Cheat Detection is a Voice AI feature that flags AI interviews where a candidate may have used unauthorized assistance — like reading from a script or getting real-time help from an AI tool — while answering interview questions.
Cheat Detection is a supplemental signal, not a verdict. It does not affect the candidate's overall AI interview score. A flagged interview always requires a human to review the recording and transcript and make the final call.
Note: Voice AI can make errors. Always listen to the recording and read the transcript yourself before deciding whether a candidate appears to have cheated. You can adjust the cheat detection level to reflect your own assessment at any time.
Where to find Cheat Detection results
On the AI Interview candidate list
Candidates whose interview may show signs of cheating are marked with a warning icon next to their name in the AI Interview candidate list. Hovering over the icon shows a summary message letting you know the interview is worth a closer look.
On a candidate's AI Interview profile
Open a flagged candidate's AI Interview to see the full assessment. Select the Cheat Detection tab from the left-hand navigation to view the likelihood scale, the specific factors that contributed to the flag, and the option to overwrite the result.
To overwrite the result, select Overwrite and select the new segment you want to adjust the assessment to.
Understanding the likelihood scale
Each flagged interview gets a likelihood rating that reflects how confident the assessment is that the candidate used unauthorized assistance:
- Very Low
- Low
- Medium
- High
- Very High
This rating is informational only. It does not factor into the candidate's overall interview score or rating.
Reviewing detection factors
Below the likelihood scale, the detection factors list explains what specifically stood out about the interview. Factors can include things like:
- Background sounds, like typing, during a pause before the candidate answers
- An answer that's repeated nearly word-for-word later in the interview
- A delivery that sounds unusually even, fast, or polished for a spontaneous response, with none of the pauses or filler words typical of natural speech
Overwriting a Cheat Detection result
If you review the recording and transcript and disagree with the assessment, select Overwrite next to the likelihood scale to set the level yourself.
Note: Overwriting the Cheat Detection level reflects your own review of the interview. It does not change the candidate's overall AI interview score.
Turning on Cheat Detection
Cheat Detection is turned on in the Company Settings. In the left hand menu, select Interviews, then AI Interviews. Select the cog icon in the top right corner.
Scroll down to find Cheat Detection and turn it on or off by selecting the toggle.
Cheat Detection vs. Fraud Detection
Cheat Detection and Fraud Detection are related but separate signals within Voice AI:
- Cheat Detection looks at the interview itself — things like speech patterns and timing — to flag answers that may have been scripted or read with AI assistance.
- Fraud Detection looks at signals outside the interview content, like device and identity information, and resume authenticity, to flag signs of fraud.
Neither feature makes an automatic decision about a candidate — both are signals for a recruiter or hiring manager to review.