Key elements and practical logic of AI-optimized EFL formative assessment in higher education: a grounded theory study

Introduction Artificial intelligence (AI) is rapidly reshaping higher education, yet formative assessment in English as a Foreign Language (EFL) teaching often remains standardized and surface-level, limiting its value for assessment-driven learning. In response, interest is growing in human–machine collaborative assessment, but coherent theoretical models are still scarce. Methods This study uses grounded theory to analyze EFL classrooms in Chinese universities, with the goal of building a theoretical model of AI-optimized EFL formative assessment. Results The findings show that optimization is a dynamic process involving five interconnected elements: (1) the functional system of AI-driven English formative assessment; (2) the transformation of English teachers' assessment roles; (3) English learners' shifts in assessment experiences and cognition; (4) reconstruction of the EFL formative assessment paradigm; and (5) constraints in the practice of English formative assessment. The resulting model proposes potentially iterative relationships among these elements, in which the functional framework of AI-driven formative assessment provides the technological infrastructure for the entire system; evolving teacher roles and learner experiences influence each other on that basis and gradually reshape assessment practices. Discussion/Conclusion Importantly, paradigm change does not follow a simple linear path. Instead, it emerges through ongoing negotiation and flexible adaptation to local educational challenges. This framework deepens understanding of how AI can optimize EFL formative assessment and offers conceptual guidance for designing precise, personalized assessment practices that align technological possibilities with effective pedagogy in diverse English educational contexts.