Cluely and InterviewCoder: What Recruiters Need to Know
Your candidate is not alone in the interview. There is software in the room with them.
Two names keep coming up. Cluely markets itself as an "invisible AI" assistant that overlays answers on the candidate's screen during interviews, calls, and exams. InterviewCoder targets technical interviews specifically, feeding coding solutions in real time. Both promise the interviewer will never know.
They are not bluffing about the demand. Cluely raised venture funding on this pitch. The tools work well enough that recruiters now assume every remote interview is compromised until proven otherwise.
How the tools work
The core trick is simple. The cheating tool shows the candidate an overlay window with AI-generated answers. The overlay is designed to be invisible to standard screen sharing: the interviewer sees a clean screen while the candidate reads answers off a hidden layer.
InterviewCoder listens to the interview audio, transcribes the coding question, and displays a solution with explanation. The candidate reads, types, and sounds competent. The interviewer sees problem-solving. The problem was solved by someone else's model.
Newer tools add voice assistance: an AI whispers answers through an earbud, or a cloned voice takes the phone screen entirely. The candidate's job is to sit still and look thoughtful.
Why standard interviews cannot catch this
A video call shows you the candidate's face. It does not show you their screen. An overlay designed to hide from screen sharing will hide from your Zoom recording, your Teams transcript, and your interviewer's eyes.
Phone screens are worse. Audio-only gives the cheating tool everything it needs and the recruiter nothing to check against. Voice cloning has crossed the point where a recruiter's ear can tell the difference.
This is the structural problem. The interview was designed for a world where the candidate's only resource was their own brain. That world is gone.
What actually catches it
Screen capture at the OS level, not screen sharing. An overlay can hide from a shared window. It is much harder to hide from full screen capture running on the candidate's machine during the interview. Capture everything the candidate sees, not just what they choose to share.
Audio forensics. AI-generated and cloned voices leave artifacts: unnatural pitch stability, missing breath noise, spectral patterns human vocal cords do not make. Analysis that tests every audio frame catches what the ear misses.
Liveness and replay detection. A candidate playing a pre-recorded video through a virtual camera fails frame-level liveness checks. Interpolation patterns, lip-sync drift, and lighting inconsistencies show up under analysis even when they fool the eye.
Behavioral signals. Long pauses before fluent answers. Eyes locked on one screen region. Copy-paste bursts during "live" coding. Each is weak alone. Stacked with screen capture and audio analysis, they complete the picture.
What recruiters should do now
Assume compromise and verify. Tell candidates up front that the interview includes screen capture and fraud analysis. Disclosure alone deters casual cheating.
Structure the interview against assistance. Live follow-up questions that build on the candidate's own answer are hard to feed through an overlay. "Walk me through why you chose that approach" defeats a tool that only supplies the what.
Watch for the tells, but do not rely on them. A recruiter's gut is not evidence. Pair human judgment with tooling that produces a timestamped record.
Move the first screen earlier and make it proctored. A 5 to 15 minute AI interview with screen capture on day one filters compromised candidates before your team spends an hour with them. The expensive interview should be the verified one.
The honest framing
Cheating tools will keep improving. Detection will keep improving. This is an arms race, and the side with layered defenses wins each round.
The recruiters losing right now are the ones running 2022 interviews in 2026: video call, no capture, no analysis, gut feel. The fix is not suspicion. The fix is instrumentation.
The line: Cluely and InterviewCoder promise the interviewer will never know. They are right about standard video calls. Full screen capture, audio forensics, and liveness checks change the answer.