integrity safeguards. The analysis situates each pathway within the constraints of private undergraduate institutions and connects it to the conditions under which differentiated judgment is preserved rather than displaced. The paper concludes by discussing the theoretical implications of framing GenAI as an engine for teacher-led differentiation and by outlining the propositions that future empirical work should test.