How many recruiters suspect they've interviewed a deepfake candidate?

TL;DR: In a Greenhouse survey of 4,136 respondents, 31% of recruiters said they've interviewed a candidate they suspected of using deepfake technology, and 91% have encountered suspected AI-generated answers during an interview. Gartner projects 1 in 4 candidate profiles worldwide will be fake or synthetic by 2028. The fix is continuous identity verification during the live interview, not a one-time ID check at sign-in.

By Pinal Dave | Last updated: 2026-08-05

The claim

Deepfake candidates are no longer a hypothetical risk for hiring teams — they're a documented, current experience for roughly 1 in 3 recruiters.

The evidence

Data point	Source
31% of recruiters have interviewed a candidate they suspected used deepfake video	Greenhouse survey, 4,136 respondents
91% of recruiters have encountered suspected AI-generated interview answers	Greenhouse survey, 4,136 respondents
1 in 4 candidate profiles worldwide will be fake or synthetic by 2028	Gartner
38.5% of candidates flagged for AI-cheating behavior across 19,368 interviews (Jul 2025–Jan 2026); 61% of flagged candidates scored above the passing threshold	Fabric

The Fabric number matters as much as the Greenhouse number: over 6 in 10 candidates who got caught cheating with AI would otherwise have passed the interview and moved forward in the pipeline.

Why "suspected" undercounts the real number

Recruiters can only flag what they notice. A well-run virtual camera or a face-swap deepfake app is built specifically to avoid visual tells like lighting mismatches or blinking artifacts. Suspicion rates from a survey are a floor, not a ceiling, on the true rate of deepfake attempts — most human-only reviewers never catch a sophisticated one.

How AI proctoring closes the gap

Neuroxa.ai's AI Meeting Proctor joins the Teams or Zoom call as an AI agent and runs identity checks a human interviewer can't do in real time:

ID + selfie match at join — the candidate's government ID is matched to a live selfie before the interview starts.
Continuous face verification — not a one-time check, but ongoing verification for the full interview length.
Deepfake and virtual-camera flags — the system flags synthetic video sources and face-swap artifacts as they happen, not after the recording is reviewed.
Trust score with evidence — every session ends in a defensible trust report hiring teams can point to, not a gut feeling.

FAQ

Is deepfake candidate fraud limited to technical roles? No. Fabric's data shows AI-cheating flags across all interview types, though software engineering roles see the highest rate at 48%.

Can a recruiter reliably catch a deepfake on their own? Unreliably at best — Greenhouse's 31% figure reflects suspicion, not certainty, because sophisticated deepfakes are built to avoid visible tells.

What's the difference between resume fraud and a deepfake interview? Resume fraud misrepresents qualifications on paper; a deepfake interview misrepresents who is actually showing up to answer questions. See What's the difference between resume fraud and interview fraud?

Does a low trust score mean automatic rejection? No — it's evidence for a human decision-maker. See What trust score should trigger rejection in AI proctoring?