AI has made it 'close to impossible' to assess university coursework

England’s higher education regulator warns that widespread generative AI use is undermining universities’ ability to judge what students know and can do. An investigation by the Office for Students and Advance HE, based on interviews with more than 300 students and staff at 127 institutions, found AI use across essays, dissertations and open-book assessments. Students reported peers producing work in minutes with AI, earning good grades without detection, and sometimes being unable to explain what they had submitted. Academics described batches of submissions with similar writing, sources and style, and questioned whether existing assessments still measure students’ skills reliably. The regulator says the credibility of degrees and students’ learning are at risk.

The OfS urges universities to make much greater use of assessments that limit or directly address AI, including traditional in-person written exams, oral presentations and vivas. Other approaches include asking students to critique AI-generated work or document the steps they took to research and produce an assignment. Institutions may also incorporate AI under controlled conditions. The regulator cautions that failing to establish valid and reliable assessment could breach registration requirements and bring sanctions; the Department for Education expects stronger, durable rules. Universities UK says many institutions have already changed their practices, though further work is needed. The central challenge is to preserve trustworthy assessment while allowing universities and students to benefit from AI.