Proctoring in the Age of AI, Why Exam Design Still Comes First

Future-proofing online exams amid rising demand and tightening privacy laws.

The global online exam software market, valued at $9.4b in 2025, is expected to reach $28.7b by 2034, as training providers, professional bodies and educational institutions increasingly shift assessments online.

The popularity of online exams is easy to understand, as educational institutions are managing funding constraints and limited campus space, while employers and professional bodies are looking for more flexible ways to assess candidates. Remote testing allows individuals to balance learning and certification with work and personal commitments, making assessments more accessible than traditional, in-person formats.

Yet the industry’s biggest challenge isn’t scale; it is the age-old commodity, trust. As AI tools make cheating harder to detect, institutions are under growing pressure to prove that online assessments are fair, secure and compliant.

Recently, the Association of Chartered Certified Accountants made headlines after announcing its decision to end remote exams amid AI cheating concerns, while Ofqual has reported that teachers are finding AI increasingly difficult to identify in assessed work. Even Princeton University, one of the world’s top universities, has amended its 133-year-old Honor Code to introduce proctoring for in-person exams.

Given these pressures, it’s easy to assume that more proctoring is the solution. However, at a certain point, more surveillance doesn’t necessarily lead to better standards or more accurate results. In many cases, excessive monitoring can harm accessibility, reduce candidate trust and raise privacy concerns.

Proctoring Is Not a Shortcut for Good Assessment Design

Proctoring is an effective tool for preventing misconduct, but it is not a standalone solution for poor assessment design. In-person invigilators cannot watch every candidate simultaneously. But online proctoring has its own limitations, as students have been known to use smart devices to evade online monitoring tactics.

When an exam is focused on testing a candidate’s memory, such as a standardised multiple-choice exam, the incentive to cheat increases because there is a limited pool of correct answers to choose from. If that exam carries significant weight towards a final grade, that incentive only rises further. These exam formats are inherently more vulnerable to AI, which excels at retrieving and reproducing this kind of information.

When exams are designed to test a candidate’s ability to analyse information and apply theory to different contexts, the immediate advantage AI offers diminishes quickly. This better reflects modern professional environments, where employees are rarely required to operate entirely from memory.

In practice, this looks like a contextual question framed around a hypothetical scenario that asks candidates to justify their answers. When combined with time limits, browser lockdowns or copy-paste restrictions, even a well-constructed AI response becomes difficult to apply usefully. These questions demand original reasoning, not retrievable facts.

By delivering open-book exams, case studies and other critical thinking tasks under controlled conditions, institutions can develop assessment models in which security measures are distributed across multiple parts of the assessment process. This will allow them to reduce their reliance on monitoring because integrity is built into the exam structure rather than imposed through surveillance.

Although memory-based testing is still critical for testing a candidate’s foundational knowledge, institutions can gain a much more detailed understanding of a candidate’s ability when it is paired with other applied testing formats.

Exam weightings can also be restructured to mitigate cheating. If grading is spread more evenly between exams, candidates feel far less pressure to attain a high score, which naturally reduces the incentive to cheat. This offers another alternative to securing exams without defaulting to heavy surveillance.

Proctoring is essential in all exams, but it should never be the sole mechanism for maintaining integrity. Instead of focusing on different ways to monitor candidates, exam formats and questions should be engineered to make all forms of cheating, not just AI, more difficult by design.

Why Compliance Is Paramount

As institutions implement proctoring measures in online exams, they must also navigate privacy and data protection laws to ensure that surveillance methods are fair and legally compliant.

Under GDPR, capturing video feeds, audio recordings and other biometric data is considered highly sensitive; therefore, collection must be justified and candidates need to know how their data is being used. 

The EU AI Act has also introduced restrictions on fully automated exam monitoring, emphasising that high-stakes assessments must retain human oversight, auditable outputs and the ability for staff to intervene or override automated decisions.

At the same time, institutions should consider the long-term risk associated with storing vast amounts of proctoring data, as the more proctoring an online exam introduces, the more data will need to be managed over time. This represents a significant risk, given the high volume of cyber security attacks targeting the UK’s education sector.

This is why a one-size-fits-all proctoring model is no longer legally defensible. To avoid blanket surveillance, institutions should adopt a data minimisation mindset and configure proctoring models based on an exam’s risk level.

When thoughtful exam design is supported by proportionate proctoring, online exams become more secure. More importantly, it demonstrates to regulators and candidates that testing conditions are fair and compliant, while still providing the audit trail needed to validate results.

The Future of Assessment

The rate at which online exams are growing presents a huge opportunity for learning and professional development. However, it also raises a fundamental challenge in how exam security is defined and managed.

The training providers, professional bodies and educational institutions that thrive will be the ones that treat exam design and proctoring as complementary rather than relying on surveillance alone. Integrity is not something that can be monitored from the outside. It has to be built in from the start.