Central South University study provides novel insights into GenAI-mediated second language writing instruction

Generative artificial intelligence (GenAI) has rapidly transformed second language (L2) writing education, serving as digital writing partners and automated feedback providers. Yet the landscape of how these technologies actually support learning has remained fragmented. A new systematic review by researchers at Central South University applies the established Activity Theory framework to analyze 55 empirical studies, revealing that the effectiveness of GenAI in L2 writing instruction depends crucially on how tools, task design, and social roles interact within the learning process.

GenAI technologies like ChatGPT can brainstorm ideas, refine arguments, and provide instant feedback on grammar and style. While researchers have cautioned that GenAI may bring negative impacts such as over-reliance and plagiarism risks, this technology has been found to facilitate the learning, teaching, and assessment of L2 writing. However, previous reviews have primarily focused on isolated user perceptions of ChatGPT specifically, rather than the wider landscape of GenAI tools and the complex ecosystem of GenAI-assisted writing classrooms. The current study, published in ECNU Review of Education, moves beyond simple "pros and cons" to examine how six key elements, including learners, tools, community settings, role divisions, procedural rules, and learning objectives, interact to shape educational outcomes.

Led by Mi Rong and Yuan Yao from Central South University, the team systematically reviewed 55 peer-reviewed empirical studies from Web of Science-indexed journals published between January 2023 and July 2025. Unlike previous reviews that treated GenAI use as a simple input-output process, this study employs Activity Theory to offer a holistic view of how human agency and technological mediation co-construct learning experiences. Their analysis reveals that tools, division of labor, and rules serve as critical mediators that can either enhance or hinder learning. For instance, while GenAI improves writing quality and motivation when students use it as a "writing collaborator," over-reliance on generic chatbots may diminish critical thinking and creativity, leading to only marginal improvements in writing skills. In contrast, custom-designed AI tools tailored to specific writing tasks demonstrate greater educational value than general-purpose applications.

"Activity Theory provides a comprehensive lens for understanding the complex interactions between humans and technology in educational settings," explains the research team. "Our findings suggest that successful GenAI integration requires moving beyond simple prompting toward staged, dialogic protocols that distribute agency between teachers, students, and AI systems."

The study also finds that collaborative use of GenAI, where students and teachers share prompt-making responsibilities, produces more durable learning gains than individual use. Specifically, when students co-create prompts with teachers or engage in collaborative feedback sessions, they develop deeper metacognitive awareness and incorporate feedback into revisions more effectively than when working alone. Conversely, simply providing example prompts without critical thinking training may lead to passive over-reliance on AI-generated content and subsequent loss of creativity. Effective training proves essential: students who receive structured guidance on both technical operations and writing strategies demonstrate stronger academic integrity and self-regulated learning abilities.

These findings provide educators and policymakers with a comprehensive roadmap for designing writing instruction that harnesses technological assistance while preserving learner autonomy and critical thinking capacities.