Interviewing
Undetectable Is a Claim About Your Screen Share, and Little Else
Interview copilots are undetectable only in the narrow sense their vendors mean, and nothing independent has tested even that. Interview Coder tells buyers its overlay is completely hidden from screen share and that the interviewer sees nothing [1], and Final Round AI says its copilot does not appear in screen shares [2]. Both sentences are advertising copy about the seller's own product. The claim covers what renders on a shared screen. It says nothing about the conversation, where an answer still has to hold up, or about the employer's rule.
The takeThe technical half of the word is the least interesting half. What the claim smuggles in is that hiding is therefore safe, and the terms of service do not say it. ParakeetAI's terms forbid using the service to gain an unauthorized advantage in any interview where AI assistance is barred or has to be disclosed [4]. The stealth is what a buyer gets. The exposure stays with the buyer.
Where Olive fits
Open a role and see what the work shows
An Olive assessment is not an interview, and nothing in it is judged from how you sound: you work a real assignment with an AI assistant open, think aloud or type, and a person writes or edits each finding in words. You read the same report the employer reads, free.
Assess your shortlistWhat Does the Undetectable Claim Actually Cover?
One narrow thing: that the tool's window does not appear in a screen share or a screen recording. Interview Coder's coding-interview page says solutions arrive in a click-through overlay "completely hidden from screen share", followed by the line "You see the answer; your interviewer sees nothing" 1. Final Round AI publishes the same shape of claim about video calls 2.
Both are the seller's copy about the seller's product. No test result sits behind either page, and neither vendor says what was tried against it, by whom, or when.
Final Round's page also lists what it says it supports: HireVue, Zoom, Google Meet, Microsoft Teams, LeetCode, CodeSignal, HackerRank, CoderPad and Amazon Chime 2. That is a compatibility claim by a vendor, not an arrangement with any of those companies.
Whether the tool escapes notice on any of them is asserted, not shown.
A claim about pixels stays a claim about pixels.
The honest version of the pitch is smaller than the headline, and it is worth reading next to what an interview copilot costs you on the second question.
Read the Terms, Not Just the Front Page
One vendor's contract forbids the use its own marketing sells. ParakeetAI's terms, last updated March 2026, prohibit using the service to cheat on or gain an unauthorized advantage in any examination, assessment, interview or certification process where AI assistance is barred or must be disclosed 4. The front page of the same site marketed the product as completely undetectable when both pages were read on 1 September 2026.
Two documents, two readers.
Terms like that stop nobody. They decide who carries the consequence, and the answer is whoever clicked accept. The same front page told users to check the rules first, which is an instruction to do the one thing the product makes easy to skip.
The company that made the word famous has stepped back from it. TechCrunch reported in July 2025 that Cluely's Roy Lee called invisibility "a nifty add-on" rather than a core feature, and said most enterprises opt to disable it because of legal implications 3. The same piece reported a rival launch: Truely, from another Columbia student, marketed as an anti-Cluely that claims to detect unauthorized applications during online meetings 3.
Cluely's tagline had moved by then, from "cheat on everything" to a line about having everything you need before you ask, which still ended on the words "This feels like cheating" 3.
Both are founders describing their own products, and Truely's detection claim is untested. Lee's "most enterprises" carries no number. What the exchange shows is that the feature is negotiable to the people selling it.
Employers, meanwhile, are being told to put the rule in the invitation rather than police it afterwards, which is the substance of permitting or banning AI during the interview.
What Happens If Someone Finds Out?
Offers were withdrawn in the clearest documented case. Where the employer has said no, the use misstates who is answering, found out or not. CNBC reported in March 2025 that after Roy Lee posted a video of himself using Interview Coder to pass an Amazon interview, Amazon and the other companies that had made him offers rescinded them 8. That is one case, publicised on purpose by the person it happened to, and no company confirmed it on the record.
Amazon's reported rule covers the interview, not AI in general. Its recruiter guidelines, as reported, say candidates should not use generative AI during an interview unless explicitly permitted, and that failure to comply may result in disqualification 5. That wording is discretionary, and nothing reported says how often it has been applied.
Suspicion, meanwhile, is common and unverified. In the Institute of Student Employers' 2025 survey, using AI during interviews without disclosure or permission was the most commonly reported form of candidate misconduct, named by 61% of employers 9. That measures what UK early-careers employers say they suspected or identified, inside a self-selected membership body, with no verification step behind the number.
An employer believing it happened is not evidence that it did.
Disclosure beat exposure in an experiment that tested both. In a study published in 2025 with 195 US panel members, a professional was trusted least when someone else exposed his AI use, more when he disclosed it himself, and most when nothing was said 10. The scenario was a tax adviser and his client, so the ranking transfers and the magnitudes do not. The silent arm is a rater who never found out, which is luck rather than a plan, and where AI is barred, silence is the breach itself.
