Tasks

Proctoring

TestInvite proctoring in one view: live human supervision during the exam, AI analysis after it, bulk triage across a whole batch, and a per-session deep review — what each tool does, when it matters, and how they work together.

Updated 2026/07/13

Proctoring in TestInvite has two halves. Capture happens during the exam and is configured in each step's anti-cheating settings: webcam photos or video, screenshots or screen recordings, lockdown, and AI detection settings. Review — what this chapter covers — is how you supervise and judge what was captured: three human surfaces and one automated analyst, each built for a different moment and scale.

The four tools at a glance

ToolWhenScopeWhat you do there
Live ProctoringDuring the examThe whole roomWatch live feeds, track progress and violations, chat, flag moments, end or reinstate sessions
AI Proctoring AnalysisAfter each session completesEvery enabled session, automaticallyNothing — the AI reviews the footage, pins findings to exact moments, and computes a traffic-light verdict
Post ProctoringAfter the examA whole batch of candidatesTriage: synchronized playback, wall review, bulk verdicts and finalization
Session ProctoringAfter the examOne candidateThe final word: verify each finding against its evidence, add manual flags, finalize the verdict

Live Proctoring — the human real-time layer

The Exam Monitor is where you supervise an exam as it happens: a live candidate roster with presence, position, and violation signals; a wall of live webcam and screen feeds; a real-time event stream; and in-exam chat. It is also the only surface where you can intervene — answer a stuck candidate, flag a suspicious moment the instant you see it, end a session, or reinstate one that broke down.

When it becomes very important: scheduled, high-stakes exam windows where a human must be able to act immediately — certification sittings, classroom exams, any exam where candidates may need live support. If nobody needs to act during the exam, you can skip live proctoring entirely and rely on the post-exam tools.

AI Proctoring Analysis — the automated first reviewer

When a session completes, the AI reviews its webcam material and flags what deserves a human look: a second person, a face that doesn't match the ID photo, sustained looking away, a phone or book in frame, talking, a muted microphone. Every finding is pinned to the exact photo or video second, and the session receives a traffic-light risk verdict. The AI never fails a candidate by itself — it collects evidence and leaves every decision to you.

When it becomes very important: any exam too large to watch by hand. The AI turns hours of footage per candidate into a short list of moments to verify — it is the multiplier that makes reviewing hundreds of sessions feasible. It can also be enabled retroactively on completed sessions when a result is disputed.

Post Proctoring — triage at batch scale

Post Proctoring loads a whole batch of candidates into one screen: play everyone's footage side by side on a single synchronized playhead, review and edit detections on a per-candidate wall, set verdicts in bulk, and finalize the clean majority in one action — opening the full session review only for the candidates that warrant it.

When it becomes very important: high-volume exams — hiring screens, admission tests, periodic compliance training — where opening sessions one by one would take days. Sort the batch by verdict, work the reds and yellows, bulk-finalize the greens.

Session Proctoring — the final word on one candidate

The deepest surface: one candidate's complete captured session, with every AI finding verifiable against its measured evidence (angle, confidence, duration, which run found it), a full approve/decline history with actor and timestamp, manual detections, and the explicit Finalize act that locks the verdict. This audit trail is what makes a proctoring decision defensible.

When it becomes very important: decisions that must hold up — failing a certification candidate, rejecting an applicant on integrity grounds, answering an appeal or a complaint. When someone asks “why was this session flagged, and who decided?”, this is where the answer lives.

How they work together

The four tools form one pipeline, and nothing is lost between stages:

  1. The step's anti-cheating settings capture the material during the exam.
  2. A live proctor (optional) watches in real time and flags moments — these manual detections survive the AI run untouched.
  3. The AI analyzes each completed session automatically and adds its findings and verdict.
  4. Post Proctoring triages the batch: false positives are declined (permanently — they never return on re-runs), clean candidates are bulk-finalized.
  5. Session Proctoring settles the contested cases: verify evidence, set the manual flag (it always wins over the AI verdict), and finalize.
Match the effort to the stakes. A low-stakes quiz may need no proctoring at all; a hiring screen does well with photo capture + AI + post-proctoring triage; a certification exam earns live proctors, video capture, full AI detections, and a finalized per-session verdict. Every layer is independent — add exactly the ones your exam needs.
Choosing your setup