Online exams have made assessment more flexible and accessible. Students can complete exams remotely, institutions can serve learners across locations, and faculty can reduce some of the logistics associated with in-person testing.
But the purpose of an exam hasn’t changed.
When an assessment is designed to measure a student’s knowledge or competency, institutions need confidence that the person taking the exam is actually the person enrolled—and that the answers reflect their own knowledge.
Generative AI has made that more complicated.
AI has changed what unauthorized assistance can look like
The risks associated with remote exams aren’t entirely new. Without the right controls, a student may consult notes or textbooks, search for answers online, communicate with another person, or even have someone else complete an assessment on their behalf.
Generative AI adds another readily available source of assistance.
And unlike traditional unauthorized resources, AI can generate answers, explain concepts, solve problems, summarize information and produce written responses within seconds.
For Canadian postsecondary institutions, this isn’t a hypothetical issue. In a 2026 Academica Group survey of 1,134 Canadian postsecondary students, 84% said they had experimented with or used generative AI. Among those students, 73% had used it for coursework.¹
That doesn’t mean those students are cheating. In fact, much of their reported use involved legitimate learning activities such as clarifying concepts and studying. But it does illustrate an important reality for universities and CEGEPs:
AI is already part of students’ academic environment.
The question is no longer whether students have access to it. The question is how institutions distinguish between appropriate AI use and situations where an assessment needs to measure a student’s independent knowledge.
AI can support learning—but it shouldn’t replace it
There are good reasons to incorporate AI into education.
Used appropriately, generative AI can help students explore a difficult concept, brainstorm ideas, receive explanations or work through material in a different way. Teaching students to use these tools critically will likely become an increasingly important part of higher education.
But there is a meaningful difference between learning with AI and allowing AI to do the thinking during an assessment intended to evaluate individual competency.
Emerging research makes that distinction worth paying attention to.
A 2025 study from Microsoft Research involving 319 knowledge workers found that greater confidence in generative AI was associated with less critical thinking. The researchers also found that AI changes where cognitive effort occurs: instead of gathering information or solving a problem directly, users may spend more effort verifying and integrating AI-generated responses.²
An exploratory 2025 MIT Media Lab study looked specifically at essay writing. Participants writing without digital assistance demonstrated stronger and more distributed brain connectivity than those using a large language model. The LLM group also had more difficulty accurately recalling and quoting from essays they had just produced.³
These findings should be interpreted cautiously—the MIT research involved a relatively small sample and further research is needed. But they reinforce a fundamental principle of assessment: if the objective is to determine what a student knows and can do independently, the assessment process needs to preserve that distinction.
A policy alone can’t secure an online exam
Clear institutional policies around AI are essential.
Students need to understand when AI is encouraged, when it is permitted with restrictions, and when its use constitutes unauthorized assistance.
But for high-stakes online exams, telling students not to use AI may not be enough.
The University of Toronto, for example, explicitly identifies the use of an unauthorized aid or unauthorized assistance during an examination as an academic offence, with its academic integrity resources specifically addressing ChatGPT and other generative AI tools.⁴
For an online exam where outside resources aren’t permitted, institutions also need mechanisms that support those rules.
Depending on the assessment, that can mean verifying the test-taker’s identity, controlling access to unauthorized resources, monitoring the exam environment and reviewing what happened during the session.
Why AI-assisted cheating makes human review more important—not less
AI is evolving quickly.
New models, applications and ways of accessing AI appear regularly. A process designed around detecting today’s tools can eventually encounter something it wasn’t designed to recognize.
That’s one reason institutions should be cautious about relying entirely on automated detection.
Technology can provide an important first layer of protection by helping restrict unauthorized activity and identify behaviours or events that warrant attention. But for consequential assessments, a human review of the session provides an additional layer of validation and context.
Rather than asking an algorithm to make the final judgement, reviewers can examine relevant evidence and help institutions make informed decisions according to their own academic integrity policies.
That combination—technology plus human oversight—can create a more defensible assessment process than relying on automation alone.
Is your online exam process ready?
AI isn’t going away, so higher education needs to focus on using it responsibly while protecting the integrity of independent assessment.
The challenge is to determine where AI belongs in the learning process and where independent assessment still matters.
For universities and CEGEPs offering remote exams, that means looking beyond whether an online exam technically works. Institutions should also ask:
- Can we verify who is actually taking the exam?
- Can we restrict unauthorized online and physical resources when required?
- How does our process address access to generative AI during an exam?
- Do we have sufficient evidence to review a questionable session?
- Is there human oversight for cases that require context or judgement?
- Can we apply our academic integrity policies consistently in a remote environment?
As AI becomes more capable and accessible, protecting exam integrity requires processes that can evolve with it.
uxpertise offers an online exam platform with integrated proctoring for institutions that need to deliver secure remote assessments. Our solutions include online proctoring powered by Integrity Advocate, combining technology with human review to support exam integrity.
Looking at how to strengthen your institution’s online exam process? Contact our team to explore online proctoring options for your university or CEGEP.
Sources
- Academica Group, How are Canadian postsecondary students using generative AI today?, May 2026.
- Lee, H.-P. et al., The Impact of Generative AI on Critical Thinking: Self-Reported Reductions in Cognitive Effort and Confidence Effects From a Survey of Knowledge Workers, CHI 2025, Microsoft Research.
- Kosmyna, N. et al., Your Brain on ChatGPT: Accumulation of Cognitive Debt when Using an AI Assistant for Essay Writing Task, 2025 preprint.
- University of Toronto, Academic Integrity, Using ChatGPT or other generative AI tool on a marked assessment.
