Exploring the use of ChatGPT-generated model texts as a feedback instrument: EFL students' text quality and perceptions

PurposeThis study investigates the usefulness of ChatGPT-generated model texts as a feedback instrument. Specifically, it focuses on the effects of ChatGPT-generated model texts on EFL students' text quality in terms of content, organisation, vocabulary, and grammar. Students' perceptions are also examined to gauge a more comprehensive understanding from a subjective perspective.MethodologyThe study adopted a pretest and post-test quasi-experimental design with three groups: one control group (CG) without any form of feedback, one experimental group (EG1) with a model text written by human teachers, and one experimental group (EG2) receiving a model text generated by ChatGPT. Data were collected from 104 Chinese high school EFL students. Students' writing tests were analysed using descriptive and inferential statistics, while perception questionnaires were analysed using descriptive statistics and content analysis.FindingsThe results showed that while there was no significant difference between the three groups at the outset, both experimental groups demonstrated better performance than the control group in all the analytical measures in the post-test with large effect sizes. In addition, the improvements made by EG1 and EG2 were similar. Students generally held positive perceptions of using ChatGPT-generated model texts as a feedback tool, although they expressed concerns about the potential lack of contextual relevance and stylistic unnaturalness of the AI-generated models.Originality/valueThis study contributes to the growing body of research on AI-assisted feedback and offers implications for integrating emerging technology, such as ChatGPT, into the use of model texts as a feedback instrument in EFL writing classrooms.