We build integrity tooling for hiring, which means we owe you honesty about what these signals are, how confident they are, and the decisions they should never make on their own.
HireBetter surfaces signals from an interview recording so a human reviewer can decide where to look. It does not determine whether someone cheated, and it should never be used to automatically reject, penalize, or disqualify a candidate. Every report is a starting point for a conversation, not a conclusion.
Our headline outcome has three states by design: Clear (no corroborated concerns), Review (evidence is inconclusive — explicitly routed to a person), and Flag (multiple independent signals corroborate a concern, still pending human confirmation). When the evidence is ambiguous we abstain to a human rather than guess.
From a single recording we look at several independent channels: speech transcription and speaker separation, on-camera face presence and gaze patterns, optional comparison against an enrolled reference photo (to catch impersonation), question-by-question answer consistency, and silence/timing patterns. We require corroboration across channels before raising a concern, so a single noisy signal does not produce a flag.
Each signal carries a confidence level and contributes to a calibrated risk estimate rather than a hard yes/no. Signals that fail to corroborate are presented as advisory and are kept out of the risk score. The goal is to make the evidence legible — you can see exactly why something was raised and judge it yourself.
Computer-vision models for faces and gaze are not equally accurate for everyone. Accuracy can vary with skin tone, lighting, camera quality, assistive devices, disability, neurodivergent behavior, and cultural differences in eye contact and expression. Industry face-detection systems have well-documented disparities across demographic groups, and we treat that as a real risk, not a footnote.
To mitigate it, we:
None of this removes the possibility of error. Treat a flag as “worth a closer look,” and give candidates a fair chance to explain before any decision is made.
We recommend you disclose to candidates that interviews may be analyzed for integrity, obtain consent where required by law, apply the tool consistently across all candidates for a role, and keep a human in the loop for every decision. Interview recordings can contain biometric and personal data subject to laws like GDPR and BIPA — handle them accordingly.
Recordings are processed to produce a report and supporting evidence. Raw uploaded video is retained only as long as needed to generate and review the report and is deleted on a rolling retention window (60 days by default); custom retention and data-region requirements are available on Enterprise plans. Biometric embeddings used for cross-session matching are stored separately from reports and scoped to your organization. Contact us to exercise data access or deletion requests.