InterviewWatch includes a built-in applicant tracking system: post jobs and track candidates. Explore our ATS →
Contact usTry now
Detection reference

Every Detection Class The Integrity Agent Runs.

This is the reference layer under the interview integrity agent: one page per signal class, covering what it catches, how it decides, and where its limits are. AI tools, hidden overlays, remote-control software, clipboard activity, extra monitors, virtual devices and timing patterns. Nothing here records screen, audio or video.

The agent Live verification AI tool detection Integrity reports
68Integrity risk
AI assistantHigh
Extra monitorDetected
Clipboard burstFlagged
DecisionReview
How the platform works

From candidate consent to a verified interview.

The product runs one continuous flow: the candidate consents, the agent monitors machine-level integrity signals live, a risk score builds in real time, and a signed report lands after the session.

01 Consentcandidate opt-in 02 Live monitormetadata signals 03 Risk scorelive integrity 04 Reportsigned evidence Metadata only Tamper-evident
Product areas

Detailed pages by capability.

Each page explains a specific product capability, from live interview verification to AI-tool, overlay, remote-control, device, and evidence reporting.

Live interview verification

Monitor machine-level integrity signals during Zoom, Teams, Meet, or other live interviews.

AI tool detection

Detect ChatGPT-style tools, local LLM runners, coding assistants, AI browser activity, and stealth helpers.

Hidden overlay detection

Identify screen-share-excluded windows and overlay tools built to stay invisible on calls.

Remote-control detection

Flag AnyDesk, TeamViewer, RDP, and proxy-interview patterns that indicate outside control.

Virtual device detection

Detect virtual cameras, audio routers, voice manipulation tools, and mid-interview device changes.

Clipboard paste detection

Review large or structured paste events without reading clipboard content.

Typing pattern detection

Flag uniform typing, injected input, and pause-then-burst answer timing.

Multi-monitor detection

Record display baselines and detect monitors added or removed during the interview.

Knowledge-source detection

Surface search, docs, Stack Overflow, GitHub, and coding-site activity during live rounds.

Integrity reports

Deliver a tamper-evident report with scores, per-signal evidence, and a full incident timeline.

Consent & privacy by design

Structured metadata only, no screen, audio, video or keystroke content, and no auto-reject.

What InterviewWatch observes

  • AI tools, browser AI tabs, local LLM apps, and AI-assistant DNS traces.
  • Hidden overlays, screen-share-excluded windows, and suspicious window activity.
  • Remote-control tools, focus changes, clipboard bursts, and answer timing patterns.
  • Virtual camera/audio devices, monitor changes, camera changes, and audio-device changes.

What InterviewWatch avoids

  • No eye tracking or movement tracking.
  • No screen, audio, or video recording.
  • No keystroke content capture.
  • No automatic rejection without human review.
FAQ

Platform questions, answered.

What does InterviewWatch actually do?
It verifies live interviews the moment they happen: the candidate consents, the agent monitors machine-level integrity signals in real time, and a tamper-evident integrity report is signed after the session. It never records screen, audio or video.
Is InterviewWatch a proctoring or surveillance tool?
No. There is no eye tracking, movement tracking, or screen/audio/video recording. It observes structured integrity metadata only and never auto-rejects a candidate.
What can it detect during a live interview?
AI tools and local LLMs, hidden capture-excluded overlays, remote-control and proxy software, virtual cameras and audio routers, monitors added mid-interview, clipboard bursts, and pause-then-burst answer timing. See the cheating-tool guide.

One platform to verify every live interview.

Verify live, build an integrity risk score in real time, and hand reviewers signed evidence, without recording the candidate's screen.

Try nowRead the methodology