摘要
Generative artificial intelligence tools such as ChatGPT are rapidly infiltrating academic writing, yet their concrete role in undergraduate thesis production remains underexplored. The present qualitative thematic study, therefore, examined how final year Information Systems students at a state university in Malang employ and experience ChatGPT while drafting their undergraduate theses. Purposive sampling yielded five participants (three undertaking implementation projects and two conducting non implementation studies). Semi structured interviews were conducted online or face-to-face between May and June 2024, audio-recorded, transcribed verbatim, coded in NVivo 14, and analyzed using Braun and Clarke’s six-step thematic approach with source triangulation to increase credibility. The analysis produced six interlinked themes. Participants reported opening ChatGPT whenever they wrote, primarily to paraphrase references, understand research methods, draft introductory and literature-review sections, translate and summarise sources, locate journals, and debug code. They valued the tool’s personalized explanations, instant 24/7 availability, and time efficiency relative to conventional search engines, noting that it accelerated idea generation and enriched vocabulary. However, students cautioned that outputs were often generic, occasionally hallucinated, poorly referenced, and stylistically robotic, requiring manual verification and editing. Over reliance was recognized as a risk that could erode critical reading, independent information seeking, and writing proficiency. Overall, ChatGPT functioned as an informal “third supervisor” or “virtual academic assistant,” improving speed and perceived quality while threatening skill development. This study suggests that universities should integrate AI literacy, embed mandatory source-checking protocols, and design staged supervisory feedback that encourages the balanced use of AI alongside traditional scholarly practice. Future research should test scalable interventions that preserve cognitive rigor while leveraging conversational AI to democratize high quality academic authorship worldwide.