GenAI Integration and EFL Academic Writing: A Quasi-Experimental Study of Performance Gains and Developmental Error Patterns in Chinese Higher Education

This study investigated the impact of Generative AI (GenAI) tools on Chinese EFL university students’ academic writing development. Seventy second-year English majors participated in a 16-week quasi-experimental intervention; the intervention group ( n = 35) used GenAI tools for writing support, while the control group ( n = 35) received traditional instruction. The intervention group demonstrated significantly greater improvement in overall writing performance, particularly in organisation (d = 0.81) and content development (d = 0.73); moderate effects were also observed across vocabulary (d = 0.69), language use (d = 0.62), and mechanics (d = 0.55). Error analysis revealed a 43.1% reduction in grammatical errors and a 42.2% reduction in lexical errors; concurrently, discourse-level errors increased by 20.3%, a pattern interpreted as reflecting developmental progression toward more complex rhetorical construction rather than declining accuracy. The findings suggest that GenAI tools functioned as cognitive scaffolds, supporting a reallocation of learner attention from surface-level accuracy to higher-order rhetorical planning, with particularly strong effects on organisation and content development—the competencies most central to academic writing quality.