Which leaves a dull move with nothing to hide.
Run the copilot the night before, against the posting's own language, then arrive with two or three things you did and can be questioned about. A follow-up about something you said thirty seconds ago is the check an overlay is worst at.
Common questions
Can an interviewer see a copilot window on a screen share?
The vendors say no. Interview Coder describes a click-through overlay hidden from screen share, and Final Round AI says its copilot does not appear in screen shares. Both statements are marketing about the vendor's own product, with no published test behind them. Even taken at face value, they cover one thing: what renders in a shared window. And a shared screen is not the only thing an interviewer has to go on.
Do employers actually run software to check for AI use in interviews?
Some say they do. In Greenhouse's 2026 vendor survey of hiring managers in the UK, Ireland and Germany, 59% said they use software to detect or monitor AI use during interviews at least some of the time. That survey publishes no fielding dates and no sampling frame, covers no US employers, and says nothing about whether the software works. Reported use is not the same as accurate detection.
Is it against the rules if nobody said anything about AI?
It depends on whether a rule was stated somewhere you agreed to, so read the invitation and anything you acknowledged when you applied. Amazon's recruiter guidelines, as reported from internal documents, say candidates should not use generative AI during an interview unless explicitly permitted, and that failure to comply may result in disqualification. If nothing is stated anywhere, the question is a fair one to put to the recruiter before the round rather than after it.
What is the worst documented outcome for using one?
Withdrawn offers. CNBC reported in March 2025 that Amazon and other companies rescinded offers to Roy Lee after he posted a video of himself using Interview Coder to pass an Amazon interview. That is a single case, and he had publicised the use himself. None of the companies confirmed the rescissions on the record. It is the clearest documented consequence available, not a base rate for anyone else.
Is admitting AI use better than being caught?
An experiment that tested all three conditions points that way. With 195 US panel members, a professional was trusted least when a third party exposed his AI use, more when he disclosed it himself, and most when nothing was said at all. That silent arm is a rater who never found out, not a plan, and where AI is barred, silence is the breach. The scenario was a tax adviser and a client, so the ordering carries over, not the size of the gaps.
What should I do with a copilot subscription instead?
Point it at preparation, where no rule is in play. Run the job posting and your own examples through it the night before, and rehearse the two or three decisions you expect to be asked about. That version of the tool produces answers you can defend under a follow-up, which is the moment a fed line has least behind it.
References
- 1. Coding Interview: Screenshot to Solution in Seconds, Invisible to Screen Share interviewcoder.co The vendor's own wording for the undetectability claim this article examines: a click-through overlay hidden from screen share, with the interviewer seeing nothing.
- 2. Interview Copilot – Get Real-Time Interview Help with AI finalroundai.com Supports the second vendor claim about screen shares and video calls, and the list of platforms the vendor says it supports.
- 3. Why Cluely's Roy Lee isn't sweating cheating detectors techcrunch.com Supports the claim that the best-known vendor now calls invisibility an add-on that enterprises disable, and that a rival launched a detector.
- 4. Terms and Conditions parakeet-ai.com Supports the claim that a copilot vendor's contract prohibits undisclosed use in interviews and puts the rule-checking on the candidate.
- 5. Amazon banned candidates from using AI tools during interviews itpro.com Supports the reported interviewer tells (typing, reading, wandering eyes) and the reported disqualification wording.
- 6. Honest and Deceptive Impression Management in the Employment Interview: Can It Be Detected and How Does It Impact Evaluations? doi.org Supports the claim that interviewers detect impression management poorly, that experience does not help, and that perceived impression management still moves ratings.
- 7. The 2026 AI in Hiring Report: How jobseekers, recruiters and hiring managers across the UK, Ireland and Germany are navigating AI trust and transparency cdn.prod.website-files.com Supports the share of hiring managers reporting monitoring software during interviews, and the survey's stated composition.
- 8. How Google is responding to AI cheating in coder interviews cnbc.com Supports the documented rescission of offers after the use was publicised.
- 9. 5 trends you need to know from ISE's Recruitment Survey 2025 ise.org.uk Supports the share of employers reporting undisclosed AI use in interviews as the most common misconduct.
- 10. The transparency dilemma: How AI disclosure erodes trust (Study 13) oliverschilke.com Supports the ordering claim: exposure by a third party is trusted least, self-disclosure sits above it, silence highest.
10 sources, numbered by first appearance. How Olive sources claims
General guidance for hiring teams. What works at one company and one volume may not transfer to yours.
Olive assesses how a person works with AI. It does not detect AI-written documents, and it never produces a score, a ranking, or a match percentage for a person. Candidates read the same report the employer reads.