New research suggests AI classroom tools may deliver different writing feedback depending on how student profiles are described by race, gender, motivation level, and learning disability status. In a study from Stanford University, researchers submitted the same middle-school essays to four AI models multiple times with varying descriptors. The study found consistent patterns across models: Black-labeled essays received more praise and encouragement, while essays labeled as written by Hispanic students or English learners triggered more grammar and “proper English” corrections. Feedback also shifted tone by gender and academic motivation signals. The findings raise immediate questions for universities adopting AI tutoring, assessment, or learning-support tools: how feedback is governed, audited, and validated for fairness before large-scale deployment. For academic leaders and academic technology teams, the report strengthens the case for transparency, bias testing, and documented oversight of AI-driven formative assessment tools.