Exploring the Effectiveness of Teacher Feedback Versus AI-Generated Feedback in L2 Writing: A Comparative Study in a Tertiary EFL Context

Author : Dr Emrah Cinkara

Abstract :The rapid advancement of artificial intelligence (AI) in education has opened new possibilities for providing automated feedback on students’ written products. However, limited research has critically compared AI-generated feedback with traditional teacher feedback in terms of pedagogical effectiveness, accuracy, and teacher perceptions. This study investigates the differences, advantages, and limitations of teacher versus AI-generated feedback on the writing of A2-level learners in a tertiary English as a Foreign Language (EFL) context. A total of 38 learners participated in the study, completing a guided writing task focused on self-description. Two groups were formed: one received written feedback from experienced EFL instructors, while the other was provided feedback generated by a widely used AI tool (ChatGPT). The students’ revised writing samples were evaluated using a standardized rubric covering content, grammar, coherence, and lexical variety. In addition to this quantitative analysis, in-depth semi-structured interviews were conducted with the instructors to gather their reflections on the usefulness, accuracy, and pedagogical alignment of AI-generated feedback compared to their own. The findings reveal both convergences and divergences between the two feedback sources. While AI feedback was found to be prompt, consistent, and useful for surface-level error correction (e.g., grammar and vocabulary), teacher feedback demonstrated superiority in addressing deeper discourse-level issues such as idea development, cohesion, and audience awareness. Furthermore, instructors expressed ambivalence regarding the integration of AI in feedback practices: they acknowledged its potential to reduce workload but raised concerns about contextual appropriateness and the absence of dialogic feedback processes. These findings offer important implications for blended feedback approaches in L2 writing instruction, highlighting the complementary strengths of human and AI-driven support. By bringing together learner outcomes and teacher perspectives, this study contributes to ongoing discussions on the pedagogical integration of AI in language education and provides evidence-based insights into how human and machine feedback can be effectively combined to enhance L2 writing development.

Keywords :AI-generated feedback, teacher feedback, L2 writing, ChatGPT, A2 learners, EFL, automated writing evaluation, tertiary education.

Conference Name :International Conference on Educational and Instructional Technology (ICEIT-25)

Conference Place Prague, Czech Republic

Conference Date 13th Aug 2025