Online Proctoring

The three types of proctoring

• Live proctoring: a person watches the candidate by webcam in real time. Hardest to beat, but expensive and needs scheduling, which kills completion rates in high-volume funnels.
• Record and review: the session is recorded and a reviewer watches flagged parts later. Cheaper, but the decision comes after the fact.
• AI proctoring: software watches the session automatically and produces flags and an integrity score. It scales to thousands of candidates a day and is now the default in remote hiring. A human should still review anything that leads to rejection.

Common signals checked: face match against an ID or earlier session, more than one face or no face on camera, voices in the background, browser tab or window switches, pasting text, unusual typing rhythm, screen sharing or virtual machine use, and answers that match known leaked content.

Why it matters in remote hiring

When assessments moved online, so did cheating. Recruiters in contact center and tech hiring now see proxy test-takers (someone else sits the test), coached candidates reading answers off a second screen, and people who pass a remote pre-employment assessment but cannot hold a basic conversation on day one. Generative AI made it worse, since a candidate can paste a written question into a chatbot in seconds.

Proctoring is one control against candidate fraud. It works best alongside identity checks at the offer stage and a short live conversation before hire. Many teams use AI proctoring across the whole remote funnel instead of only on the final test, because fraud tends to start at the first step.

A worked example

A remote customer support program hires 120 work-at-home agents a quarter. In one quarter, 9 new hires turn out to be different people from those who passed the assessment and video interview. Each failed hire costs roughly $4,000 in recruiting, equipment and training time.

The team adds AI proctoring to the assessment and asynchronous interview, with face match between the two and the onboarding ID check. Flags trigger a human review, not an automatic rejection. In the next quarter about 6% of sessions are flagged, two thirds are cleared on review (poor lighting, a family member walking past), and the rest are confirmed. Impostor hires drop to one. The cost is a few minutes of reviewer time per flagged session.

Privacy, consent and fairness

Proctoring collects video, audio and sometimes biometric data, so the legal side matters:

• Tell candidates before they start what is recorded, what is analyzed and how long it is kept.
• In Illinois, the Biometric Information Privacy Act (BIPA) requires informed written consent before collecting biometric identifiers such as face geometry, plus a published retention policy. Other states and cities have their own biometric and AI rules. See BIPA.
• Under GDPR, proctoring data is personal data and face matching can be special category data, so you need a lawful basis, a clear notice and a retention limit.
• Offer accommodations. Candidates with some disabilities may trigger flags (eye movement, needing a screen reader, a carer in the room), so the ADA's reasonable accommodation duty applies.

Rules vary by state and country, and this is a fast-moving area.

Common mistakes

• Auto-rejecting on a flag. Flags are evidence, not a verdict. Webcam glare, shared housing and slow connections all create false positives.
• Proctoring too heavily too early. Requiring a 360-degree room scan for a 10-minute screening test drives candidates away. Match the strength of control to the stakes.
• No retention limit. Recordings kept forever add legal risk with no benefit.
• Ignoring the rest of the process. If identity is never checked again after the test, a proxy can pass a proctored test and still not be the person who shows up.