Does the impact of GenAI persist in learning? comparing AI and instructor feedback on accessibility

The continued marginalization of over 1.3 billion people with different abilities is partly due to limited accessibility education in computing curricula. To address this gap, effective pedagogical strategies are needed, yet there is little empirical research on the role of feedback in this domain. Using a quasi-experimental design, this study investigated the comparative effectiveness of AI-generated and instructor-delivered feedback on teaching inclusive design to (N = 81) undergraduate computer science students. In a web development course, participants were assigned to either a control (Instructor-delivered feedback) or experimental (AI-generated feedback) group. Using pre- and post-assessments, assignments, and delayed-exams, we evaluated how students understand the concept of accessibility through auditing of educational websites. Our findings show that while there was no significant difference in short-term assignment performance between the modalities, the instructor-led group performed significantly better on the exam assessing long- term retention (p < 0.001). No significant differences based on gender were observed in students' assessment and the exam across feedback modalities. This result suggests that while AI-tools can support immediate learning tasks in accessibility, direct instructor feedback is more effective for promoting durable, and understanding required to embed inclusive practices in future professionals.