Universities in the US are redesigning teaching and testing approaches amid rising AI usage among students, raising concerns about skills development and assessment validity at a critical stage of young adults’ growth.
American universities are beginning to redesign how they teach and test students as evidence mounts that AI use is already routine during the very years when many young adults are still building judgement, self-control and academic independence. What had often been framed as a problem of schoolchildren and teens is now moving squarely into higher education, where the stakes include not just cheating but whether students are outsourcing the mental work a degree is supposed to develop.
The scale of that shift is already clear. Gallup reported in April that 57% of US college students use AI in coursework at least weekly and about one in five use it daily, even though 42% say their institution discourages AI and 11% say it prohibits it altogether. A separate Cornell and University of California, Berkeley study, published in Science and covered by Cornell Chronicle in May, drew on responses from more than 95,000 students at 20 US public research universities and found that 37% used generative AI at least monthly while 9% said they had used it to cheat. Use was far from evenly spread: Cornell said it reached 62% among computer science students, compared with 24% in the arts, and was higher among male students than female students. Rene Kizilcec, one of the study’s authors, said “Assessment reform is necessary and urgent” because misuse of AI threatens “assessment validity” and, with it, “the credibility of university credentials”.
Psychologists are also warning that the timing matters. In a cover story published on 1 September, the American Psychological Association said more than half of US adolescents now use AI chatbots for schoolwork, citing a 2026 Pew Research Center report, and stressed that research into the developmental effects is only just beginning. Anne Maheux told the APA that “Everything around youth and AI is essentially an open question right now” and that while there are benefits, the potential risks are more urgent to understand. Michael Robb, head of research at Common Sense Media, told the same publication that AI will be so embedded by the time today’s teenagers reach adulthood that society may barely notice it, making the real challenge working out where it belongs “without replacing the skills young people need to learn, grow, and function in society”.
Evidence focused specifically on young adults suggests those risks do not end at 18. A Microsoft Research paper presented in an extended version for AIES 2026 examined testimonials from chatbot users aged 18 to 25 and identified three ingredients of AI dependence: chronic use, efficiency and delegation. Researchers said the combination pointed to a pattern in which users increasingly hand work over to the system because it is faster and easier. Participants themselves reported “feelings of atrophy in abilities” and, in some cases, “feelings of inadequacy”. The authors argued that preventing lasting harm to young adults’ personal and social development should be treated as paramount.
That picture is more complicated than a simple crackdown on misconduct. Gallup found that the most common reason students gave for using AI was to understand difficult material, with nearly nine in 10 monthly users saying that mattered and 46% calling it extremely important. Saving time and getting better grades followed closely behind. In day-to-day study habits, 64% of student AI users said they relied on it daily or weekly to help with coursework they did not understand, 60% to check answers, and 54% both to improve writing and to summarise lectures or notes. Those figures suggest students increasingly see AI not as a forbidden shortcut but as ordinary academic infrastructure.
The technology is also reaching into parts of student life that universities have barely begun to address. Inside Higher Ed reported in April on research from The Rithm Project, based on a nationally weighted survey of nearly 2,400 people aged 13 to 24, including more than 500 young adults aged 18 to 24. The study found that 39% mainly used AI for information and tasks, but 18% used it for personal or relational support and 15% interacted with AI characters, described as the most intimate form of use. Alison Lee of The Rithm Project said many adults are overestimating dramatic cases of AI companionship while underestimating the quieter ways AI is reshaping relationships and connection. Laura Erickson-Schroth of the Jed Foundation added that adults are confronting these tools alongside young people rather than ahead of them, making it harder to offer guidance or set boundaries with confidence.
Colleges are already responding by trying to bring thinking back into the room. The Associated Press reported in March that universities including Cornell, the University of Pennsylvania and New York University are expanding oral exams, presentations and in-person assessments. At Cornell, biomedical engineering professor Chris Schaffer now uses oral defences and says, “You won’t be able to AI your way through an oral exam.” At Penn, Emily Hammer said the point is not merely policing dishonesty: “It comes across as if we’re trying to prevent cheating. That’s not why we’re doing this. We’re doing this because students are actually losing skills, losing cognitive capacity and creativity.” Cornell’s Centre for Teaching Innovation has begun offering an “Oral Assessment Workshop”, while NYU professor Panos Ipeirotis has gone further by pairing written work with an AI-powered oral exam, saying, “I don’t trust written assignments anymore to be the result of actual thinking.”
Even so, few researchers are arguing for blanket bans. Gallup said students expect universities to prepare them for a labour market in which AI will be unavoidable, and warned that institutions that ignore the tools may leave graduates behind. The APA struck a similar note, arguing that the crucial question is not whether students use AI but how and why they use it. Nicole Barnes told the association that “Teaching students when not to use AI, and requiring reflection on their choices, may help mitigate the risks around healthy development.” That points towards a middle course: preserve the parts of learning that require struggle, judgement and original thought, while teaching students how to use AI deliberately rather than reflexively.
Taken together, the latest reporting suggests that higher education may be the place where the debate over AI and development becomes most consequential. Children still have parents, teachers and school rules around them. University students have far more freedom, but they are also the group most likely to be told to manage an increasingly persuasive technology on their own. If campuses treat AI only as a plagiarism problem, they may miss the broader question raised by psychologists, technologists and educators alike: what happens when a generation reaches adulthood with powerful new tools in hand, but before its habits of thought are fully set?
Disclaimer: This content is for informational purposes only and is not intended to be a substitute for professional medical judgment, advice, diagnosis, or treatment.
