One-third of college students use AI for assignments; 9% admit to cheating with it
About one-third of undergraduates at 20 major U.S. research universities regularly use generative AI and LLM tools like ChatGPT to complete assignments, according to a Cornell University study published in Science. Nine percent have used the technology to cheat.
The survey of more than 95,000 students, conducted during the 2023-24 academic year, represents the largest assessment of AI use among college students to date. Researchers found that daily AI users cheat at rates five times higher than monthly users-26% versus 7%.
Usage patterns vary sharply by field. Sixty-two percent of computer science students report regular AI use, compared to 24% in the arts. Students in data-heavy disciplines adopt the tools at higher rates.
Demographic gaps in AI access and use
The research revealed inequities in how students access and use AI. Forty-five percent of male students used AI regularly, compared to 33% of female students. White and Asian students reported usage at 39%, while students from underrepresented racial minorities reported 29%.
These gaps may reflect unequal access to AI tools and training. Researchers warned the disparities could widen as AI becomes more specialized and expensive, creating long-term disadvantages in both college learning and job market readiness.
Universities must change how they assess students
The study calls for urgent reform in assessment practices. Rene Kizilcec, associate professor of information science at Cornell, said the findings raise questions about whether university credentials remain valid when students use AI to produce work.
The researchers propose three approaches: return to controlled testing environments with proctors and pen-and-paper exams; establish clearer guidelines for acceptable AI use; or redesign assignments to incorporate AI in ways that demonstrate professional skills.
Professional societies should help define assessment standards within their disciplines, the authors said. Universities must also address equity concerns to avoid widening educational disparities as AI literacy requirements increase.
As AI for Education becomes more common, misuse will likely grow. The challenge now is designing assessments that maintain academic integrity while preparing students to work alongside these tools.
Your membership also unlocks:
