Sertdemir Kabakçı’s thesis explores teachers’ perceptions and experiences of LLM use in EFL writing instruction. Using an explanatory sequential mixed-methods design, Sertdemir Kabakçı presented EFL instructors’ familiarity with and use of LLMs, as well as their perceptions and experiences in writing instruction, combining an online questionnaire with follow-up semi-structured interviews. The main findings include generally positive perceptions of LLMs, diverse pedagogical uses, and concerns about reliability, feedback quality, and ethical use. The thesis brings new insights into LLM integration, human–AI collaboration, and AI-supported EFL writing instruction. The abstract of the thesis is as follows:
Title of Thesis: A Mixed-Methods Study of Teachers’ Perceptions and Experiences of LLM Use in EFL Writing Instruction
Supervisor: Asst. Prof. Dr. Duygu F. Şafak
Jury Members: Asst. Prof. Dr. Duygu F. Şafak, Assoc. Prof. Dr. Ufuk Balaman (external), Asst. Prof. Dr. Mehmet Sak (external)
Defence Date: 28.08.2026
Employing an explanatory sequential mixed-methods design, this study explores instructors’ experiences of integrating large language models (LLMs) into EFL and writing instruction at a single public institution, as well as the affordances, challenges, and future directions of LLM utilisation. Guided by the third generation of Activity Theory (Engeström, 1987, 2015) and informed by the AWE-P scale (Wilson et al., 2024), the study combines data from an online questionnaire (N = 100), the AWE-P scale (n = 24), and one-on-one semi-structured interviews (n = 13). The data were analysed using descriptive and inferential statistics, followed by hybrid content analysis. The findings revealed that instructors generally perceived LLMs positively and adopted them for diverse pedagogical purposes in EFL teaching in general and in writing instruction in particular. Although LLMs were perceived as useful for reducing workload, providing detailed feedback, and assisting with scoring, concerns remained regarding feedback quality, scoring reliability, and ethical AI use. The study also highlights the importance of human-AI collaboration and the need for institutional AI policy adoption.
Keywords: large language models, LLM-mediated feedback, EFL writing