Protect hiring integrity across screening and interviews. Fraud Detection Agent analyzes video, audio, and conversation for identity fraud and answers generated by AI. Recruiters get a real-time Identity Match Score and timestamped evidence to review, never an automatic decision.
Resume fabrication, AI-generated interview answers, and identity switching between screening and final rounds are hard to catch by ear. Background checks confirm a candidate's history. They do not confirm that the person answering interview questions is the person who applied, or that the answers are the candidate's own.
Fraud Detection Agent runs inside Screening and Interview Intelligence, analyzing face, voice, and conversation for identity mismatches and responses generated by AI. Every finding lands in the Integrity Check Panel with a real-time Identity Match Score, a severity rating, and timestamped evidence. Recruiters can jump straight to the flagged moment, not just read a bare warning.
Try itPoint solutions check a single moment, usually one interview or one background report. Fraud Detection Agent combines video, voice, and conversation into one identity picture, built into Screening and Interview Intelligence, not bolted on. And it tracks identity across every interview, comparing each round to the last, so a stand-in later on doesn't slip through.
Multi-modal detection Video, voice, and conversation analyzed together, not as separate checks
Explainable by design Every score comes timestamped, with recruiter override on every flag
Built into Screening and Interview Intelligence with no separate tool or workflow to manage
Built for recruiters and compliance teams, wherever candidate verification matters.
Fraud Detection Agent compares facial geometry and voice characteristics across interview rounds and generates a real-time Identity Match Score, accounting for normal variation in lighting, camera angle, and audio quality. Recruiters see a score and its context rather than a pass or fail, so they can weigh the evidence before deciding.
Face and voice compared across every interview round, not just the first
ID cross-checked against application details, when collected at intake
Real interview footage, timestamped, backs every identity flag
The agent reviews interview language for vocabulary diversity, fluency patterns, and natural disfluencies like filler words and self-corrections, and flags contradictions between what a candidate says in one round, in another round, and on their resume. A high signal means the interview did not produce enough verifiable substance. It does not mean the candidate is lying.
Flags unusually polished answers that follow a telling pause
Compares claims across rounds and against the resume for contradictions
Rates depth of detail, not just correctness, on open-ended answers
At application, the agent screens a candidate's profile against your Do Not Hire, No Poach, and No Rehire lists. It also compares the profile to your existing candidate records, catching a duplicate that reapplies under a different name or contact details. The check runs on exact identifiers plus configurable fuzzy matching, so a variant spelling doesn't slip through.
Checks new applicants against your own restricted-entity lists
Flags a duplicate profile registered under a different identity
Runs before a recruiter ever opens the application
From application to final interview, the agent works in the background.
The agent activates when a candidate submits an application and again at each screening or interview stage, drawing on whatever data is available at that point: resume, application fields, and recorded or live video and audio.
It checks face and voice consistency across rounds and analyzes conversational authenticity, then correlates findings across categories rather than judging each one alone.
Each signal lands in the Integrity Check Panel on the candidate's profile as Null, Low, Medium, or High, tied to the exact moment in the transcript or recording that produced it, so the recruiter reviewing it can see, not just read, what was flagged.
One serious flag, or a few smaller ones that line up across categories, can be routed to your compliance or audit team, following rules you set. Every time someone dismisses, acknowledges, or escalates a flag, it's recorded for the audit trail.
Application or interview. Runs automatically at application submission and interview stage.
Nothing about the hire. The agent assigns a severity tier to a signal. It never rejects, advances, or scores a candidate's fit for the role.
Every time, always. Every signal reaches a recruiter or a configured review team before any hiring action is taken.
Every action logged. Every signal, dismissal, acknowledgment, and escalation is timestamped and attributed for the audit trail.
Every signal can be dismissed or acknowledged by a recruiter or hiring manager, and no candidate is ever automatically rejected because of a fraud signal. Compound high-severity patterns can trigger a configured workflow action such as a pipeline hold, but a human still makes the final call before any hiring decision is finalized.
Fraud Detection Agent runs as a layer inside Phenom Interview Intelligence, drawing on the same candidate and interview data your team already works from, so integrity checks show up where recruiters are already looking instead of in a separate tool.
Fraud signals appear on the same interview record your team already reviews for scoring and feedback. Fraud Detection Agent runs as part of that Screening and Interview Intelligence workflow, not a separate one.
The agent's detection models run on the same X+ infrastructure that powers other Phenom agents, so signal calibration improves as more interviews are processed across the platform.
Every signal category can be turned on or off, consent language is configurable by region, and every recruiter action is logged for compliance and audit review.
85%
human-validated confidence rate across flagged cases
537
signals surfaced across identity and language
87%
facial-match confidence on known impersonations
Most of what the agent found, especially the high-severity cases, were definitely issues that would have prompted us to stop the process with that candidate. We found it reliable.
© 2026 Phenom People, Inc. All Rights Reserved.
