This study explores how artificial intelligence (AI) can support feedback practices in undergraduate English for Academic Purposes (EAP) blended courses. It focuses on the challenge of providing timely and varied feedback that helps students understand goals, assess progress, and improve in the online asynchronous component of learning environments. Using a framework addressing different types and levels of learning support, the study compares instructor- and AI-generated feedback through content analysis of archived comments on student writing. Results show that instructor feedback was often personalized and context-specific but mainly addressed surface-level issues, offering limited support for learning strategies or independent skill development. In contrast, AI-generated feedback provided more balanced support and clearer forward-looking guidance but lacked personalization. Combining both forms of feedback may lead to a more comprehensive learner-centered experience. These findings offer insights for developing scalable and sustainable feedback systems that foster student engagement, autonomy, and long-term progress in higher education.