Live Interviews Can Be Faked Too — Here's the Guard

[Meta excerpt: Learn how deepfake and impersonation detection can stop real‑time interview cheating before it slips through your hiring funnel.]

Tags: AI-Security

The Rise of Skillfishing

Recruiting teams have long worried about polished resumes, but the newest threat lives inside the video call itself. A recent audit of 19,368 live interviews found that 38.5 % of candidates displayed signs of cheating, with the figure jumping to 48 % for technical roles. What’s more, the rate climbed from a modest 9 % in July to a striking 45 % by September—showing that this isn’t an isolated trick but a spreading tactic.

The industry has even coined a name for it: **skillfishing**—the act of feeding a candidate answers in real time so they can pretend to possess expertise they don’t actually have. Tools such as Cluely, Interview Coder, and Final Round AI overlay invisible pop‑ups on the candidate’s screen, feeding them scripted replies while they maintain eye contact with the interviewer.

Why Traditional Interviews Fall Short

Standard interview scripts are practically a gift to these AI‑assisted cheaters. Questions like “Walk me through how you’d optimize this” or “Explain the difference between these two approaches” have predictable, lookup‑friendly answers. A candidate can simply read the supplied response, score high enough to pass (61 % of cheaters still clear the threshold), and walk away looking like a stellar hire.

Detecting a lie after the fact is too late; the damage is done once a false candidate moves forward in the pipeline. The solution isn’t to build a better lie detector—it’s to redesign the screening gate so that real‑time AI has nothing to feed on.

Adding Deepfake + Impersonation Detection to Your Gate

A robust defense starts at the very first video interaction. By integrating deepfake and impersonation checks directly into the screening stage, you can flag manipulated feeds before a human ever spends time evaluating them.

1. **Real‑time video authentication** – Use facial‑landmark analysis and micro‑expression tracking to spot signs of digital overlays, screen‑share injections, or deepfake avatars.

2. **Voice‑biometrics verification** – Compare the candidate’s speech patterns against enrollment samples to detect voice‑cloning or audio‑injection tools.

3. **Contextual anomaly scoring** – Combine video and voice flags with behavioral cues (e.g., unnatural eye‑gaze latency, delayed responses) to produce a risk score that triggers a manual review or a secondary challenge.

These layers don’t require new interviewing software; they work alongside your existing video‑conferencing platform, silently vetting each feed in the background. When a candidate attempts to