Artificial Intelligence (AI), in particular the novel possibilities supplied by GPT-4, is increasingly being utilized in educational settings, yet its impact on student activation and learning progress remains controversial. We use a quasi-experimental design in a lab-in-the-field setting in undergraduate tutorial classes in economics to precisely understand how well AI feedback performs as compared to lecturer feedback and peer feedback concerning student activation and learning progress. Over the course of the semester and across 18 tutorial classes, we asked students to answer eight open-ended questions. We programmed an interface between an advanced audience-response system and GPT-4 to organize the data collection. Depending on the treatment, students received lecturer feedback at the classroom-level only (LF), additional individual feedback from their peers (PF) or from the AI (AIF), generated by GPT-4. The feedback referred to both the content and the style of the student answers. We find that PF and AIF have a significant positive effect on student activation. Compared to LF, participation in the tasks is significantly higher in AIF and PF. Across the eight tasks, AIF stands out for inducing the longest written answers and the most significant improvement in style. Compared to PF, we find mild evidence for algorithm aversion, yet AI demonstrated more reliable feedback provision. Our findings demonstrate that AI, specifically GPT-4, can be as effective as peer feedback in stimulating student engagement and participation, thereby offering a scalable and consistent solution for providing individual feedback in educational settings.