AI interview cheating tools explained: Cluely, Interview Coder and the rest

AI interview cheating tools are apps that run on the candidate's computer or phone during a live interview: they listen to the interviewer, can read what is on the screen, and show suggested answers in a window that, their makers say, does not appear in screen shares or recordings. As of October 2026 the best-known examples include Interview Coder, Final Round AI's live interview assistant, LockedIn AI and Parakeet AI; Cluely, the name most people have heard, now markets itself as an "undetectable" AI assistant for meetings. They make polished, generic answers easy. What they cannot supply is the candidate's own memory of their own work, which is where a good interview should spend its time.

This page explains how these products work and what they claim. Detecting AI-written answers is covered in how to detect AI-generated interview answers, reading from a script in how to tell if a candidate is reading answers, and software that flags AI help in tools to detect AI use in interviews.

How do AI interview assistants work?

Most of them combine three pieces:

• Listening. The app captures the call audio, transcribes the interviewer's question and sends it to a language model, often with the candidate's resume and the job description already loaded.
• Reading the screen. For coding interviews, the app captures the problem shown on screen, such as a shared editor or an assessment platform, and returns code and an explanation.
• A hidden window. The answer appears in a small overlay the candidate can place near the camera, so their eyes stay roughly on the interviewer.

The hidden window relies on features operating systems already provide. On Windows, Microsoft documents a window setting under which "the window is displayed only on a monitor. Everywhere else, the window does not appear at all," intended for things like recording controls. Microsoft also notes it is not a security feature and does not stop someone photographing the screen. In practice it means a candidate can share their whole screen while a window you cannot see sits on top of it.

Some products go further. LockedIn AI describes a feature that lets a friend join the candidate's private session as a remote helper who is "never visible to the interviewer," which is a proxy helper delivered as software.

Which tools are candidates using, and what do they claim?

Everything in this table comes from each company's own site, checked in October 2026. The claims are theirs; we have not tested whether the tools are in fact undetectable.

Two details stand out. Cluely presents itself today as a meeting assistant rather than an interview tool, and Parakeet AI tells users the product is for conversations "where AI assistance is permitted" and to check the rules first. The technology is the same either way; whether using it is cheating depends on what the candidate was told.

How common is live AI help in interviews?

• In Greenhouse's 2025 AI in Hiring survey, 32% of US hiring managers said they had caught applicants reading from AI-generated scripts (November 2025, vendor survey of 665 US recruiters and hiring managers).
• In Greenhouse's July 2025 survey of 2,200 job seekers in the US, UK and Ireland, only 19% of US candidates said using AI in a live interview crosses the line.
• Of the 39% of candidates who used AI during the application process in Gartner's late-2024 survey, 29% used it to generate answers to assessment questions.

The second number matters most for interviewers: many candidates do not think live AI help is wrong unless someone tells them it is. More figures, each linked to its source, are in candidate fraud statistics.

Where do these tools help a candidate, and where do they fail?

The pattern is consistent: the more the interview depends on the candidate's own past and on back-and-forth with you, the less any assistant can do. That is also what a well-run structured interview rewards anyway. Questions built on that principle are collected in interview questions that expose AI-assisted answers.

What should you change about how you interview?

1. Say the rule before the interview. Whether live AI help is allowed should be in the invitation, in writing, for every candidate. A tested template is in candidates using AI in interviews: a fair policy.
2. Decide what you are testing. If the job uses AI every day, consider allowing it and assessing how the candidate uses it. Canva said in June 2025 that it now expects backend, machine learning and frontend candidates to use AI coding tools such as Cursor and Claude in technical interviews, and evaluates how they clarify requirements and catch problems in the AI's code.
3. Spend time on follow-ups, not openers. Score the second and third answer on a topic, not the first. Fluent openers are exactly what the tools produce.
4. Do not lean on screen sharing. As the Windows documentation above shows, a shared screen can hide a window, and a phone on the desk is not on screen at all. Asking for a camera tour of someone's room is intrusive and proves little.
5. Keep one stage the tools cannot reach, if the role justifies it: an in-person conversation, or a supervised session. The trade-offs are in should we bring back in-person interviews.
6. Treat software flags as prompts. Interview Signal, which we make, has a private answer check that flags answers that read as AI-written or read out and suggests a question to try; it never changes a score and never appears in a client submittal. Any flag, from any tool, is a reason to ask a better question.

What should you do if you think a candidate is using one?

The same thing you would do for any suspected outside help: do not accuse, ask follow-ups about their own experience, and score only what they showed. Neutral lines that keep the interview going:

• "Let's set the polished version aside. What actually happened on that project?"
• "Before you go on, who asked you to do that, and what did they say?"
• "Can you walk me through it the way you'd explain it to a new teammate?"

Write down observations, not conclusions ("could not say what went wrong after two follow-ups"), and add the same next step for every candidate at that stage if you remain unsure. If the issue is that the person on camera may not be the applicant at all, that is identity fraud, and the response is in what to do if a candidate used a proxy.

Is using an AI interview assistant illegal?

In most cases the consequences come from the employer's rules rather than from a specific law: a rejected application, or a withdrawn offer if the candidate broke a rule they had been told about. It becomes a different matter when the tool is used to pretend to be someone else, such as altering a face or voice to impersonate another person, which is identity fraud. Because the line depends on what you told the candidate, write the rule down and apply it to everyone. This is general information, not legal advice.