Fraud Detection Compared: RightMatch vs HireVue vs Pillar vs Harver
Everyone is cheating now. The question is which interview platform can prove it.
Fraud detection is becoming a standard checkbox in AI interview software. But "cheating detection" means different things on different platforms. Some analyze the video call. Some capture the screen. Some publish their accuracy. Most do not.
Here is how the four leading platforms compare on the capabilities that matter, based on public documentation.
The comparison
| Capability | RightMatch | HireVue | Pillar | Harver |
|---|---|---|---|---|
| Full screen capture during interview | Yes, continuous | Partial: employer-enabled screen sharing, scoped to coding assessments, not continuous in standard video interviews | Not publicly documented | Not publicly documented (periodic photo capture, not screen capture) |
| AI-generated voice detection | Yes, per-frame audio analysis | Not publicly documented (no dedicated AI-text detector) | Not publicly documented | Not publicly documented |
| Camera replay and liveness checks | Yes, every video frame | Periodic webcam snapshots for identity consistency | Not publicly documented | Periodic photo capture during assessments |
| Resume duplication across applicants | Yes, hash-based, cross-customer | Not publicly documented | Not publicly documented | Not publicly documented |
| IP / geo / VPN flagging | Yes, optional auto-block | Yes: IP address, tab-switch records, and login counts shown to reviewers | Not publicly documented | Focus-change detection (tab/app switching); IP flagging not publicly documented |
| Composite risk score with threshold | Yes, 75% default, configurable per role | Flags plus human review; no published composite score | Not publicly documented | Behaviors reported on the Candidate Detail Page; no published composite score |
| Published false positive rate | Yes, under 0.4%, quarterly | Not publicly documented | Not publicly documented | Not publicly documented |
| Timestamped audit trail per flag | Yes, every signal and action logged | Yes: event logs visible to reviewers | Not publicly documented | Yes: detected behaviors reported on the Candidate Detail Page |
| SOC 2 Type II | Yes | Yes | Not publicly confirmed | Yes |
What the table tells you
RightMatch is the only platform in this set that publishes its false positive rate. That matters more than any feature row. A detection system without a published error rate is an accusation machine with no accountability.
Screen capture is the dividing line. Most cheating tools, from Cluely to InterviewCoder, work by showing the candidate something the interviewer cannot see. Platforms that only analyze the video call are blind to the overlay. Full screen capture during the interview is what makes the invisible visible.
Resume duplication across applicants is rare. Most platforms check the candidate in front of them. Checking whether the same resume is working five identities across your pipeline, and across other customers, catches organized fraud that single-candidate analysis misses.
What to ask in a demo
Do not ask "do you detect cheating." Everyone says yes. Ask these instead:
- Do you capture the candidate's full screen during the interview, or only the video call?
- What is your published false positive rate?
- Show me the evidence trail for one flag. What does the recruiter actually see?
- Can a single signal block a candidate, or does it require corroboration?
- Is every signal, score change, and action logged with timestamps?
The vendor that answers all five with specifics is the vendor to shortlist.
The honest summary
HireVue has the longest track record in video interviewing and markets interview integrity as part of its enterprise platform. Evaluate it on the five questions above like anyone else.
Pillar and Harver are strong in their lanes, AI interviewing and assessment respectively. Their public documentation on fraud detection specifics is thinner. That is not a verdict. It is a reason to ask harder questions in the demo.
RightMatch built fraud detection as a core layer, not a feature: continuous screen capture, per-frame voice analysis, cross-applicant resume matching, a composite risk score with a published under-0.4% false positive rate, and a timestamped audit trail on every flag.
The line: Every vendor says it detects cheating. Only one in this comparison publishes its false positive rate. Ask for the number before you believe the claim.