News Article

Universities Urged to Rethink Assessment as AI Threatens Confidence in Qualifications

November 12, 2026

Students sit at individual desks throughout a large hall, taking a university examination while an invigilator stands on a stage in the background.

Universities and colleges in England are being urged to reconsider how they assess students following new research warning that artificial intelligence could undermine confidence in the knowledge and skills represented by higher education qualifications. The Office for Students (OfS) says institutions must ensure that assessments remain reliable as increasingly capable AI tools make it possible to produce convincing academic work without demonstrating the understanding that qualifications are intended to recognise.

Published on 11 November 2026, the research was undertaken jointly by the OfS and Advance HE. It draws on discussions with more than 300 students and staff from over 120 higher education institutions, examining how AI is changing teaching, learning, assessment and the development of digital skills.

The findings present a challenge for education providers across different delivery models. While the research focuses primarily on higher education, its questions about assessment integrity, learner independence and the responsible use of technology are relevant to the wider distance learning sector.

What has the Office for Students found?

The report, Making AI Work for Students: Reflections on Institutional Practice, examines the experiences of students and staff as generative AI becomes increasingly common in higher education.

Researchers found considerable variation in how institutions are responding. Some have developed innovative approaches to using AI within teaching and assessment, while others remain at an early stage of establishing policies, staff training and appropriate educational practices.

One of the most significant concerns relates to unsupervised assessment. Students and staff described situations in which AI could produce work meeting academic expectations without the learner necessarily demonstrating the knowledge or reasoning that the assessment was designed to measure.

The OfS also found that students are often uncertain about their institution’s rules on acceptable AI use. Although many universities have introduced policies, differences between departments and individual assignments can make expectations difficult to understand.

The regulator’s central concern is not simply whether students use AI. It is whether the methods used to assess them continue to provide trustworthy evidence of their own achievements.

Why AI presents a challenge for traditional coursework

Written assignments have long played an important role in education because they allow learners to research a subject, develop arguments and demonstrate understanding through extended work.

Generative AI changes the circumstances in which those assignments are completed. A student can now ask a system to produce an essay, summarise source material or construct an argument that appears academically credible, potentially without undertaking much of the underlying intellectual work.

The difficulty for education providers is distinguishing between a learner who has used AI appropriately to support their thinking and somebody who has relied on it to replace the thinking being assessed.

The OfS research identifies particular risks where assessments can be completed independently without supervision. A finished submission may appear to meet the required standard while providing insufficient assurance that the student possesses the relevant knowledge.

The problem extends beyond deliberate misconduct. If an assessment can be completed successfully without engaging with the intended learning, its design may no longer provide a reliable measure of achievement.

The regulator wants more secure assessment

In response to the findings, the Office for Students is encouraging institutions to make greater use of assessment methods that allow them to establish what students can demonstrate independently.

These can include supervised examinations, oral presentations, practical assessments and other activities conducted in controlled environments. Such approaches may provide stronger evidence of individual achievement where unrestricted AI assistance would otherwise undermine the purpose of an assessment.

The regulator is not proposing that every assessment should return to traditional written examinations. Different subjects require different ways of demonstrating competence, and some learning outcomes are better assessed through projects, practical work or professional activities.

The important requirement is that assessment remains valid and reliable. Institutions need to consider whether their chosen methods genuinely test the knowledge, understanding and skills that students are expected to develop.

This may involve redesigning existing assignments rather than simply adding restrictions to assessment instructions. A task that can be completed entirely by an AI system may need to be reconsidered if its purpose is to establish the learner’s independent capability.

AI detection software is not a complete solution

One possible response to AI-generated assignments is to rely on software designed to identify whether submitted work was produced using artificial intelligence.

The OfS research cautions against treating such systems as a dependable answer. Evidence reviewed by the researchers indicates that AI detection tools do not consistently or reliably identify AI use.

That creates difficulties for institutions attempting to establish whether individual students have breached academic rules. An unreliable detection result cannot provide the same assurance as an assessment method that directly demonstrates the learner’s knowledge.

There are also questions about fairness. Students need to understand what is permitted, how concerns will be investigated and what evidence may be used when academic integrity is questioned.

The findings therefore support a broader approach centred on assessment design, clear policies and appropriate academic judgement rather than expecting detection technology to resolve every problem.

What does this mean for distance learning?

Distance education has long relied on a combination of independent study, tutor feedback and assessments completed away from a traditional classroom. These arrangements can provide considerable flexibility, particularly for adults balancing education with employment or caring responsibilities.

The growth of generative AI makes it increasingly important to distinguish between flexible learning and assessment without sufficient safeguards.

A learner may study independently at home while still completing examinations or practical assessments under controlled conditions. Other qualifications may use portfolios, observed activities or assessment methods that allow a provider to establish the authenticity of the work.

As our guide to recognised distance learning qualifications explains, the credibility of a qualification depends on its awarding and assessment arrangements rather than simply where teaching takes place.

The OfS findings do not establish that distance learning is inherently less reliable than classroom education. They instead raise questions about assessment methods that can arise wherever students complete substantial work without direct supervision.

For distance learning providers, the challenge is to preserve the flexibility learners value while ensuring that assessment remains a meaningful demonstration of achievement.

Assessment needs to reflect the subject being taught

A single approach to AI and assessment is unlikely to work across every academic and vocational discipline.

In some subjects, independently recalling information or demonstrating a calculation without assistance may be essential. In others, the ability to use digital tools responsibly could form part of the professional capability that a qualification is intended to develop.

The OfS recognises that assessment can sometimes incorporate AI appropriately, particularly where learners are expected to demonstrate how they apply knowledge in realistic situations.

