AI & Manual Proctoring for Secure Online Exams

Not every online exam needs video proctoring. For most school quizzes, coaching mocks, and basic hiring screens, built-in controls — timing, shuffle, fullscreen, tab blocking — are sufficient. Proctoring adds meaningful value when the stakes are high enough that a disputed result creates a real institutional, legal, or career-affecting problem.

Final semester exams that affect academic standing, professional certifications with career implications, and senior hiring assessments where selection decisions are contestable are the right contexts for AI proctoring. The $5/mo cost is calibrated to be accessible for these situations without pricing out everyday use.

The question to ask before enabling proctoring: if this result were challenged, would I need video evidence to resolve the dispute fairly? If yes, enable proctoring. If the result challenge can be addressed with score data and attempt metadata alone, the built-in controls are sufficient.

When proctoring is enabled on an exam, candidates are prompted to allow camera access at the start of their session. AI monitors the video feed throughout the attempt for behaviors associated with malpractice — looking away from screen for extended periods, presence of a second person in the frame, phone usage, or document reference.

Flags are logged against the candidate's session record with a timestamp. Recordings are stored securely for five days. After the exam, you review flagged sessions from the proctoring dashboard. The review interface makes it straightforward to jump to the flagged timestamp rather than watching the full recording.

Manual review is where the human judgment happens. The AI flags are probabilistic — they identify sessions that warrant closer attention, not confirmed malpractice. A reviewer watches the flagged clip and makes the determination. EasyEvaluate provides the evidence; the decision remains with you.

After a proctored exam session, the dashboard shows a list of flagged candidates ranked by flag count and severity. Click any candidate to see their session recording with flagged timestamps highlighted. Navigate directly to each flagged moment rather than reviewing full-length recordings for every candidate.

The manual review decision is binary: the flag is consistent with malpractice (escalate or invalidate the result) or the flag is a false positive (clear the flag, result stands). Document your review decision in the session record so there is an audit trail if the result is challenged later.

Five-day retention is designed to cover the post-exam investigation window. Most result disputes are raised within 24-48 hours of score release. Five days gives enough runway for the review to happen, the decision to be made, and the result to be confirmed or adjusted before the evidence window closes.

AI proctoring reliably detects: extended gaze away from the screen, a second person's presence in the frame, phone or document use that is visible to the camera, and long periods of inactivity. These are the highest-confidence behavioral indicators of malpractice in remote exam settings.

AI proctoring has meaningful limitations: it cannot detect a second screen outside the camera's field of view, a candidate receiving whispered assistance from someone off-frame, or a candidate using an AI text tool on the same screen in a browser tab. Browser-level controls address this last case; the others are harder to detect without additional physical controls.

Communicate the proctoring setup clearly to candidates: that their camera will be active during the exam, that the recording is stored for 5 days, and that flagged sessions will be manually reviewed. Transparency about proctoring is both ethical and practically effective as a deterrent.

• Reliably detected: looking away from screen, second person in frame, phone use, document reference
• Partially detected: unusual screen behavior, signs of distraction or stress (false positive risk higher)
• Not detected by camera: second screen off-frame, whispered assistance from off-camera, AI tool on a separate device
• Browser controls cover: tab switching, copy-paste, and in-browser AI tool use during the session
• Combined controls (browser + proctoring) cover a significantly wider range than either alone

Video proctoring involves collecting and temporarily storing personal biometric data — facial video. Candidates should be informed clearly before the exam begins: that proctoring is active, what is recorded, how long it is retained, who has access, and how it is used in result evaluation.

Consent to proctoring should be explicit, typically via a pre-exam acknowledgment step. Candidates who decline camera access should be unable to start the proctored session — configure the exam to require camera permission for entry rather than allowing unproctored attempts on the same link.

Five-day retention limits the privacy footprint. Recordings are not kept longer than necessary to resolve disputes that arise in the post-exam window. After 5 days, recordings are deleted automatically — reducing the long-term data risk associated with storing video of identifiable individuals.

AI proctoring works alongside — not instead of — EasyEvaluate's standard anti-cheating controls. Fullscreen, shuffle, tab blocking, and copy-paste restrictions remain active. Proctoring adds a video evidence layer in addition to the behavioral controls, not as a replacement for them.

The combination covers both behavioral signals (browser controls) and observational evidence (video recording). Together they provide a stronger integrity case than either approach alone — and in a dispute, having both types of evidence is more defensible than having either type alone.

For organizations running high-stakes exams, the recommended setup is: enable all browser-level controls as baseline, add proctoring for the exam sessions where video evidence is required, and maintain a clear manual review process for flagged sessions. This three-layer approach is more comprehensive and more operationally manageable than video-only proctoring without browser controls.