<![CDATA[The application of Generative Artificial Intelligence (GenAI) in academic writing is increasing. However, limited research has examined how students interact with AI during the writing process, especially among EFL postgraduate students in Indonesia. This study investigated the interaction patterns between three postgraduate students enrolled in the English Language Education Study Program at an Islamic university in Jambi, Indonesia, and ChatGPT as they completed academic writing assignments. A qualitative approach with a case study design was employed. Data were collected from ChatGPT chat logs and written reflections and subsequently analysed using the human-LLM interaction taxonomy and thematic analysis. The results identified three distinct interaction patterns: a balanced pattern in P1, where Planning and Iterating occurred in equal proportions; a facilitative pattern in P2, marked by the use of ChatGPT for direct academic support; and an evaluative pattern in P3, defined by consistent verification of the validity and credibility of ChatGPT output. Additionally, written reflections revealed cognitive dimensions not evident in the chat logs, such as information selection, source verification, and attention to academic integrity. The findings suggest that the quality of human-AI interaction may play an important role in GenAI-assisted academic writing. Pedagogically, this study recommends integrating AI literacy training into postgraduate EFL curricula. Methodologically, the combination of chat logs and written reflections offers insights into human-AI collaboration in academic writing.]]>
