The debate about the impact of generative artificial intelligence (GenAI) on higher education is polarized. Some portray GenAI as normalizing cheating at scale (1), whereas others argue that misconduct patterns have changed little (2). These competing narratives underscore the need for reliable data on where GenAI use is concentrated and where misuse is most likely. Existing studies of GenAI adoption and perceptions in higher education provide useful early signals but often rely on small samples and lack measures designed to capture sensitive behaviors such as cheating across fields (3–5). We addressed this gap with survey data from 95,513 students in a representative sample of 20 major public researchintensive universities in the United States and an indirect method for estimating GenAI-assisted cheating across disciplines. We found substantial heterogeneity in GenAI use and misuse across disciplines and student groups. These patterns call for discipline-specific assessment reform, not blanket bans or universal detection regimes.