A systematic review of human- and emerging AI-generated reformulation in L2 writing

Reformulation has long served as a feedback approach in second language (L2) writing. With the emergence of Generative Artificial Intelligence (GAI), particularly ChatGPT, AI-generated reformulation has recently begun to appear in L2 writing research, although this area remains nascent. Guided by the PRISMA framework, this study reviews 26 empirical studies on reformulation in L2 writing published up to January 13, 2026. Unlike previous reviews of positive evidence feedback or model texts more broadly, this review provides a reformulation-focused synthesis and maps both human-generated and emerging AI-generated practices. The review focuses on three key dimensions: (1) research contexts and trends, (2) methodological characteristics, and (3) primary focus areas and key findings. Results indicate that: (1) research has predominantly targeted university-level English as a Foreign Language (EFL) learners, with emerging interest in AI-generated feedback; (2) methodologies have mainly employed quasi-experimental designs situated in classroom-based, three-stage narrative tasks; and (3) while reformulation-only studies have tended to conceptualize writing as a process and emphasize learners’ attention to lexical and grammatical features, reformulation-integrated studies have produced mixed findings regarding the transition from cognitive processing to written products. These findings underscore the multifaceted role of reformulation in L2 writing development and identify critical directions for future research, particularly concerning the integration of emerging AI-assisted pedagogical practices.