What is interview cheating detection?
Interview cheating detection spots outside help during an interview, such as answers read from ChatGPT or a second screen. InterviewFlowAI's Trust Score checks for tab switching, signs of LLM or Google use, and eye and face behavior, and scores each interview out of 100. Below 70 shows Likely cheated and 70 to 84 shows Worth a look. It is included at no extra cost.
Trust Score is InterviewFlowAI's built-in interview integrity layer, included at no extra cost. During AI interviews, the platform monitors for behavior signals that suggest the candidate may not be responding authentically: tab switching, possible LLM usage or Google search, and eye and face direction during video interviews, and you can block extra screens and monitors. These signals are aggregated into a Trust Score out of 100 that sits alongside the candidate's competency scorecard — not as a pass/fail judgment, but as context that helps recruiters decide whether an interview warrants closer review.

Set expectations before the interview starts.
Confirm camera, microphone, and interview readiness before a candidate begins responding.
- Camera and microphone ready
- Eye and face direction monitored
- Tab switching and external help checks explained
Investigate context instead of auto-rejecting.
Review tab switching, possible LLM usage or Google search, and eye and face direction alongside the interview evidence.
- Tab switching timeline
- Possible LLM usage or Google search
- Eye and face direction signals

Challenge second-screen AI interview copilot scripts
A polished answer can hide a real-time AI interview copilot on a second screen. For technical hiring managers, the question is whether candidates can explain their own work and apply their knowledge when the conversation changes. InterviewFlowAI combines three ways to challenge live script-reading and give reviewers evidence to assess.
Resume-specific follow-up probing
Adaptive follow-ups probe the candidate's actual past projects, using their resume and answers to ask about personal contributions, technical trade-offs, failures, and measurable results. For example: on the network migration you listed, what failed during cutover, what did you change yourself, and how did you verify recovery? Each answer shapes the next probe, making generic copilot scripts harder to rely on.
Live scenario AI Role Plays
Pair the interview with an AI Role Play that asks candidates to work through a realistic scenario in real time. Introduce a changing constraint or a new symptom and ask them to explain their next step. Review how they apply their knowledge as the situation develops, beyond a polished opening answer.
Question-level integrity flags
Trust Score, included at no extra cost, scores each interview out of 100 from tab switching, answer authenticity, and eye and face direction, and the candidate report shows which questions were flagged. You can also stop candidates from using extra screens. Human reviewers check those moments against the transcript, recording, and competency scores to assess possible external assistance in context.
Zero automatic rejections from integrity signals
An integrity flag never automatically rejects a candidate. A second-screen copilot may not trigger a tab switch, and a pause or inconsistent answer alone does not prove script-reading. A human reviews the evidence, considers innocent explanations and candidate accommodations, and records the reason for any hiring decision. This supports a documented, accountable review process; it does not guarantee legal compliance.
Remote hiring needs integrity signals, not just competency scores
Remote interviewing solved the scheduling problem but created a new one: how do you know the person answering your questions is actually the person you're hiring — and that they're answering without external help? In live interviews, eye contact, hesitation patterns, and environmental cues give interviewers some signal. In asynchronous AI interviews, those signals disappear unless the platform actively monitors for them.
Trust Score fills that gap. It's not a lie detector — it's a behavior monitoring layer that surfaces signals your team can use to decide which interviews need a closer look. A candidate who stays in the interview tab and keeps their attention on the screen? High trust. A candidate who tabs out before every technical answer and keeps looking away? That interview gets flagged for recruiter review.
What Trust Score monitors and what it means
LLM usage or Google search
Signals of possible LLM usage or Google search for answers are flagged for recruiter review alongside the interview evidence.
Eye and face direction
During video interviews, eye and face direction signals help reviewers identify moments that may need a closer look. Looking away alone does not establish cheating.
Tab switching and window focus tracking
The platform detects when a candidate leaves the interview window — to search for answers, reference notes, or receive external help — and how frequently it happens.
Extra screens and monitors
Turn off multiple or extended screens to stop candidates from running the interview with a second monitor attached.
Configurable checks and flags
Choose which checks to run and how many tab switches to allow. Interviews that score below 70 get a Likely Cheated badge, scores from 70 to 84 get Worth a look, and the candidate report shows which questions were flagged.
How to use Trust Score without over-relying on it
Trust Score is a decision-support tool, not an automated rejection system. The strongest teams use it as one data point among many — flagging low-trust interviews for human review rather than auto-rejecting candidates based on a score alone.
- Use Trust Score as a review prioritization signal — review flagged interviews more carefully, not automatically reject them.
- Consider context: a candidate interviewing from a noisy coffee shop may trigger different patterns than one in a quiet home office.
- Combine Trust Score with competency scores: a candidate with strong answers AND high trust is a stronger signal than either alone.
- Look for patterns across your candidate pool — if many candidates trigger the same flag, your question format may need adjustment.
- Use Trust Score data to improve your interview design — questions that consistently trigger flags may need to be reworked for originality.
Trust Score vs. traditional proctoring solutions
| Approach | Traditional proctoring | InterviewFlowAI Trust Score |
|---|---|---|
| Cost model | Per-exam or per-candidate fees. Expensive at scale. | Included at no extra cost. |
| Candidate experience | Intrusive. Webcam monitoring, locked browsers, room scans. Feels accusatory. | Background monitoring. Candidates focus on the interview. Integrity signals surface in review. |
| Review workflow | Proctors review flagged sessions. Added time and cost. | Trust signals appear alongside competency scores. Recruiters review in the same dashboard. |
| Decision role | Often used as a pass/fail gate. Candidates can be auto-rejected. | Decision-support tool. Highlights interviews for human review. Humans make the call. |
Recommended Trust Score workflow
Enable Trust Score for your interview workflow
Open the Interviewer's Proctoring tab, turn on proctoring, choose your checks, and set how many tab switches to allow based on your role's integrity requirements.
Review trust signals alongside competency scores
Look for the Likely Cheated badge on the Applicants tab, then open the Proctoring tab in the candidate report to see the score and which questions were flagged.
Investigate flagged interviews, not auto-reject
Watch the recording, read the transcript, and check the flagged questions. Make a human judgment about whether the interview reflects genuine candidate ability.
Use trust patterns to improve your process
If certain questions consistently trigger flags, redesign them. If certain candidate sources produce lower trust scores, adjust sourcing or screening instructions.
Guides on interview integrity
Make your next hiring flow faster.
Start a 7-day trial with 10 interview credits on Pilot, Growth or Scale, or book a demo to see it on one of your roles.
