Cluely and Interview Coder: What They Are and How to Detect Them

A wave of purpose-built apps now promises to make candidates "undetectable" while an AI solves their coding interview in real time. If you hire remotely, you should understand exactly how tools like Cluely, Interview Coder, and LeetCode Wizard work — and what actually catches them.

What these tools actually do

Despite different branding, this category works the same way:

1. The tool captures the interview question — by screenshotting the screen or listening to audio.
2. It sends that to a large language model (GPT, Claude, or similar).
3. It displays the answer back in a floating window positioned over the interview — often near the webcam so the candidate appears to keep eye contact.

Cluely markets itself as a general "invisible" real-time assistant. Interview Coder and LeetCode Wizard are tuned specifically for coding rounds — screenshot the problem, return a working solution with explanation. The pitch is always the same: solve any interview without appearing to.

Why they claim to be "undetectable"

Their core trick is screen-capture exclusion. Modern operating systems let an application mark its own windows as excluded from screen recording and screenshots. These tools use that flag so their answer overlay is invisible to:

• The interviewer's screen share view
• Screen recordings
• Screenshots the candidate might otherwise be caught taking

Because the overlay produces no paste event, no browser tab switch, and no window that shows up in a normal screen share, basic monitoring misses it entirely. That's the whole value proposition.

How to detect them anyway

"Invisible to screen capture" is not the same as "undetectable." A layered approach surfaces these tools:

1. Require a full-screen share — and treat exclusion as a flag

When you require the candidate to share their entire screen, most overlays face a choice: render into the captured pixels (and become visible to you) or exclude themselves. The very act of a window being hidden from capture while everything else is visible is itself a strong signal worth investigating.

2. Use a native lockdown client for high-stakes rounds

A browser tab can't see other applications — but a native desktop client can. It can enumerate running processes and match them against a signature database of known cheating tools (Cluely, Interview Coder, and others), detect always-on-top or capture-excluded windows at the OS level, and flag virtual machines. This is the layer that directly detects the tool itself.

3. Watch the behavioral tells

Even with an invisible overlay, humans leave patterns: long silence followed by a sudden, fully-formed optimal answer; reading cadence in the eyes; answers that don't survive a follow-up. A live risk score that combines these makes the pattern visible.

4. Ask the candidate to change their solution

The most reliable tell of all: once code is on the screen, ask them to modify a requirement live ("now handle an empty input," "make this run in constant space"). A candidate reading from an overlay stalls; one who wrote it adapts instantly.

The honest limits

No detection is absolute, and we won't pretend otherwise. If a candidate runs the AI on a separate phone or laptop off to the side, no software on the interview machine — browser or native — can see it. The countermeasures there are a webcam gaze check (which flags repeated looking-away), requiring earphones so a nearby device can't hear the audio, a second "room camera," and adaptive questioning.

Related reading: How candidates cheat with AI in technical interviews · Is remote interview proctoring legal?

Tool names are referenced for identification and defensive purposes only. InterviewGuard provides consent-based integrity monitoring as a triage aid, always paired with human judgment.