<p>Artificial intelligence-generated content feedback (AIGCF) has become increasingly valuable in the field of learning. Although research exists on AIGCF’s effectiveness, with some studies showing improved student writing and others showing minimal or negative effects, their overall impact remains unclear. This study aimed to examine the effect of AIGCF, exemplified by ChatGPT-4, on non-native English students’ writing quality and evaluate the quality of AIGCF itself. We conducted a single-group experiment with undergraduates. Thirty-two participants completed a series of writing tasks over ten weeks and received AIGCF for their work. We assessed the writing quality based on syntactic complexity, lexical complexity, accuracy, and fluency. We also evaluated the quality of AIGCF with respect to criteria-based feedback, clarity of improvement directions, accuracy, prioritization of essential features, and supportive tone. Preliminary findings suggested that AIGCF might be useful in influencing syntactic and lexical complexity, but its impact on improving accuracy and fluency was variable. The study revealed strengths and weaknesses in the quality of AIGCF, with criteria-based feedback emerging as a notable strength. The study also showed that the quality of feedback based on criteria and the clarity of suggestions for improvement got better over time. However, the prioritization of essential features, the accuracy of the feedback, and the tone of support decreased. It was concluded that the effectiveness of AIGC varies depending on the specific writing area. This study provided valuable insights into the potential of AIGCF in writing instruction and highlighted areas for future research.</p>