As effective as the teacher, but less motivating: a randomized controlled comparison of AI-generated and teacher feedback on students’ writing performance and intrinsic motivation

Introduction Generative artificial intelligence (AI) offers scalable feedback on student writing; however, its effects on writing performance and motivation relative to teacher feedback remain insufficiently understood. Drawing on Feedback Intervention Theory and Self-Determination Theory, this study examined the comparative effectiveness of AI-generated and teacher-provided feedback on students’ writing performance and motivational outcomes. Methods A randomized controlled study assigned 312 undergraduate students to one of three conditions: AI-generated feedback, teacher feedback, or a control condition without substantive feedback. A total of 287 participants were included in the analyses. The study evaluated revised-essay quality, feedback engagement, writing self-efficacy, intrinsic motivation, perceived relatedness to the feedback source, and AI literacy. Results Both AI-generated feedback and teacher feedback produced higher revised-essay quality than the control condition, with effect sizes of d = 0.60 and d = 0.59, respectively. The two feedback conditions were statistically equivalent within a prespecified equivalence margin of ±0.35 standard deviations. Feedback engagement, but not writing self-efficacy, statistically accounted for the observed performance advantage. Teacher feedback elicited higher intrinsic motivation than AI-generated feedback, although AI-generated feedback also yielded higher intrinsic motivation than the control condition. This motivational difference was associated with perceived relatedness to the feedback source. The indirect effects identified in the analyses are associational rather than causal. AI literacy did not significantly moderate the effect of feedback on writing performance ( p = 0.058). Discussion Within the context of a single feedback-and-revision cycle at one institution, the findings suggest that AI-generated feedback can achieve writing-performance outcomes comparable to those of teacher feedback, while providing weaker motivational support. These results highlight the potential value of blended feedback approaches that combine the scalability of AI-generated feedback with the relational and motivational strengths of teacher–student interaction. Further research across multiple institutions and repeated feedback cycles is needed to establish the generalizability and longer-term implications of these findings.