Grades without learning? How AI reshapes student incentives and assessment

Generative AI helps students earn higher grades in less time, but AI-assisted performance does not necessarily mean greater learning. As grades become less closely tied to actual skills, they also become less informative to employers. We develop a model of how students divide their time among genuine study, AI-assisted graded work, and leisure. When grades matter for scholarships, future wages, and other opportunities, AI use becomes more attractive to students. Accordingly, course design is critical in countering this incentive by redirecting time from AI-use toward assessments that require students to demonstrate their actual skills. Detection and penalties reinforce these incentives.