Effects of three levels of AI integration on second language academic writing: Evaluating restricted, guided, and free use of ChatGPT

This study evaluates the impact of different ChatGPT usage protocols as instructional tools for academic writing among L2-English learners by exploring how three levels of AI integration affect learning outcomes. Three groups of Korean college students ( n = 103) were taught by the same instructor in English essay writing courses: The Restricted group did not use ChatGPT, the Guided group accessed it only during supervised in-class activities, and the Free group had unrestricted access to ChatGPT throughout the semester. The results indicate that both the Guided and Free groups consistently outperformed the Restricted group, demonstrating the educational benefits of AI engagement as an innovative approach to language learning. A close examination of essay scores finds that the Guided group excelled in content quality, although there were no significant differences in organization and language use between the Guided and Free groups. This suggests that while structured guidance is crucial for content development, more autonomy may be advantageous in other areas of writing. An interview with the instructor provides further insights into how different levels of autonomy in engagement with AI affected student engagement, and suggests the benefits of strategically integrating AI to improve L2 academic writing skills.