Such assessments still need careful design. Allowing AI does not remove the requirement to establish what the student understands, what decisions they have made and whether they can evaluate the results.

A learner might, for example, be asked to assess the accuracy of an AI-generated explanation or justify decisions made during a practical project. The educational value would lie in the learner’s reasoning rather than the system’s ability to produce an answer.

This distinction is particularly relevant to vocational education, where professional competence often involves using tools appropriately while remaining accountable for the quality of the work produced.

Students also need opportunities to develop AI skills

Although assessment integrity is a central concern, the OfS report does not recommend excluding artificial intelligence from education altogether.

The research found that students increasingly regard AI literacy as relevant to their future employment. Many would welcome clearer guidance and more opportunities to develop the knowledge needed to use AI effectively and responsibly.

The report also identifies uneven access to AI training. Some institutions have introduced substantial programmes, while students elsewhere reported receiving little structured instruction.

Our recent analysis of AI literacy for learners and workers examined why the ability to question, verify and responsibly use AI is becoming an important part of wider workplace readiness.

The OfS findings reinforce that distinction. Teaching students to use AI should involve more than familiarity with particular applications. Learners also need to understand their limitations, evaluate outputs and recognise when independent reasoning is essential.

Education providers consequently face two connected responsibilities: protecting the credibility of assessments while helping students develop capabilities they may need in employment.

The risks of allowing AI to replace learning

Artificial intelligence can help learners understand unfamiliar material, generate practice questions and explore different explanations of a difficult concept. Used carefully, these functions may support independent study.

The concern arises when assistance replaces the activities through which learning takes place.

The OfS research discusses the risk of students transferring too much intellectual effort to AI systems. If a learner repeatedly asks a tool to complete reading, analysis or problem-solving tasks, they may produce satisfactory outputs without developing the knowledge those activities were intended to build.

For distance learners, who often manage much of their own study, understanding that distinction is particularly important.

Independent education requires more than completing tasks. It involves engaging with material, identifying misunderstandings and gradually developing the ability to apply knowledge without continuous assistance.

The relationship between independence and support is central to effective self-paced learning. AI may become another useful resource within that process, provided it does not remove the learner’s responsibility to understand the subject.

Staff training and institutional leadership are also under scrutiny

The research identifies differences not only between students but also among the staff responsible for teaching and assessment.

Some academics are experimenting with AI-supported learning activities and teaching materials, while others use the technology infrequently. Staff training is often voluntary, with participation varying considerably between institutions.

The report also suggests that institutional responses have frequently developed through the efforts of individual departments or enthusiastic staff members rather than a consistent organisation-wide approach.

The OfS and Advance HE argue that stronger leadership can help institutions establish clear expectations, develop staff capability and ensure that AI is considered within teaching, assessment and quality assurance.

This has implications for education providers beyond universities. Where learners encounter different instructions from different tutors or departments, uncertainty about acceptable AI use can undermine confidence in the rules.

Clear policies are therefore only part of the solution. Staff need the knowledge and support to apply them consistently and explain their purpose to learners.

What action could the Office for Students take?

The OfS has stated that institutions must continue meeting their existing regulatory obligations concerning the quality of education and the validity of assessment.

The regulator intends to provide further clarity and share effective practice rather than introduce new AI-specific registration conditions at this stage.

Where evidence indicates that a registered provider is failing to meet relevant requirements, the OfS can use its existing regulatory powers. Depending on the circumstances, these can include imposing specific conditions or taking other enforcement action.

The regulator has also indicated that assessment resilience will be relevant to its proposed future approach to quality regulation.

These statements should not be interpreted as an immediate prohibition on unsupervised coursework or a requirement for every institution to adopt the same assessment model. Providers remain responsible for determining appropriate methods that satisfy the applicable standards.

The regulatory direction is nevertheless clear: institutions need to demonstrate that the qualifications they award continue to reflect students’ actual knowledge and capabilities.

What should learners expect from their education providers?

Students should expect clear guidance about when AI may be used and when work must be completed independently.

That guidance should distinguish between ordinary learning activities, such as using a tool to explain a difficult topic, and formal assessments where assistance may be restricted or subject to disclosure requirements.

Learners should also understand how their work will be assessed and why particular methods are used. A supervised examination, oral discussion or practical demonstration may be intended to provide assurance that the learner can apply knowledge independently.

Where AI is incorporated into teaching, students should receive sufficient support to use it responsibly. Access to a technology does not automatically mean that somebody understands its limitations or can judge the accuracy of its outputs.

The quality of communication between learner and provider remains important. As our article on good distance learning tutor support explains, clear explanations and constructive feedback are essential to helping learners develop genuine understanding.

For prospective students, questions about assessment arrangements and academic integrity policies are becoming increasingly relevant when comparing courses.

Protecting qualifications while preparing learners for the future

The OfS and Advance HE research reflects a fundamental tension facing education. Artificial intelligence is becoming more capable and more widely used, yet qualifications must continue to provide credible evidence of what individuals know and can do.

Simply preventing all AI use would overlook opportunities to improve learning and develop relevant workplace skills. Allowing unrestricted use without reconsidering assessment, however, could weaken the relationship between completing a course and demonstrating genuine achievement.

The challenge for providers is to distinguish between appropriate technological assistance and activities that students must undertake themselves.

For the wider distance learning sector, this is an important quality assurance issue. Flexible delivery can remain a credible route to recognised qualifications, but assessment methods must continue to establish the learner’s own knowledge and competence.

The Office for Students’ latest findings underline the importance of balancing innovation with educational standards. As AI becomes a routine part of studying and working, confidence in qualifications will depend increasingly on how convincingly providers can demonstrate that learners have achieved what their awards claim to recognise.