A randomized classroom trial reported online in June finds that providing teachers access to an AI teaching assistant can reduce student motivation and depress exam outcomes, especially where instructors were already weaker. The study, “Generative AI Can Harm Teaching,” led by Wharton’s Alp Sungu, draws on a spring 2025 experiment across 193 teachers and more than 2,800 middle and high school students at a private school chain in Turkey. Teachers were randomly assigned either AI access—customized to Turkey’s national curriculum—or business-as-usual instruction. Over 10 weeks, teachers primarily used the tool to generate lecture notes, assignments, and exams. Students whose teachers had the AI rated classes as less enjoyable, less interesting, and less important; researchers also reported lower standardized final exam scores in key subgroups. The result is a caution for institutional AI adoption: even when AI accelerates content production, it may degrade teaching quality if it substitutes for instructor engagement and planning.