Peer feedback has been widely employed in writing instruction in various contexts. However, studies into its effects on improving students’ writing performance, particularly its long-term effects, are under-represented. To fill the research gap, we adopted a pre-, post-, and delayed post-test quasi-experimental design to investigate the long-term effects of a peer feedback intervention. Two intact classes of EFL students from a Chinese university were recruited, with one class participating in peer feedback activities and the other class receiving collective feedback from their teacher for one semester. Their compositions for the pre-, post-, delayed post-tests were compared in terms of the overall text quality, content, organization, complexity, accuracy, and fluency (CAF). The results show that the experimental group participants made significant improvements in the overall text quality, content, organization, and accuracy in the post-test and they retained the improvements 12 weeks after the intervention. In addition, they outperformed their counterparts receiving collective feedback in the overall text quality and organization in the post- and delayed post-tests, and in accuracy only in the delayed post-test. Results suggest that peer feedback could help Chinese EFL learners make progress in writing.