ChatGPT and Generative AI in Writing Instruction: Literature Review

Abstract

This study is motivated by the rapid adoption of ChatGPT and other generative artificial intelligence tools in writing instruction across educational levels. The growing use of these tools raises questions about their actual pedagogical value, the reliability of the feedback they generate, and the risks they pose to academic integrity. This study aims to map the role, feedback quality, and challenges of ChatGPT use in writing instruction, and to identify implications for Arabic writing instruction. The method used is library research with a descriptive-analytical qualitative design. Data were collected from journal articles, books, and scholarly reports indexed in Scopus, Scite.ai, and Google Scholar, then analyzed through thematic content analysis. The results show that ChatGPT supports idea development, drafting, and surface-level language correction, but its contribution to higher-order feedback depends on prompt design and revision structure. AI-generated citations remain unreliable and require independent verification. User acceptance is generally positive but conditional on transparency and human oversight. Based on these results, it can be concluded that ChatGPT functions best as a complementary tool guided by teachers, not as a replacement for human judgment, and that Arabic writing instruction needs its own empirical evidence base for AI-assisted feedback design.