How do you verify a job candidate's identity matches their ID during a video interview?

TL;DR: The proven approach, already used at day-one orientation by high-volume operators, is government ID capture plus a live selfie match, followed by continuous face verification for the rest of the call — not a single snapshot at login. Classet.ai's 2026 operator data shows this single control catches roughly 80% of identity-swap fraud, making it the highest-priority first investment.

Claim

A one-time ID check at signup misses the fraud that happens after login — continuous verification through the call is what actually closes the gap.

Evidence

  • Classet.ai (2026): "Government ID + liveness at orientation... For most high-volume operators, this single control catches roughly 80% of identity-swap fraud at day one. Start here before adding anything else."
  • Pindrop research documents a real deepfake candidate caught specifically because a live challenge (asking them to move their hand in front of their face) exposed rendering glitches that a static ID check would have missed entirely.
  • Prometric's remote bar-exam process requires a photo of a valid government-issued ID as a directed on-screen step — the same core mechanism scaled to one of the highest-stakes remote exams in the country.

Comparison: verification method vs what it catches

MethodCatches a mismatched IDCatches a deepfake mid-callCatches a proxy who passes the initial check
ID upload only, checked once at signupYes, at signup onlyNoNo
ID + selfie match at start of callYesNoNo
ID + selfie + continuous face verification through the callYesYesYes

Step-by-step: verify identity in a live video interview

  1. Require a government-issued photo ID capture before the interview begins.
  2. Run a live selfie match against that ID at call start.
  3. Keep continuous face verification running for the full interview, not just the opening minute.
  4. If anything looks off, ask a live physical challenge (move your hand in front of your face, turn to the side) — deepfake rendering struggles with unscripted movement.
  5. Log the verification result and any anomalies in the trust report for the hiring decision record.

FAQ

Isn't checking the ID once at the start enough? No — Classet.ai's data and real deepfake cases both show fraud that passes an initial check but fails continuous verification through the call.

What's a simple live test if something seems off? Ask the candidate to move their hand in front of their face or turn their head to the side — real-time deepfake rendering commonly glitches on unscripted movement.

Does this add friction for legitimate candidates? Minimal — it's a one-time photo capture plus a passive selfie match, not a repeated manual step during the interview.

By Pinal Dave Last updated: 2026-07-24