Universities in England are being urged to rethink how they assess students as generative AI makes it harder to tell what coursework demonstrates. The Office for Students (OfS) says the problem puts fair grading and confidence in degrees at risk, and warns institutions that fail to develop reliable assessment methods could face sanctions.
Coursework can conceal how students learned
A joint OfS and Advance HE survey drew more than 300 respondents at 127 institutions. It found that students use AI extensively for essays and dissertations, often without being detected, and can receive grades similar to students who complete the work without it.
The survey points to a gap between the submitted work and what some students can explain. The report says some students produced entire papers with ChatGPT but could not recall basic content afterward. Instructors also described batches of submissions with the same sources and similar writing styles.
Those patterns make written work harder to interpret as evidence of a student's own understanding. A polished essay may no longer tell an instructor, by itself, how much of the research, reasoning or writing the student carried out.
OfS calls for assessments that show understanding
OfS strategy director Josh Fleming described the issue as a “generational risk” to the credibility of degrees. He called for “far greater use of AI-free assessment,” while the regulator says institutions need dependable ways to assess student work.
The concern is not limited to whether AI appears in a submission. If a university cannot reliably judge what a student knows, it becomes more difficult to be confident that grades reflect learning. The regulator's warning puts pressure on institutions to show that their assessment methods can still support that judgment.
Traditional exams and in-person presentations are among the formats the OfS is urging universities to bring back. These approaches can require students to demonstrate knowledge directly, rather than relying only on a piece of coursework whose authorship or process may be unclear.
Universities are changing how work is assessed
Some universities are already moving toward oral exams, process-based assessments and exercises that ask students to critically revise AI-generated text. These methods place more emphasis on how students explain decisions, develop an answer or evaluate material.
Each approach changes what an assessment asks students to demonstrate. An oral exam can make understanding visible through a student's responses; a process-based task can focus attention on how work develops; and revising AI-generated text can test a student's ability to assess and improve it.
The challenge for institutions is to choose formats that provide reliable evidence of learning. The survey's findings suggest that simply collecting essays may not be enough when AI-assisted work can go undetected and look similar to work completed without it.
Government expects a durable response
Britain's Department for Education expects universities “to take this issue seriously.” It also wants the OfS to ensure its regulatory framework is “explicit, strong and durable on the use of AI.” The department's position adds a broader expectation that universities address AI use through clear, lasting oversight.
For universities, the pressure now comes from both the practical problem of grading coursework and the regulator's warning about sanctions. They are being asked to develop assessment formats that make student understanding easier to judge as AI becomes part of the academic environment.
The shift toward exams, presentations and work that exposes a student's reasoning is one response to that problem. Whether those formats can restore confidence in assessment will depend on institutions using them in ways that reliably show what students have learned.