Providing timely, high-quality feedback at scale remains a major challenge in higher education. Artificial intelligence (AI) enables scalable feedback delivery, yet uncertainty remains about how different AI-mediated feedback designs support diverse learners. This study examines students’ perceptions of two AI-mediated feedback approaches: fully AI-generated feedback and a hybrid human–AI feedback model that integrates automated delivery with rule-based, educator-authored pedagogical input. Survey data were collected from 118 undergraduate students and analyzed across four dimensions of feedback perception: utility, supportiveness, motivation, and support for future learning. Student performance level was examined as a pragmatic and scalable indicator of learner variability. Results show no overall differences between AI and hybrid feedback at the cohort level, suggesting perceived pedagogical parity between feedback approaches. However, performance-based differences emerge. Medium-performing students perceive AI feedback as more supportive, valuing its clarity and structured guidance. In contrast, high-performing students report stronger support for future-oriented learning from hybrid feedback that incorporates contextualized, educator-informed guidance. These findings demonstrate that student performance can inform adaptive feedback design in AI-enabled learning environments. By aligning feedback modality with performance-based learner needs, the study provides empirical guidance for designing scalable, personalized feedback systems that strategically integrate human and AI contributions in higher education.