Summary
- Ofqual will propose changes to non-exam assessment in A-level history and all three A-level English qualifications because of difficulties authenticating unsupervised work.
- Awarding organisations have until 30 November 2026 to explain how they are responding to the misuse of wearable technology.
- The regulator expects appropriate staff training, including for examination invigilators, ahead of summer 2027; no final coursework reforms have been adopted.
England’s qualifications regulator is preparing changes to A-level coursework as generative artificial intelligence makes it increasingly difficult to establish whether students have completed unsupervised assessments independently.
Ofqual has confirmed that it will bring forward proposals covering non-exam assessment in A-level history and all three A-level English qualifications. Chief Regulator Sir Ian Bauckham outlined the position in a letter to awarding organisations on 18 November, alongside new expectations for tackling the use of connected wearable devices during examinations.
Exam boards have been asked to explain the measures they are taking to address wearable technology risks before 30 November. The regulator expects relevant staff, including invigilators, to receive consistent training ahead of the summer 2027 examination series.
The approach marks a shift towards reconsidering assessment arrangements where authenticating students’ work has become difficult. Ofqual has not abolished coursework or announced the replacement assessments for the qualifications identified, and any substantive reforms will require further consultation.
Unsupervised writing presents a growing problem
Non-exam assessment allows students to demonstrate knowledge and skills through work completed outside a conventional timed examination. Extended essays and research assignments can assess abilities that are difficult to measure through shorter written papers, including the selection of evidence and development of sustained arguments.
Generative AI complicates the process of determining whether submitted work represents a student’s own understanding. A student may use a system to develop an argument, prepare an outline, interpret sources or produce passages of text without that assistance being visible in the final document.
Teachers have traditionally relied on knowledge of a student’s work, the development of assignments and their professional judgement when identifying possible malpractice. Those methods remain relevant, but they may become less reliable when automated systems can reproduce increasingly convincing written responses.
Ofqual’s concerns do not establish that every use of AI in education is inappropriate. Students may use digital tools legitimately while studying, and assessment rules must distinguish permitted support from assistance that compromises the skills being measured.
Where an assignment is intended to demonstrate independent analytical writing, however, extensive unauthorised AI involvement can undermine the meaning of the resulting grade. Universities and employers may then be relying on a qualification that provides less reliable evidence of the student’s abilities.
Bauckham has indicated that attempting to detect AI-generated material after an assignment has been completed may not provide a sufficiently dependable long-term solution. Changing the conditions under which assessed work is produced could offer stronger assurance, although the appropriate arrangements will differ between subjects.
The regulator has identified history and the three A-level English qualifications as priorities because their non-exam assessments involve substantial unsupervised written work. It has not confirmed that every affected component will be removed or replaced by a timed examination.
Teachers favour stronger supervision
Survey evidence commissioned by Ofqual illustrates the pressure on current arrangements. In recent Teacher Tapp polling, 70% of secondary teachers surveyed said unsupervised written coursework should not contribute to final grades, while 83% supported greater supervision or other controls.
The findings represent teachers’ views of assessment arrangements rather than a measurement of how many students have engaged in malpractice. They nevertheless indicate concern among the professionals responsible for administering and assessing school qualifications.
In his speech at the Schools and Academies Show on 18 November, Bauckham linked the reliability of assessment to confidence in the qualifications awarded. His argument was that grades need to represent what individuals know and can do, rather than the ability of available technology to complete the assessed work.
Greater supervision could help establish a clearer connection between a student and the work submitted. However, additional supervised assessments would require examination space, staff time and procedures for supporting students with particular educational or accessibility needs.
There are also legitimate differences between the skills measured through timed examinations and those developed through extended research. An assessment completed under strict time constraints may not test the same ability to investigate complex information or refine an argument over several weeks.
Ofqual will therefore need to consider both authenticity and educational validity when developing its proposals. The objective is to strengthen confidence in assessment without unnecessarily removing activities that contribute to the value of the qualification.
Connected wearables challenge examination security
Alongside coursework, the regulator has raised concerns about devices that can be concealed or worn during examinations. Smart glasses and other connected accessories may provide recording, communication or information-display functions without requiring a student to handle an obvious electronic device.
Some products can also connect to AI services, introducing opportunities to obtain assistance or transmit information in ways that conventional examination checks may not readily detect.
Ofqual has asked awarding organisations to identify the safeguards they are introducing and explain where further action may be needed. The regulator expects a coordinated response that can be communicated consistently to schools and colleges.
Staff training forms an important part of that response because examination invigilators must understand the rules governing permitted equipment and know how to handle suspected malpractice. Changes to the design and appearance of wearable technology mean that relying on a fixed list of recognisable devices may become progressively less effective.
Controls will also need to distinguish prohibited assistance from legitimate accessibility technology. Students may require authorised equipment as part of reasonable adjustments, so examination procedures must address inappropriate use without assuming that every wearable device indicates misconduct.
The regulator has not mandated a particular electronic detection system or proposed routine automated surveillance of examination candidates. Its immediate requirement is for awarding organisations to explain their risk assessments and preventive arrangements, with appropriate training in place before the summer 2027 examination series.
Assessment technology faces greater scrutiny
Automated detection tools are sometimes used to estimate whether writing was produced by AI, but their findings can be uncertain. An indicator that text resembles generated material does not establish which system produced it or whether a student breached a particular assessment rule.
Such limitations become particularly consequential when an allegation could affect an entire qualification. Evidence used to investigate malpractice needs to be considered alongside the student’s work, the applicable rules and the circumstances in which an assignment was produced.
Ofqual has already issued guidance explaining how existing regulatory requirements apply to AI misuse. Its November intervention goes further by identifying assessment design as an area requiring potential reform rather than relying principally on detection and enforcement.
For awarding organisations and technology suppliers, this could eventually create different requirements for assessment administration, authentication and the management of examination environments. The precise implications will depend on the proposals published and the decisions taken after consultation.
Schools and colleges will also need clear information about any changes sufficiently far in advance to prepare teaching and assessment arrangements. The regulator has not established an implementation date for the proposed coursework reforms, so current qualification requirements remain in place.
Its next confirmed deadlines concern wearable technology: awarding organisations must respond before 30 November and prepare relevant examination staff for the summer 2027 series. The separate proposals for A-level coursework will determine how Ofqual intends to preserve reliable assessment as generative AI becomes more capable.












