ChatGPT-assisted formative feedback and argumentative essay writing performance in Moroccan secondary schools

This quasi-experimental study examined the effect of ChatGPT-assisted formative feedback on Moroccan secondary school students' argumentative essay writing performance. Thirty-one second-year Baccalaureate students participated in a four-week intervention: 22 students received teacher-mediated ChatGPT feedback, whereas 9 students received traditional teacher feedback. Pre-test and post-test essays were evaluated by two independent raters using a 20-point analytic framework based on Toulmin's Argumentation Model. Recalculation from the reported group sizes, means, and standard deviations showed no pre-test difference between the groups, t(29) = 0.08, p = .938. At post-test, the experimental group (M = 11.06, SD = 1.77) outperformed the control group (M = 9.62, SD = 1.72), t(29) = -2.07, p = .047, with a large standardized mean difference (Cohen's d = 0.82). A Welch sensitivity analysis produced a similar effect estimate but a marginal p value (p = .053), indicating that the inferential result should be interpreted cautiously because of the small, imbalanced groups. The findings provide preliminary evidence that carefully scaffolded ChatGPT-assisted feedback may support EFL argumentative writing, while underscoring the importance of teacher mediation and replication with larger, balanced samples.