GoodFit

AI Proctoring

Catch cheating before it costs you a bad hire

Real-time tab, face, copy-paste, and AI-voice detection. Every flag timestamped on a scrubbable timeline - review in minutes.

Live · Rec
08:47
Tab switch detected
02:14
Copy-paste flagged
03:47
Single face · OK
now

Trusted by fast-growing companies

Customer logo
Customer logo
Customer logo
Customer logo
Customer logo
Customer logo
Astuto
The Sleep Company
Hudle
Customer logo
Customer logo
Customer logo
Customer logo
Customer logo
Customer logo
Customer logo
Customer logo
Customer logo
Customer logo
Customer logo
Customer logo
Customer logo
Customer logo
Customer logo
Customer logo
Customer logo
Astuto
The Sleep Company
Hudle
Customer logo
Customer logo
Customer logo
Customer logo
Customer logo
Customer logo
Customer logo
Customer logo
Customer logo
Customer logo
Customer logo

Live monitoring

Watches the whole screen, not just the webcam

Proctoring that stops at "is the webcam on" misses almost everything that actually matters. We track tab switches, fullscreen exits, copy-paste, print attempts, right-click menus, and keyboard shortcuts in real time. Flags fire the instant suspicious behavior happens, not when a reviewer gets around to it.

  • Tab and window focus tracking - catches the ChatGPT tab in a second
  • Copy-paste monitoring for both directions (into and out of the app)
  • browser tools and console detection - candidates who know the tricks get caught
  • Fullscreen exit detection flags candidates who leave the assessment window
Live · Rec
08:47
Tab switch detected
02:14
Copy-paste flagged
03:47
Single face · OK
now

Face detection

Catches impersonation and multiple people on camera

The AI runs continuous face detection through the interview. If the candidate disappears, a second face enters the frame, or the person on camera changes halfway through, the timeline gets flagged. Works without storing biometric data - flags are behavioral, not identity-based.

  • Missing-candidate flags when the face leaves the frame
  • Multiple-faces flags when a helper enters or is visible
  • Face-swap flags when the primary candidate appears to change
  • No biometric storage - detection runs in-session, flags are event-based

Candidate review · 10:00

4 flags
0:0010:00
Tab switch02:14
Copy-paste03:47
Lip-sync mismatch05:22
Head turned away07:55

AI voice detection

Catches ChatGPT reading the answers out loud

The biggest proctoring blind spot of 2026: a candidate has ChatGPT generate answers while they mouth along. We catch it with lip-sync analysis - if mouth movement doesn't match the voice being produced, the interview is flagged for review. Built for the era you're actually hiring in.

  • Checks that the person speaking matches the person on camera
  • Detects both pre-recorded responses and live AI-voice generation
  • Flags confidence level so reviewers can triage false positives
  • Works for voice and video interviews alike
Live · Rec
08:47
Tab switch detected
02:14
Copy-paste flagged
03:47
Single face · OK
now

Fast review

Every flag timestamped, review in minutes

A 30-minute interview with 4 flags takes ~3 minutes to review - not the full 30. Every flag is a direct jump-link into the recording. Reviewers see the context before and after, mark false positives, and move on. No scrubbing through hours of video.

  • Scrubbable timeline with flags at exact timestamps
  • One-click jump to the moment in the recording
  • Mark as false positive with reason - trains the per-org filter over time
  • Reviewer notes feed back into the hiring decision log

Candidate review · 10:00

4 flags
0:0010:00
Tab switch02:14
Copy-paste03:47
Lip-sync mismatch05:22
Head turned away07:55

Full coverage

12 violation types, nothing slips through

Every behavior that could indicate cheating is tracked: tab switches, fullscreen exits, copy attempts, paste attempts, print attempts, blocked keyboard shortcuts, right-click menus, developer console access, missing face, multiple people on camera, head turned away, and lip-sync mismatches. Each violation is logged with a timestamp and confidence level so your team reviews facts, not hunches.

  • Browser-side signals: tab switch, fullscreen exit, copy, paste, print, keyboard shortcuts, right-click menu, developer console
  • Camera-side signals: no face detected, multiple people, head turned away
  • Voice-side signal: lip-sync mismatch with confidence score
  • All signals combined into a single integrity score per candidate
Live · Rec
08:47
Tab switch detected
02:14
Copy-paste flagged
03:47
Single face · OK
now

Customer story · Xcelore

Xcelore relies on proctored engineering assessments to 5× their screening throughput without adding TA headcount.

Before GoodFit, the process was entirely manual. Recruiters reviewed every resume, then invited shortlisted candidates to full-day in-person rounds. There was no early filtering layer, a lot of time spent before anyone knew whether a candidate could actually code.
Sakshi Srivastava
Sakshi SrivastavaCampus Hiring Recruiter, Xcelore, Xcelore
Read the full story

6 weeks

Hiring time saved

2,600+

AI interviews conducted

What you get

12

violation types tracked

3 min

average flagged-interview review

Timestamped

every flag, every time

On by default

zero config to enable

FAQ

Questions hiring teams ask about AI Proctoring

Do you store candidate biometrics?
No. Face detection runs in-session to produce event flags (e.g., "face missing at 04:22"). We do not store face embeddings or identity vectors. Video recordings are retained per your retention policy.
What if a flag is a false positive?
Reviewers can mark any flag as a false positive with a reason. Over time, the per-organization filter learns which patterns in your setting are benign (e.g., a shared workspace camera catching coworkers). False positives do not fail a candidate unless a reviewer confirms.
Does it work for voice-only interviews (no camera)?
Yes. Voice-only interviews get the AI-voice detection, copy-paste monitoring, and tab-switch detection. Face-detection is skipped. Most of the cheating patterns we catch are browser-side, not face-side.
How accurate is the lip-sync detection?
Very accurate on high-confidence flags, which is what we surface. Low-confidence flags are logged but not raised to reviewers. The model improves per-org as you mark false positives.
Does proctoring slow down the candidate experience?
No. Candidates see no degraded performance.

Get started for free

Start hiring smarter today

Every account comes with 20 free credits. No credit card, no lock-in, no surprises.

Start free with 20 credits