Given that generative AI can instantly produce a wide range of feedback types, it is increasingly important to identify appropriate types of AI-generated feedback. This exploratory study thus compared ten college students’ attitudes toward two types of ChatGPT-generated feedback on their writing: commentary and reformulation. Participants’ attitudes were examined through an online survey and analyzed using both quantitative and qualitative methods. Overall, students expressed positive attitudes toward both feedback types in terms of satisfaction, usefulness, helpfulness for improving English proficiency, and helpfulness for improving English writing skills. However, most participants did not strongly support replacing teacher feedback with AIgenerated feedback. Despite the absence of statistically significant main or interaction effects, qualitative findings suggest that students’ attitudes may differ depending on their proficiency levels. Some students also cautioned against providing reformulation on interim drafts due to its potential risk to academic integrity. These findings suggest that selecting appropriate AI-generated feedback types for particular learners and determining when to use them may require teachers’ professional judgment.