This exploratory study examined how undergraduate students interacted with ChatGPT during essay-writing tasks, with a primary focus on identifying recurring student-AI interaction patterns. Participants were 21 undergraduates enrolled in an eight-week writing course at a private university in South Korea, where ChatGPT use was optional. Interaction transcripts from the 11 students who chose to use ChatGPT served as the primary data source, supplemented by surveys and interviews to contextualize how these patterns were experienced during writing. The analysis identified four recurring interaction patternsRepeatedStructure (RS), Creating Structure (CS), Modifying–Creating Structure (MCS), and Creating–Repeating Structure (CRS)-capturing how students repeated, generated, refined, and iteratively applied prompts while developing their essays. Students primarily used ChatGPT to develop body paragraphs, write conclusions, and revise drafts. They reported benefits such as improved idea organization and refined wording, while also noting redundancy, inaccuracies, and risks of overreliance. Survey and interview data further contextualized these patterns, highlighting students’ perceptions of ChatGPT as a tool for idea expansion, wording refinement, and revision. These findings underscore the dual role of generative AI in writing education and highlight the importance of guided and balanced integration to foster higherorderthinking and writing development. 

Keywords : Generative AI, ChatGPT, Student–AI interaction, Essay Writing, AI literacy