Interviews run with the InterviewWatch agent get a signed integrity report on the candidate card. See how verification works →
Contact usTry now
ATS with integrity checks

ATS with interview integrity checks

Applicant tracking records that a round happened, who ran it and what they thought. It cannot tell you whether the answers were the candidate's own. InterviewWatch is an interview integrity platform with a pipeline inside it, so the evidence and the decision live on the same card.

The gapHow it closesWhat is detectedConsent and privacyVersus a bolt-onFAQ
K. Fontaine, OnsiteScore 97
S. Haddad, OnsiteScore 41
AI tool active in sessionCritical
Second display attachedReview
Signed reportOn card
Short answer

InterviewWatch is the applicant tracking system where interviews come back verified. Schedule a round from a candidate card and it runs with consent-first integrity monitoring. The score and the signed, timestamped report attach back to that same card, so the person deciding sees the evidence without opening another tool.

Nothing is recorded: no screen, audio, video, typed text or clipboard contents. Findings are evidence for a human reviewer, and no candidate is ever automatically rejected.

The gap

The gap in a standard hiring record

Sourcing, stages, scheduling and scorecards are all instrumented. The interview itself, the single event the whole hire rests on, is an unaudited conversation.

38.5%of candidates flagged for AI-assisted cheating across an analysis of 19,368 technical interviews, with the rate climbing from 9% to 45% in one quarter. Source
$15Mventure funding raised by Cluely, the overlay tool that began as Interview Coder. Undetectable interview assistance is a funded category, not a fringe hack. Source
0fields in a conventional ATS that record whether the answers in a remote round were the candidate's own work.

This is not a screen-share problem. An assistant can run on a second monitor, on a phone, or as an overlay excluded from capture, and the shared window stays clean throughout. What the machine knows and what the call shows have come apart. More on why screen share misses it.

How it closes

How verification attaches to the candidate card

The reason the evidence actually gets used is that it never leaves the object the recruiter is already working in.

1Candidate on the boardApplied from your career page, imported, or added by a recruiter
2Schedule from the cardTime, interviewer and meeting link. A join link is minted for the candidate
3Consent, then sessionThe candidate agrees, the agent runs for the interview only
4Signals correlatedIndependent signal classes combined into scored findings, live
5Score on the cardIntegrity score plus a signed, timestamped report, tied to the pipeline
6Human decidesMove the stage, or review the finding first. Nothing is automatic
Why the last step matters.

A finding is an observation, not proof of intent. Your team decides what an active AI tool means for your policy, and the report is there months later if that decision is ever questioned. See what a report contains.

What is detected

What the detectors look at

No single signal convicts. The value is in what fires together, and when, relative to the interview.

AI tool activity Hidden overlays Remote control Virtual devices Displays and focus Clipboard sources Correlated no single signal decides Signed report on the candidate card Reviewed by a human Nothing auto-rejected

Six independent signal classes, correlated into one finding a human reviews.

AI tool activity

Desktop AI apps, browser AI tabs, coding copilots, local model runners and browser-extension assistants active during the session.

Chat AICopilotsLocal LLMs

AI tool detection

Hidden overlays

Answer overlays designed to be invisible to screen capture, which is precisely the class of tool a screen share cannot reveal.

Overlay apps

Overlay detection

Remote control and proxy

Remote-desktop and screen-control software, the signature of somebody else driving the machine or sitting the interview.

Remote desktop

Proxy detection

Virtual devices

Virtual cameras and microphones, used to feed prepared or relayed media into a live call.

Virtual cam

Virtual device detection

Displays and focus

Monitor changes and window focus movement during the round, which is context for where attention was going.

Multi-monitor

Multi-monitor detection

Clipboard and knowledge sources

Clipboard event metadata (size and source, never contents) and knowledge-source activity during the session.

Paste events

Clipboard detection

Scoring, weighting and known limitations are documented in full on the methodology page, including the things this cannot see: a second device off-network is not directly observable, and signal availability differs by operating system.

