Negotiating Academic Voice with Artificial Intelligence: A Qualitative Inquiry into Gemini-Assisted Task-Based Writing
DOI:
https://doi.org/10.36312/jllet.v1i2.73Keywords:
Academic voice, Authorship, Gemini-assisted writing, Generative artificial intelligence, Task-based writingAbstract
The increasing use of generative artificial intelligence in academic writing raises important questions about how learners maintain academic voice, authorship, ownership, and ethical responsibility while incorporating AI-generated language and ideas. This qualitative study investigated how EFL learners negotiated academic voice during Gemini-assisted task-based writing and how structured AI-supported activities shaped their perceptions and practices of authorship. Thirty-three EFL learners at Universitas Pendidikan Mandalika participated through purposive sampling. Data were collected using semi-structured interviews and structured classroom observations and analyzed iteratively through data condensation, data display, and conclusion drawing and verification. Methodological triangulation was employed to compare participants’ reported experiences with their observable writing practices. The findings revealed three interrelated processes through which learners negotiated academic voice: selective acceptance, deliberate rejection, and purposeful transformation of Gemini-generated content. Students primarily accepted AI suggestions for linguistic clarity and academic expression, rejected content that was irrelevant, inaccurate, unverifiable, or inconsistent with their intended stance, and transformed AI output through paraphrasing, contextualization, reorganization, and integration with their own reasoning. The findings further showed that learners increasingly understood authorship as responsibility for meaning, decisions, verification, and final claims rather than exclusive production of every sentence. Initial independent reasoning and reflective justification strengthened perceptions of ownership, while extensive uncritical copying weakened authorial identification. Ethical concerns centered on verification, transparency, disclosure, and uncertainty regarding acceptable levels of AI assistance. The study concludes that Gemini can support academic writing without necessarily diminishing student voice when task design foregrounds critical evaluation, transformation, accountability, and reflective human–AI interaction.






