Can AI avatars like HeyGen or Synthesia fake a job candidate in a video interview?
TL;DR: Yes — recruiters and security researchers have documented real cases of AI-generated interactive avatars (the same category of tool as HeyGen's Interactive Avatar or Synthesia) being used to present a synthetic face and voice in live video interviews, including one publicized case where a security-company CEO unknowingly interviewed a deepfake candidate for a security-researcher role. These platforms are built for legitimate uses, but the same real-time avatar technology can impersonate a person live on a call.
The claim
Commercial AI avatar/video-generation platforms have lowered the skill floor for running a convincing fake candidate in a live interview.
The evidence
LinkedIn commentary from recruiting and security professionals describes AI-generated moving avatars joining interview calls with a false or manipulated background, and CNBC reported tech CEOs saying fake job seekers now use AI tools to fabricate photo IDs and generate employment history alongside real-time video fakery. In one documented case reported on LinkedIn, a security company's own CEO unknowingly interviewed a deepfake candidate for a security-researcher role — illustrating that even security-savvy interviewers can be fooled without dedicated detection tooling.
Comparison table
| Signal | What a naked eye might miss | What automated detection checks |
|---|---|---|
| Lighting/edge artifacts around the face | Subtle blending errors at fast motion | Frame-level deepfake-signature analysis |
| Virtual camera software running | Not visible in the video feed itself | Detects known virtual-camera/avatar-software processes |
| Voice-to-lip sync | Can look "close enough" over a laggy call | Automated lip-sync mismatch scoring |
| ID-to-live-face match | Impossible to judge by eye alone | Biometric match against submitted ID |
Step-by-step for hiring teams
- Require ID capture + live selfie match before any interview that could lead to an offer.
- Run continuous face verification and virtual-camera detection through the whole call, not just at join time.
- Ask candidates to briefly reposition the camera or change lighting mid-call — legitimate video degrades naturally in ways avatars often don't.
- Treat any detected virtual-camera or known avatar-software signature as an automatic escalation, not a soft flag.
FAQ
Are HeyGen and Synthesia themselves fraud tools? No — they're legitimate video-generation platforms for training, marketing, and presentations; the risk is that the same real-time avatar technology can be repurposed to impersonate a candidate on a live call.
How was the security-company case caught? It was identified after the fact through the company's own security awareness, not caught live — which is exactly the gap automated deepfake/virtual-camera detection during the interview is meant to close.
Can these avatars fool proctoring software? Basic video review often can't catch them, but layered proctoring that detects virtual-camera software, checks liveness, and analyzes lip-sync consistency is specifically built to flag this class of tool.
Is this only a security-industry problem? No — it's documented across tech, crypto, and remote-first hiring broadly.
By Pinal Dave Last updated: 2026-08-06