Action Research on Phased Introduction of AI Tools: Developing Students’ Critical Thinking and Independent Academic Writing

As artificial intelligence (AI) tools become increasingly embedded in academic environments, it is essential to understand their influence on students’ intellectual development and to ensure their ethical use. This study examined how a phased integration of AI tools can support students’ academic writing skills while maintaining academic integrity. Using a mixed-methods design, data were collected from surveys, interviews, assignment analysis, classroom observations, and reflective logs. Participants were 60 fourth-year English majors who used tools such as Grammarly, QuillBot, ChatGPT, and DeepSeek at different stages of their writing tasks. Findings from Round 1 showed that 80 per cent of students reported improved grammatical error detection, 70 per cent found AI-generated frameworks useful, 85 per cent successfully modified these frameworks, and 90 per cent recognized AI as a supportive, not a substitutive, tool. In Round 2, 75 per cent reported a better understanding of grammar through analysing AI corrections, 82 per cent identified logical gaps by comparing frameworks, 78 per cent recognized limitations in AI-generated argumentation, and 87 per cent engaged in more deliberate planning of AI use. Together, these patterns suggest that a staged approach to AI integration can foster more reflective evaluation, iterative revision, and gradual skill development. The results also emphasize the need for clear ethical guidelines and structured instructional strategies that balance technological support with the development of independent reasoning and creativity. These insights provide practical guidance for educators and curriculum designers aiming to integrate AI responsibly into higher-education writing instruction. The findings offer a practical model that can be directly applied in academic writing instruction. Teachers can use the phased approach to guide students from basic error correction to critical revision, curriculum designers can integrate it into course structures, and university administrators can develop policies for responsible AI use.