1　Introduction

This section introduces the application background of generative artificial intelligence in foreign language education, highlighting its advantages in writing feedback in terms of speed, personalization, and cost, while also noting issues such as dependence on prompt quality, erroneous feedback, and the potential to weaken students' independent thinking abilities. Existing research has predominantly focused on static comparisons, neglecting the dynamic evolution and value-added benefits of students' multi-round use of GAI feedback. Therefore, this study aims to explore the specific value-added benefits of multi-round GAI feedback on college students' English writing quality through a pedagogical experiment.

2　Literature Review

This section defines the technical connotation and core characteristics of generative artificial intelligence (GAI), identifying its "creative" nature that distinguishes it from traditional AI and its multi-modal data processing capabilities. It reviews the evolution of GAI applications in foreign language writing instruction: from automated writing evaluation (AWE) to GAI feedback, covering dimensions such as curriculum reform, academic writing, grammar correction, and essay assessment, emphasizing its pedagogical value throughout the entire writing process including pre-writing, writing, evaluation, and reflection. In the section on current research status of GAI writing feedback, empirical studies from both teacher and student perspectives are systematically reviewed: the former focuses on quantitative comparisons of feedback quantity, types, and quality as well as teacher attitude surveys, while the latter examines changes in students' writing performance, motivation, engagement, and user experience. Subsequently, four limitations of existing research are identified: first, insufficient attention to the improvement effects on students' essay quality; second, the predominance of researcher-provided fixed prompts, lacking authentic contexts where students autonomously generate feedback; third, the absence of longitudinal research designs tracking the dynamic development of writing quality under GAI feedback; and fourth, inadequate examination of the value-added benefits of "teacher + GAI" composite feedback compared to single teacher feedback. Based on this, the core objective of this study is proposed—to reveal the value-added benefits of GAI feedback on college students' English essay quality improvement beyond teacher feedback through longitudinal comparison of four timed in-class writings, providing empirical evidence and practical solutions for the deep integration of GAI into English writing instruction.

3　Research Methodology

This section introduces the research design of a pedagogical experiment comparing the effects of generative artificial intelligence (GAI) feedback and teacher feedback on college students' English writing quality. The study involved 87 non-English major freshmen from a Project 985 university, conducted within a 16-week "College English III" course. Two parallel classes were randomly assigned to an experimental group (teacher + GAI feedback) and a control group (teacher feedback only). Data collection was completed through the natural course delivery process, encompassing four argumentative essays. Two raters independently scored the essays according to CET-4 standards and averaged the scores. The experimental group sought GAI-assisted revision based on teacher feedback, while the control group revised based solely on teacher feedback, with both groups completing the tasks within class time limits. Data analysis employed SPSS 27, utilizing normality tests, paired-sample t-tests (for within-group pre-post comparisons), and independent-sample t-tests (for between-group comparison of improvement magnitude) to investigate whether GAI feedback possesses value-added benefits.

4　Research Results

Both groups showed steady improvement in their first draft scores across four essays, with revised draft scores consistently higher than first drafts. The teacher + GAI feedback group demonstrated greater improvement magnitude than the teacher feedback group throughout, and significantly outperformed the latter except for the second essay; in the fourth essay, the improvement magnitude for both feedback methods was considerably lower than in the first three. Paired-sample t-tests revealed that post-feedback scores were significantly higher than pre-feedback scores for both groups (p<0.01), with large effect sizes (|Cohen's d|>0.7). Independent-sample t-tests indicated no significant difference in improvement magnitude between groups for the first two essays (p>0.05), while significant differences emerged for the latter two (p<0.05) with medium effect sizes, suggesting that the value-added benefits of GAI on writing quality become increasingly evident with accumulated feedback instances.

5　Discussion

Based on quantitative analysis of first and revised draft scores from four essays by Chinese college students under teacher + GAI feedback and teacher feedback conditions, this section examines the value-added benefits of GAI corrective feedback. The findings reveal that both feedback methods effectively improve writing quality, with GAI feedback demonstrating potential cumulative effects: no significant difference in improvement magnitude between the two feedback methods for the first two essays, while significant differences emerged for the latter two, indicating that GAI's value-added benefits gradually become prominent with accumulated feedback instances. This difference stems from freshmen's initial lack of GAI experience and feedback literacy, leading to focus on linguistic aspects while neglecting structure and content, as well as difficulty processing dual feedback; with accumulated experience, students become better able to generate and utilize feedback, and the advantage of GAI's direct indication of revision points can be fully realized. The study concludes that GAI feedback cannot replace teacher feedback; teachers should collaborate with GAI and provide long-term guidance to enhance students' GAI feedback literacy to fully realize its benefits.

6　Conclusion

This section discusses pedagogical implications of GAI feedback in English writing instruction, noting that both GAI feedback and teacher feedback can effectively improve writing quality. It recommends that teachers utilize GAI for timely first-draft feedback to optimize time allocation, while emphasizing that teachers must monitor GAI feedback quality and collaborate with GAI, prioritizing student agency to avoid direct copying. Additionally, limitations including small sample size, single genre, ceiling effects in scoring, and individual differences in students' feedback literacy are acknowledged, with suggestions proposed for future research.