Hiring

How to Detect Cluely in an Interview

Cluely says it never shows up in shared screens. Here is how to detect Cluely by its effects, and how to run an interview it can't keep up with.

RE

CEO, Ruzora

6 min read

If you want to detect Cluely in an interview, start with what it claims. Cluely's homepage today calls itself the "#1 Undetectable AI for Meetings" and says it "never shows up in shared screens, recordings, or external meeting tools." So screen sharing won't show it. Asking the candidate to share their whole desktop won't show it. You have to catch its effects, not the app.

Here's what Cluely and its predecessor Interview Coder are, what the assessment vendors say they can and can't see, and the interview moves that make an invisible assistant much less useful.

Key Takeaways

  • Cluely grew out of Interview Coder, a tool built to cheat on software engineering interviews. Both market themselves as invisible to screen capture.
  • Screen sharing and recordings won't reveal an overlay app that hides from capture. HackerRank's desktop app mode is one of the few tools that says it detects and closes Cluely.
  • In a normal video interview, you detect Cluely by its pattern: reading eyes, delayed answers, polished first answers and weak follow-ups.
  • The strongest counter is an interview the assistant can't keep up with: interruptions, changed constraints and questions about the candidate's own past work.

What Cluely and Interview Coder are

Roy Lee built Interview Coder while a student at Columbia. NBC News reported in March 2025 that he recorded himself using it in an Amazon internship interview. Columbia suspended him; its stated reason was publishing documents from his disciplinary hearing. In April 2025 he raised a $5.3 million seed round for Cluely under the tagline "cheat on everything," followed by a $15 million Series A led by a16z in June.

Within days of launch, TechCrunch reported, Cluely dropped the references to cheating on exams and job interviews and leaned into sales calls and meetings. Interview Coder still exists separately and still says it's "100% Invisible to Screen-Recording", naming Zoom, Teams, Meet, HackerRank and CoderPad.

How it works in short: a transparent window sits over the candidate's screen, listens to the conversation or reads the problem, and shows suggested answers that the meeting software can't capture.

What platforms can and can't see

SettingCan it see an invisible overlay?Source
Zoom / Meet screen shareNo, per the tools' own claimsCluely and Interview Coder sites
Browser coding test (paste, tab, similarity checks)Not directly; may catch side effects like odd paste patternsHackerRank, CoderPad, CodeSignal docs
HackerRank Desktop App ModeSays it detects and closes "invisible cheating tools such as Cluely"HackerRank help center
Your own live follow-up questionsIndirectly, through the answersYou

I couldn't find CoderPad or CodeSignal documentation that names Cluely or Interview Coder. That doesn't mean they catch nothing. It means they don't claim it, and you shouldn't assume it.

A developer's screen with code, where an invisible overlay would sit on top
A developer's screen with code, where an invisible overlay would sit on top

How to detect Cluely in an interview without special software

Watch where the eyes go. Overlay text sits on the screen, often near the camera or the shared window. Reading looks different from thinking: steady horizontal movement, then a fluent answer.

Notice the delay pattern. The assistant needs the question first. Candidates using one often repeat the question aloud, pause for a few seconds, then answer smoothly. Thinking people start talking messily and correct themselves.

Interrupt. Cut in mid-answer with "wait, why not a hash map here?" The overlay is still showing the old answer.

Change the problem. "Same function, but the input no longer fits in memory." A good engineer adapts out loud. A reader stalls.

Ask about their own past. "What was the hardest bug in your last job, and what did you look at first?" An assistant can invent a story. The fourth follow-up about it is where invented stories fall apart.

Ask them to explain code they didn't write. Paste 40 lines into the chat and ask what it does, line by line, and where it would break.

None of these is proof. Together, over an hour, the pattern is usually clear. More moves are in how to prevent AI cheating in interviews.

A Concrete Version

A CTO at a 45-person SaaS company interviews a senior TypeScript candidate on Meet, 60 minutes.

  • Minutes 0 to 15: a standard debounce-and-throttle question. Perfect answer, textbook structure, about four seconds after each question. Eyes move left to right while answering.
  • Minutes 15 to 30: the CTO interrupts: "Skip the code. Tell me about the last time debouncing caused a bug for you in production." Long pause. A generic story.
  • Minutes 30 to 45: a constraint change on the first problem: "Now it runs in a worker and can't touch the DOM." The candidate restates the original answer.
  • Minutes 45 to 60: the CTO pastes a 40-line function from their own codebase. The candidate reads it aloud but can't say what the retry logic is for.

Decision: no. The CTO didn't need to prove Cluely was running. The interview measured what it was supposed to measure, and the candidate didn't have it.

The Honest Counterpoint

You will sometimes be wrong. Some honest candidates repeat questions to buy time, look away when they think, or answer textbook questions perfectly because they studied. Non-native English speakers pause more. Judge the whole hour, not one moment, and never accuse a candidate of cheating based on eye movement alone.

And the tools will get better at hiding. Lee told TechCrunch he could move the assistant into hardware like smart glasses. Detection of the app is a losing race. Designing interviews that don't reward a reading assistant is not.

Where Ruzora fits

At Ruzora, every engineer on our bench has passed a graded coding challenge and an AI interview answered on camera, as separate evaluations. And you still run your own interview, after the agreement, with the engineer's full identity.

Frequently Asked Questions

Can you detect Cluely in an interview through screen sharing?

No. Cluely says it never shows up in shared screens or recordings, and Interview Coder makes the same claim. You detect it through the candidate's behavior and answers, or through lockdown software like HackerRank's desktop mode that says it closes such apps.

Does HackerRank detect Cluely?

HackerRank's help center says its Desktop App Mode detects and closes invisible cheating tools "such as Cluely and Ultracode." That applies to its desktop app for AI interviews, not every HackerRank test.

Can CoderPad detect Interview Coder?

CoderPad's public fraud page lists paste alerts, IDE-exit tracking, playback, similarity and optional webcam proctoring. It doesn't mention Interview Coder or overlay apps, so don't count on it. See our overview of interview cheating detection tools.

The Bottom Line

You probably can't see Cluely. You can make it useless: interrupt, change constraints, and ask about the candidate's own work until the answers have to come from them. If you'd rather start with engineers who already passed two separate evaluations on camera, see our bench or request a shortlist.

Roberto Espinoza is CEO of Ruzora, which helps US startups hire pre-vetted senior LATAM engineers, with a vetted shortlist in 72 hours. See available engineers.

RE

CEO, Ruzora

Roberto is the founder and CEO of Ruzora. He works directly with US startup founders and CTOs on staff-augmentation and software-factory engagements, and personally reviews senior engineer placements.

Senior LATAM engineers

Get a vetted shortlist within 72 hours.

Tell us the role and the stack. Within 72 hours you're reviewing a vetted shortlist of senior LATAM engineers who match it.