A randomized classroom study suggests AI teaching assistants may harm student motivation and learning outcomes when teachers use generative tools to generate materials. Researchers reported that students whose teachers had access to a ChatGPT-based teaching assistant customized to Turkey’s curriculum rated classes as less enjoyable and less important, and that students of weaker instructors scored lower on final exams. The draft study, “Generative AI Can Harm Teaching,” follows earlier research arguing that student use of AI as an “answer machine” can damage learning. Lead author Alp Sungu said teachers may be using AI as a crutch—delegating lecture notes, assignments, and exams—reducing the interaction and instructional quality that support learning. For higher education and K-12 partners, the evidence is relevant to teacher training, AI procurement, and policy guidance: AI for content production may not translate into instructional improvement without guardrails and pedagogical accountability.