Consent and privacy

How consent and privacy work

An ATS that quietly surveilled candidates would be a worse product and a legal problem. The boundaries are part of the design, not a policy page.

What never happens

  • Screen recording
  • Audio or video recording
  • Keystroke content capture
  • Clipboard contents captured
  • Eye or body tracking
  • Automatic rejection of a candidate

What does happen

  • The candidate consents before the session begins
  • The agent runs during the interview and then exits
  • Only disclosed technical metadata is observed
  • Findings are timestamped and signed
  • A named human reviews and decides
  • Candidates keep only their name and email in the pipeline

If you need the language for your process, we publish a policy template and a consent-language guide you can adapt.

Versus a bolt-on

Why a separate proctoring tool does not stick

Most teams that decide to verify interviews buy a second product. Six weeks later the reports exist and nobody reads them, for entirely mechanical reasons.

Step
ATS plus separate vendor
InterviewWatch
Scheduling
Schedule in the ATS, then set the session up again in the other tool, keyed by whatever identifier both happen to share.
One action from the candidate card. The interview is the card's interview.
Getting the result back
A human downloads a report and pastes a summary onto the candidate, or nobody does and the result stays in the other system.
The score and report attach to the card automatically, because the interview was created from it.
Cost
Two subscriptions, often two pricing models, frequently a per-candidate assessment fee on top.
One product. A free or flat-priced ATS plan, plus a $5-per-interview verification add-on with no separate contract.
Candidate experience
Two consent flows, two sets of instructions, two support paths when something does not launch.
One consent flow, one join link, one session that ends when the interview does.
Six months later
Someone asks why a hire went wrong and the evidence has to be reconstructed across two systems.
The card links to a signed, timestamped report that was never separated from it.

Already committed to a full ATS? That is a good reason to keep it. The built-in pipeline hides itself, interviews sync in from your system, and signed reports post back onto the candidate there. See how this compares to proctoring tools.

FAQ

Interview integrity: frequently asked questions

Is there an ATS with AI cheating detection built in?
Yes. InterviewWatch is an interview integrity platform with a built-in applicant tracking system, so pipeline and evidence are one product. A round scheduled from a candidate card runs with consent-first monitoring, and the score plus signed report attach back to that card.
How is this different from bolting proctoring onto my ATS?
A bolt-on means two vendors, two logins and a person copying results between them, which stops happening within weeks. Here the interview belongs to the pipeline card, so the result has nowhere else to land.
What does it detect?
AI application and window activity, hidden overlays, remote-control software, virtual cameras and microphones, display and monitor changes, window focus changes, clipboard event metadata, knowledge-source activity and agent heartbeat state, correlated into scored findings rather than isolated verdicts.
Does it record the screen or video?
No. No screen, audio, video, typed text or clipboard contents are recorded. Detection is metadata only, the candidate consents first, and the agent runs only for the duration of the interview.
Can a candidate refuse?
Yes, consent is real consent. How you handle a refusal is your policy decision, and the policy template covers how teams usually word it in advance so it is not decided in the moment.
What shows up on the candidate card?
The interview's status and integrity score, with the signed report behind it. It appears on the same board the recruiter uses to move stages, so the evidence and the decision are in one place.
Can we use one without the other?
Yes, in both directions. Verification is opt-in per interview, so scheduling without turning it on records the meeting with no candidate join link and no charge. Or keep your external ATS: the built-in board hides itself and verification runs on interviews synced from it.
Does this replace a coding assessment platform?
No, it answers a different question. An assessment platform judges the quality of an answer; InterviewWatch reports on whether the round was genuine. The comparison is here.

Put the evidence where the decision happens

Pipeline, scheduling and verified interviews in one product. Free with unlimited seats, and your first five interviews verified free.

Start free Read the methodology
Keep reading
ATS for technical hiringWhat to verify in an engineering round Live verificationHow signals correlate in real time Detecting AI assistanceThe signal-correlation approach What integrity monitoring isThe category, defined