The Impact of ChatGPT as a Formative Feedback Tool on Moroccan Secondary School Students’ Paragraph Writing Skills: A Quasi-Experimental Design

Abstract

This study investigates the impact of ChatGPT-generated formative feedback on Moroccan secondary school students’ paragraph writing performance along with their perceptions of the tool’s effectiveness. Adopting a quasi-experimental mixed-methods design, the study involved 87 students divided into an experimental group (n = 45) and a control group (n = 42). Over an eight-week intervention, the experimental group received AI-generated feedback on their writing tasks while the control group was provided with traditional teacher feedback. Quantitative data from pre-test, post-test, and delayed post-test assessments revealed statistically significant improvements in the experimental group’s writing scores (pre-test M = 6.69, post-test M = 10.01, delayed post-test M = 10.27; p < .001). Paired-samples t-tests confirmed substantial within-group gains (t = -34.38 and -35.69) while independent-samples t-tests revealed significant between-group differences in post-test (t = 22.71) and delayed post-test scores (t = 16.69) favoring the experimental group. Effect size calculations showed a large impact (Cohen’s d = 5.13 for pre- vs. post-test; d = 5.40 for pre- vs. delayed post-test). Qualitative data from semi-structured interviews with nine students in the experimental group indicated that ChatGPT feedback was perceived as clear, specific, and motivational enhancing student engagement, autonomy, and metacognitive awareness. These findings underscore the pedagogical value of integrating AI feedback into EFL writing instruction particularly in resource-constrained educational settings.