Quick answer: A camera watching a student is not enough to catch modern exam cheating, because AI-assisted shortcuts happen in browser tabs and copy-paste actions a webcam never sees. New academic research shows that analyzing how a student actually behaves during a test, things like losing focus, copying text, and right-clicking, flags suspicious patterns that proctoring alone misses. The smart approach is behavior analytics that surfaces risk signals, paired with a human who makes the final call. That balance of automated detection and human judgment is what keeps an online exam credible.
Online exams have a trust problem, and AI made it sharper. When a student can quietly ask an AI tool for an answer in another tab, watching their face on a webcam tells you very little. A peer-reviewed study presented in late 2025 put numbers to this: even in proctored settings, about a third of students showed behaviors that analytics flagged as suspicious. The camera was on the whole time.
The finding points to a better model for a smart online exam: stop relying on the camera alone, and start reading the behavioral signals that reveal how the test was actually taken.
What the Research Found
The study analyzed behavioral data from students during proctored online exams. By clustering signals like text selection, right-clicks, and moments when the exam window lost focus, the researchers separated normal test-taking from patterns that looked like outside help. Several of the suspicious clusters also scored noticeably higher than their peers, which is exactly the pattern you would expect if some students were getting answers from somewhere else.
The headline is not that proctoring is useless. It is that proctoring plus behavior analytics catches what proctoring alone cannot. A webcam sees a face. Behavior data sees the actions.
Why AI Should Assist, Not Judge
It is tempting to let an algorithm hand down a verdict. That is a mistake. An automated flag is a starting point, not a conclusion. A focus change might be a student checking the time, or it might be a second tab. The only way to tell the difference fairly is for a person to look at the context before any integrity decision is made.
This is the principle a credible online exam platform should be built on: surface the signals, rank the risk, and put a human at the decision point. It protects honest students from false accusations while still catching the patterns that matter.
Building This Into a Smart Online Exam
ICTLMS approaches assessment integrity as a layered system rather than a single camera feed. A secure exam environment limits what a student can do during the test, identity checks confirm who is sitting the exam, and behavioral signals give reviewers the context to spot patterns that matter. Advanced analytics capabilities continue to develop, and we present them honestly as they mature rather than overstating what is available today.
Industry practice lines up with the research. Combining a secure browser, identity verification, and analytics is reported to cut remote cheating dramatically compared with an honor-system exam, while keeping a human in the loop on every flagged case.
Related reading:
AI online exam software · How a smart online exam stays credible
Frequently Asked Questions
Why is webcam proctoring alone not enough?
A camera watches the student’s face, but AI-assisted cheating happens in browser actions, copied text, and tab switches a webcam never captures. Research shows behavior analytics catches suspicious patterns that proctoring on its own misses.
What behavioral signals indicate possible cheating?
Patterns like frequent focus loss on the exam window, unusual copy and right-click activity, and timing that does not match honest test-taking. None of these prove cheating on their own, which is why a human reviews the flagged cases.
Should an algorithm decide if a student cheated?
No. An automated flag is a signal, not a verdict. A person should review the context behind every flag before any integrity decision, both to be fair to honest students and to make the call defensible.
Does behavior analytics invade student privacy?
Done well, it focuses on exam-relevant actions rather than constant surveillance, and it pairs with clear disclosure to students. The goal is to read how the test was taken, not to monitor a person’s every move.
How does ICTLMS support exam integrity?
Through a layered approach: a secure exam environment, identity verification, and behavioral signals that give reviewers context, with a human at the decision point. Analytics features continue to develop and are presented honestly as they mature.
Get Started
Want to run online exams that stay credible without over-policing students? Contact our team and we will walk you through it.