From algorithms to annotations: Rethinking feedback practices in academic writing through AI-human comparison

While ChatGPT has gained prominence in education for its efficiency, little is known about how its feedback compares to that of human instructors, particularly in English for Academic Purposes (EAP) settings with second language (L2) writers. This study examined how ChatGPT and instructors used epistemic strategies (assertiveness vs. mitigation) and delivery methods (monologic vs. dialogic) in their evaluative comments on 200 academic introductions written by Malaysian L2 students. The findings revealed that both feedback sources employed mitigation comparably to foster a face-saving environment for receiving academic criticism. However, instructors demonstrated greater assertiveness and relied more on dialogic feedback, reflecting their authoritative role in guiding revisions and prompting interaction. Conversely, ChatGPT predominantly provided monologic feedback, offering immediate and grammatically sound but static comments. The most common blend in both sources was mitigation combined with dialogic delivery, which highlights a balance between flexibility and engagement. Overall, the study indicates that while AI feedback can complement human input, it often lacks the adaptability and interpersonal depth of instructor responses. These insights inform EAP and L2 writing pedagogy by pointing to the value of blended feedback models that leverage AI’s efficiency while retaining the interactive and authoritative qualities of human expertise.