Aim/Purpose: The present study explores the use of AI-powered ChatGPT as a feedback tool for automated writing evaluation in a higher education context. Background: Artificial intelligence (AI) is increasingly applied in the field of education, offering new opportunities with its evolving capabilities. One area where AI tools have promising potential is the field of second/foreign language (L2) writing. Methodology: Participants, Turkish pre-service teachers of English enrolled in an academic writing course, received feedback from ChatGPT, peers, and the course instructor. Data came from focus-group interviews focusing on the participants' perceptions of ChatGPT as a feedback tool in comparison to peer and teacher feedback. Thematic analysis was conducted on the transcribed interview data. Contribution: Students can benefit from AI feedback by independently refining their writing. They should develop critical thinking skills when using AI tools like ChatGPT for feedback. Findings: Thematic analysis revealed that participants thought ChatGPT had several affordances, such as practicality, interactivity, and adaptability; however, it also had certain constraints, such as occasional inconsistencies and dependence on the quality of prompts. Overall, the study highlights that ChatGPT could be a valuable feedback tool for L2 writing. Recommendations for Practitioners: The findings suggest that ChatGPT can significantly facilitate the L2 writing process but cannot fully replace feedback from peers or teachers. Teachers should provide initial training and ongoing support to students in using AI effectively for feedback. Policymakers should consider the advancements in AI and their implications for L2, anticipate students' increasing access to advanced writing assistance, and seek ways to offer suitable guidance. Future Research: This study tapped into students' experiences as a snapshot. Future research may focus on the long-term influence of ChatGPT feedback on L2 writing and how to combine it with other types of feedback to optimize learning.