AI Crisis Forces UK Universities to Rethink Exams

The rapid expansion of artificial intelligence in higher education has created a serious challenge for universities across England, with the country’s higher education regulator warning that traditional methods of assessing students are becoming increasingly unreliable. As growing numbers of students use AI tools to prepare essays, dissertations and other academic assignments, university lecturers are finding it increasingly difficult to determine whether submitted work genuinely reflects students’ knowledge, understanding and independent abilities.

The Office for Students (OfS), England’s higher education regulator, has warned that widespread and potentially undetectable AI use could undermine the credibility of university qualifications. Following an extensive investigation involving students, academics and higher education institutions, the regulator has called for universities to reconsider their assessment systems, including a greater return to traditional written examinations and face-to-face presentations.

The warning comes at a time when generative AI platforms are becoming increasingly accessible to students. Tools capable of producing essays, analysing academic literature, answering complex questions and generating research-based content within minutes have transformed the way many students approach their studies. While these technologies offer opportunities to improve learning and academic productivity, their growing use has raised fundamental questions about academic integrity and the value of university qualifications.

According to the OfS, some lecturers believe that assessing coursework has become close to impossible in certain academic areas because they cannot reliably establish how much of a student’s submitted work has been produced independently. The concern is particularly significant for essays, dissertations and open-book examinations, where students may have access to AI tools throughout the preparation process.

The regulator has warned that universities failing to establish valid and reliable methods of assessing students’ knowledge and skills could breach their registration conditions. Such failures could potentially result in regulatory action and sanctions.

Josh Fleming, the OfS director of strategy and delivery, described the increasing use of AI in higher education as a generational risk to the existing university system. He warned that the technology could undermine both the learning process and institutions’ ability to assess students fairly.

Fleming said universities needed to make significantly greater efforts to address these risks, including expanding the use of assessments that can be completed without access to AI tools.

He also acknowledged that the speed and scale of technological change presented an unusual challenge for universities and regulators alike. According to him, the higher education sector must learn to benefit from AI while establishing effective safeguards against its potential misuse.

The regulator’s investigation was conducted in partnership with Advance HE, a charity focused on improving university governance, teaching and higher education standards. Researchers interviewed more than 300 students and university staff across 127 universities and colleges during 2026 to examine how institutions were responding to the growing presence of AI in education.

The findings revealed that AI use had become widespread across higher education. Students interviewed during the research described situations in which their classmates used artificial intelligence to complete academic assignments in extremely short periods, sometimes within just a few minutes.

Some students also reported that individuals who relied heavily on AI received grades comparable to those achieved by students who completed their work independently. This has created concerns about fairness, particularly among students who invest considerable time and effort in researching, writing and preparing their assignments without relying on automated systems.

The investigation also identified concerns about whether students genuinely understood the work they submitted. Researchers reported that some students were unable to recall basic details about assignments they had recently completed.

Such findings have intensified questions about whether existing assessment practices can accurately measure academic knowledge when students can use AI to produce sophisticated answers without necessarily understanding the underlying subject matter.

One student told researchers that classmates had openly admitted to writing entire essays using ChatGPT without being detected by university staff. According to the student, receiving satisfactory grades encouraged some individuals to repeat the practice in subsequent assignments.

These accounts highlight a growing problem for universities: students who use AI dishonestly may gain an academic advantage over those who complete their assignments independently. If such practices become widespread, the distinction between genuine academic achievement and artificially generated performance could become increasingly difficult to establish.

University lecturers have also raised concerns about similarities between student submissions. Some academics reported receiving large numbers of assignments that contained similar arguments, referenced the same academic literature and followed remarkably comparable writing styles.

Although similarities between academic submissions do not automatically establish misconduct, the increasing availability of generative AI has made it more difficult for lecturers to distinguish independent work from assignments produced with automated assistance.

One academic involved in the investigation warned that academic integrity was facing serious risks, arguing that coursework was now frequently written either partly or entirely with the assistance of artificial intelligence.

Another university staff member suggested that the widespread adoption of AI was exposing fundamental weaknesses in existing assessment methods. The academic questioned whether some university assessments remained suitable for measuring students’ actual abilities in an environment where AI-generated work had become increasingly common.

In response to these concerns, the OfS has highlighted several alternative assessment methods that universities are already introducing. These include traditional examinations conducted under supervised conditions, oral examinations known as vivas and face-to-face presentations in which students must demonstrate their understanding directly to academic staff.

Traditional written examinations could become particularly important because they allow universities to assess students under controlled conditions without routine access to external AI tools. In-person presentations and oral examinations can also help lecturers evaluate whether students understand the arguments, research findings and conclusions contained in their assignments.

However, the regulator’s recommendations do not suggest that universities must abandon every form of coursework or reject AI technology altogether.

Some institutions are experimenting with assessment methods that require students to critically examine, edit or challenge AI-generated material. Rather than simply producing an essay, students may be asked to identify inaccuracies, assess the reliability of AI-generated arguments and explain why particular conclusions are academically appropriate.

Another emerging approach involves process-based assessment. Under this method, students are required to demonstrate how they developed their assignments, including their research methods, sources, drafting decisions and analytical processes.

Such approaches could allow lecturers to gain a better understanding of how students approach academic problems, even when AI tools are used as part of the learning process. They may also encourage students to develop critical thinking skills rather than relying exclusively on automated content generation.

The Department for Education has also emphasised the need for universities to take the issue seriously. A spokesperson said the government expected institutions to address the risks associated with AI and expected the OfS to enforce existing academic standards effectively.

The department also called for the regulatory framework to be strengthened quickly so that expectations surrounding the use of artificial intelligence become clear, robust and sustainable.

The challenge facing universities is not simply identifying students who have used AI. It is establishing clear boundaries between acceptable technological assistance and academic misconduct. Students may legitimately use AI to improve grammar, organise research, understand difficult concepts or receive feedback on their writing. However, submitting AI-generated work as evidence of personal knowledge without meaningful independent contribution raises serious questions about academic honesty.

Universities must therefore develop transparent policies explaining which forms of AI assistance are permitted and which activities violate academic regulations. Students also need appropriate guidance to understand their responsibilities and avoid unintentionally breaching institutional rules.

There are broader implications for graduate employment and the credibility of higher education qualifications. Employers depend on university degrees as indicators of professional knowledge, analytical ability and subject-specific competence. If institutions cannot confidently establish whether graduates possess the skills represented by their qualifications, public and professional confidence in higher education could weaken.

At the same time, completely excluding AI from university education may not be a realistic long-term solution. Artificial intelligence is increasingly being incorporated into workplaces across numerous industries, making the ability to use these tools responsibly an important professional skill.

Universities therefore face the difficult task of preparing students for an AI-driven economy while ensuring that academic qualifications continue to represent genuine learning and individual achievement.

The findings of the OfS investigation suggest that the traditional relationship between coursework, academic independence and university assessment is undergoing significant change. The coming years could see institutions introduce more supervised examinations, direct presentations and practical evaluations alongside carefully designed coursework.

Ultimately, the objective will be to create an assessment system that recognises the benefits of technological innovation without allowing artificial intelligence to replace the intellectual development universities are expected to provide.

As universities across England reconsider their examination systems, the central question is becoming increasingly important: how can institutions ensure that a degree represents what a student actually knows and can do? The regulator’s warning suggests that finding a credible answer will be essential to protecting the future of higher education.