Human and AI-Generated Written Feedback in Online English Instruction: A Comparative Discourse Analysis

Written corrective feedback (WCF) plays an important role in second language instruction. As AI tools are increasingly used to provide feedback, however, it is worth asking how their responses differ from those of experienced human tutors. This study compares written feedback given to adult learners of English in an online tutoring context. It examines feedback written by the researcher, who is an experienced tutor, during six online lessons, alongside the AI-generated Lesson Insights reports that the tutoring platform Preply produced for the same lessons. The six lesson-matched pairs, dated January to August 2026, are analysed qualitatively using categories established in WCF research: the focus of the feedback, whether corrections are direct or indirect and how they are explained, how specific the comments are, and the tone and affective framing they convey. The analysis shows that the tutor's feedback works as a neutral record of the learner's errors, vocabulary and pronunciation. The AI reports summarise the lesson, praise the learner, frame weaknesses positively and quantify performance. The two sources select largely different errors, and the AI reports sometimes relay the tutor's own teaching or attribute the tutor's words to the learner. The study considers what these differences may mean for tutors and learners on platforms where both kinds of feedback are received together. Rather than arguing that one source is better than the other, the study aims to show what each can offer and to help tutors and course designers make informed decisions about using AI tools in their feedback practices.