Remote assessment solved access and broke certainty. When a candidate takes a test from their own device, the honest question is not "did anyone cheat?" but "how much confidence should I place in this attempt?" Modern AI proctoring answers that question with evidence — and the best way to communicate it is a band, not a verdict.
What the proctoring layer actually captures
On AssessAll, a proctored attempt starts with an identity baseline: a reference photo captured before the first question. During the attempt, on-device face detection watches for absence and for additional faces; the platform monitors tab switches and blocks screenshots; resume attempts are limited and logged. None of this requires installing a lockdown browser — it runs in the web session.
Why bands beat verdicts
A binary "flagged / clean" forces a threshold decision at the worst possible layer — the machine. A High/Medium/Low integrity band, backed by the specific events that produced it, moves the judgement to the human who owns the decision. A Medium band on a low-stakes practice test can be ignored; the same band on a final-round hiring assessment prompts a review of the evidence, or a re-test.
AssessAll adds an AI review pass on top: the model reads the event log and snapshots and writes a short narrative of what happened — "candidate left frame twice, both under ten seconds, no second face detected" — so a recruiter reviews a paragraph, not a spreadsheet of timestamps.
Proportionate proctoring
Not every assessment deserves surveillance. Self-insight profiles need none; certification exams need a lot. Usage pricing helps here too: proctoring is a +5-credit add-on per attempt, so programmes turn it on exactly where the stakes justify it, instead of paying for it everywhere because the plan includes it.
The integrity questions to ask any platform
Does proctoring establish identity, or just presence? Are events stored with evidence you can audit later? Can you set the strictness per assessment rather than per contract? And does the output help a human decide, or just accuse? Get those four right and remote assessment is not a compromise — it's an upgrade.