Introduction: Generative artificial intelligence offers new possibilities to strengthen formative feedback and improve the accessibility of evaluation processes in higher education. Objective: To evaluate the effect of AI-mediated formative feedback on academic performance and the perception of clarity, usefulness, accessibility, and equity of assessment in university students. Methodology: A quantitative, explanatory, and quasi-experimental study was developed with a pre-test-posttest design and non-equivalent groups. A total of 128 students participated, divided into an experimental group and a control group of 64 participants each. The intervention was developed over eight weeks and the effect on performance was analyzed by controlling the initial score. Results: The experimental group increased its mean score from 59.8 to 79.6 points, while the control group went from 60.1 to 69.1. Likewise, the experimental group obtained higher scores in clarity (4.47), utility (4.51), accessibility (4.43), and equity (4.32). The advantage was greater among students with low initial performance. Conclusion: Feedback mediated by artificial intelligence showed potential to improve learning and favor a more accessible, useful, and inclusive evaluation experience when integrated with teacher supervision.