Impact of proficiency on Chinese EFL learners’ interaction with AI-generated feedback for translation revision

AI-generated feedback has become a growing focus in second language education research. Its effectiveness, however, in translation-based language learning has not been fully explored, and the impact of English proficiency on students' uptake of AI-generated feedback remains debated. To address this, we investigated 83 Chinese EFL students' interaction with AI-generated feedback for translation revision. Drawing on both quantitative and qualitative data, the results showed that: (1) both high- and low-proficiency learners actively implemented AI-generated feedback, with high-proficiency learners being more selective; (2) AI-generated feedback led to significant improvements in translation quality and five textual features, particularly benefiting low-proficiency learners; (3) translation quality improvement was positively correlated with feedback uptake rate, but negatively correlated with English proficiency; (4) most learners held positive attitudes toward AI-generated feedback and expressed strong interest in future use, though some raised concerns about AI feedback accuracy and overreliance. This research sheds light on the pedagogical potential of AI-generated feedback in translation-based language learning and the moderating effect of proficiency on students' interaction with AI. AI literacy instruction is therefore recommended for inclusion in the EFL curriculum, with attention to students' proficiency differences to ensure that AI feedback offers pedagogically meaningful support.