Ethnic AI instructor: when a minority professor uses AI voice-over tools

Purpose Student evaluations of teaching influence faculty promotion and tenure decisions, yet research demonstrates these evaluations systematically reflect instructor demographic characteristics independent of teaching quality. As online education expands and instructors increasingly adopt AI technologies like text-to-speech voice narration, questions arise about whether racial bias and AI skepticism interact to create compounded disadvantages for minority instructors. Design/methodology/approach Using a 2×2 experimental design (Instructor race: White vs. Asian × Voice type: Human vs AI-generated), 172 undergraduate students evaluated identical online lecture content. Instructors were identified through photographs and names. Findings Results showed significant racial bias with the White instructor rated higher than the Asian instructor across evaluation dimensions. Students also rated the AI-voiced lecture significantly worse than the human-voiced lecture across all evaluation dimensions. While voice type did not moderate the race effect statistically, simple effects analyses revealed that racial bias was manifested within the AI voice condition. Perceived social presence mediated voice type effects on evaluations. Students' comfort with AI technology moderated voice type effects, with the AI discount effect weakening among high-comfort students. Research limitations/implications Racial bias in student evaluations persists regardless of voice types, while AI-generated voices substantially reduce evaluations through decreased social presence. These findings have practical implications for equitable faculty assessment as institutions adopt AI technologies and for instructors from racial minority groups who face evaluation penalties independent of teaching quality. Originality/value This study is among the first to examine how racial bias and AI voice technology intersect in student evaluations.