Comparing instructor designed rubric aligned chatbot feedback and instructor feedback in higher education regarding student perceptions of clarity usefulness supportiveness and satisfaction

As generative artificial intelligence increasingly shapes pedagogical practice, automated feedback in higher education requires careful evaluation that extends beyond technical performance to students’ learning experiences. This study compared traditional instructor feedback with feedback generated by an instructor-designed, rubric-aligned chatbot (I-GenF-Bot). The chatbot was developed by the course instructor and configured to provide qualitative, criterion-referenced feedback aligned with a course rubric, without grades. One hundred and six undergraduate students in a third-year Digital Marketing course received iterative, formative feedback from I-GenF-Bot, followed by summative feedback from the instructor using the same rubric. Students completed a questionnaire assessing functional dimensions of feedback quality (clarity and usefulness), an emotional–personal dimension (supportiveness), and overall satisfaction with the chatbot. Results indicated no significant differences between instructor and chatbot feedback on clarity and usefulness, while instructor feedback was perceived as more supportive. Clarity, usefulness, and supportiveness were each positively associated with students’ overall satisfaction with I-GenF-Bot. An exploratory two-step cluster analysis suggested that student heterogeneity was driven primarily by perceptions of instructor feedback rather than chatbot feedback. Qualitative comments highlighted the chatbot’s strengths in immediacy, clarity, and accessibility, alongside limitations related to contextual sensitivity, repetition, and empathy. Overall, the findings suggest that while rubric alignment enables AI-based feedback to approximate instructor feedback on functional quality dimensions, an ‘empathy gap’ remains a key constraint. A hybrid feedback model that combines scalable AI-supported guidance with instructor judgment and relational support therefore appears particularly promising for scalable feedback practices in higher education.