A major study using millions of student interactions found generative AI is changing math work patterns after ChatGPT’s arrival—students are completing word problems faster while learning less. The analysis drew on data from McGraw Hill’s ALEKS platform used by more than four million students annually. Researchers compared word problems, which can be copied into AI chatbots for quick answers, with graphing problems that are harder to outsource because they require screenshot uploads and recreation inside ALEKS tools. After late 2022, time spent on word problems fell steadily while graphing problem time stayed more constant. The report says unsupervised placement exams also showed similar declines in time on word problems, but proctored exams restored time to historical norms—suggesting monitoring affects behavior change. The study attributes average learning shortfalls to students who may spend only seconds on word problems by relying on AI. For higher education, the findings raise immediate assessment and learning-support questions: how institutions test readiness, detect AI-assisted shortcuts, and redesign math practice environments to preserve skill development.