Online exam integrity becomes more credible when schools design assessments around evidence of learning, then apply secure supervision at the points where independent performance must be verified. That wider approach allows school assessment leaders to distinguish between the conditions needed for learning and those required for assurance.
Monitoring remains valuable, but it no longer has to carry the entire burden of integrity. Generative artificial intelligence has made it more important to consider what an assessment reveals about a pupil’s knowledge, reasoning and ability to apply what they have learned. The central question is no longer simply whether unauthorised tools were available, but whether the evidence supports a confident judgement of mastery.
AI Makes Evidence Of Learning The Stronger Standard
Use of generative AI is becoming part of ordinary academic practice. The Student Generative AI Survey 2025 from the Higher Education Policy Institute found that 92 per cent of surveyed undergraduates used AI in some form, while 88 per cent had used it to support assessed work. Although the research concerns higher education, it points to the environment schools are preparing pupils to enter.
Access to AI does not make reliable assessment impossible. It creates a clearer distinction between tasks intended to support learning and tasks intended to verify independent capability. Pupils may use AI responsibly during research, drafting or formative work while still completing selected assessments under conditions that establish what they can do without assistance.
School assessment leaders can make that distinction explicit when planning a programme. Each task should have a defined purpose, permitted forms of assistance and an agreed standard of evidence. Clear boundaries give teachers a stronger basis for interpreting results and give pupils a more consistent understanding of what responsible AI use means.
Integrity Now Begins With Assessment Design
Assessment becomes more resistant to borrowed or generated answers when it asks pupils to reveal how they reached a judgement. Questions that require a choice of method, an explanation of reasoning or the interpretation of unfamiliar evidence make understanding more visible.
A 2025 review in Contemporary Educational Technology examines assessment strategies that foster critical thinking and deeper cognitive engagement. Its focus is higher education, but the design principle is relevant to schools: evidence of reasoning is often more informative than a polished final response alone.
In practice, an assessment director might combine a written response with a brief oral explanation, ask pupils to connect an answer to earlier classroom work or introduce new evidence that requires them to adapt a familiar method. None of these formats needs to be complicated. Their value lies in making the path to the answer part of the evidence.
Confidence also grows when judgement rests on more than one assessment event. A sequence can combine supervised tests, classroom work, project evidence and oral explanation. Consistent results strengthen the interpretation. A sharp divergence gives teachers a reason to look more closely at the context, rather than relying on an automated suspicion score or a single unusual response.
Supervision Works Best When It Is Proportionate
Purposeful design does not remove the need for controlled examinations. Schools still require selected moments when identity, conditions and independent performance can be verified, particularly where results influence qualifications, progression or access to further study.
The appropriate level of supervision depends on the consequence and purpose of the assessment. A lower consequence classroom quiz may need few controls, while an important examination may justify identity verification, defined room rules, technical checks and trained invigilation. Institutions managing dispersed candidates or large cohorts may consider outsourced proctoring services as one operational option within that broader model.
Consistency matters more than the volume of monitoring. Procedures should define what is recorded, which events require escalation and who reviews the evidence. Automated flags can draw attention to an incident, but they should not determine its meaning.
Ordinary behaviour can appear unusual when reduced to isolated data. A pupil may look away while thinking, lose connectivity or require an approved adjustment. Human review allows the school to consider the assessment rules, available evidence and individual circumstances before reaching a judgement. Documented review processes also help similar incidents receive comparable treatment across subjects and cohorts.
Secure Assessment Adds Confidence To A Layered System
Although written for tertiary providers, Enacting Assessment Reform in a Time of Artificial Intelligence, published by Australia’s higher education regulator, offers several models that are relevant to school assessment planning.
One pathway includes at least one secure task in every unit or subject, supported by straightforward identity verification and controlled conditions. It also recognises varied formats, including oral presentations, supervised practical demonstrations and classroom tasks. The report presents this as one possible pathway rather than a universal formula, allowing institutions to select an approach suited to their context.
The underlying principle transfers readily to schools. Not every task needs to become a locked examination. Selected assessments can establish independent achievement, while other work supports collaboration, research, feedback and responsible AI use.
School leaders can put this layered model into practice by coordinating curriculum design, assessment technology, teacher preparation and incident review. Rules should identify permitted assistance, secure tasks should test the capabilities they are intended to verify, and review procedures should preserve professional judgement.
Online integrity is therefore becoming less about recreating an examination hall through a webcam. A more durable model combines purposeful assessment design, secure evidence at meaningful points and informed human interpretation. Its aim is not to create the appearance of control, but to ensure that a result remains credible as the technology surrounding assessment continues to change.
