The Field Guide to AI Interview Cheating Tools (2026)

A reference for hiring teams. What's out there, how it reaches the candidate, and what it means for the integrity of your remote interviews.

Why we built this

If you run remote interviews, you're now hiring across a landscape that includes a large, fast-growing market of software built to feed candidates answers in real time. Most hiring teams know one or two names — usually Cluely — and assume that's the extent of it. It isn't. We've catalogued more than fifty distinct tools, commercial and open-source, and new ones appear roughly every week.

This is a reference for the defending side. It's organized not alphabetically but by the one thing that actually matters for integrity: where the answer is shown to the candidate, because that determines whether anything can catch it. We've deliberately left out pricing, setup steps, and "which is best" rankings — this is a guide to understanding the threat, not a guide to using it.

A note on completeness and accuracy: this market churns constantly. Tools get renamed, forked, abandoned, and relaunched. Treat this as a living snapshot as of early 2026, not a permanent registry. Where a tool's behavior is only partially documented publicly, we've said so rather than guessed.

How to read the landscape

Almost every one of these tools does the same four things under the hood: it captures the interview (the spoken question, the coding prompt on screen, sometimes the candidate's resume), sends that to an AI model, gets back an answer, and shows the answer to the candidate through a channel the interviewer can't see. The variation that matters is the last step — the delivery channel — because it splits the entire market into a few families with very different implications for detection:

• Invisible desktop overlays — the answer appears in a hidden window on the candidate's own screen.
• Browser-based helpers and extensions — the tool lives in a browser tab or add-on.
• Second-device and mobile tools — the answer appears on a separate phone, tablet, or laptop.
• Open-source clones — free, self-hosted versions of all of the above.

What this means for a hiring team, in one line: tools that run on the candidate's interview machine can be detected; tools that deliberately run on a separate device can't be detected by any on-machine software and have to be handled through how the interview itself is conducted. Keep that distinction in mind as you read — it's the difference between a software problem and a process problem.

Family 1 — Invisible desktop overlays

The largest and most mature category. A desktop application captures the screen or audio, sends context to an AI model, and renders the answer in a window engineered to be invisible to screen sharing — click-through, always-on-top, hidden from the taskbar or Dock, often with global hotkeys and a "panic" key. These run on the candidate's machine, which means they're detectable by software built for it — though the better ones disguise or rename themselves specifically to defeat detectors that look for known program names.

Commercial:

• Cluely — the category's flagship and lightning rod; listen-mode audio capture, real-time answers, invisibility toggle controlling screen-share and taskbar appearance.
• Interview Coder — the original; screenshot capture by hotkey, click-through overlay, hidden Dock and Activity Monitor, disguised process name, captures problems from CoderPad/HackerRank/LeetCode.
• LockedIn AI — system-audio capture, opacity-controlled overlay, process aliasing (e.g. presenting as "Ghost.exe"), OS-level hotkeys; also offers a remote human-helper "Duo" mode.
• Final Round AI — audio listening with resume/job context, desktop app marketed as invisible during screen share, plus a coding copilot.
• InterviewMan — markets 20+ stealth features, process-name masking, browser-detection prevention, invisibility to system/activity monitor, across desktop, mobile, and Chrome.
• ULTRACODE AI — emphasizes solving spoken and on-screen questions together; claims full-screen invisibility on Windows/Mac and across coding platforms.
• ShadeCoder — coding copilot with a screen-share-proof overlay and "undetectable" hotkeys; no tab-switching or mouse movement needed.
• ShadowCoder — invisible overlay with a screenshot-based "Code Mode" and a live-audio "Voice Mode" for system-design and behavioral rounds.
• Interview Browser — a "truly invisible AI browser"; a stealth shell with published OCR and visibility hotkeys, running as Hidden_Browser.app on macOS.
• Linkjob AI — markets 100% invisibility, never appears in the Dock, integrates with the major meeting and coding platforms, claims sub-second responses.
• Interview Solver — captures both sides of the interview audio plus the screen, undetectable global hotkeys, and a companion second-screen/remote-helper mode.
• StealthCoder — reads visible on-screen text and returns the solution in a stealth overlay.
• Interview Hammer — listens to Zoom/Meet/Teams and delivers instant answers; iPhone/iPad/Mac presence.
• Beyz AI — always-on-top invisible desktop app with a click-through hotkey; listens and pushes suggestions to the user's devices.
• PhantomCodeAI — native Mac/Windows assistant, multi-language, invisible to screen recording, with published hide/solve hotkeys.
• Hiding — hidden AI for screen-sharing with real-time transcription, screenshot solver, background audio recording, and claimed invisibility to Zoom/Teams/Meet/OBS.
• Control — desktop-native assistant for interviews and assessments that captures the question and overlays the answer, with a phone-remote workflow.
• CoPilot Interview — desktop app with a "Ghost Mode" opacity-adjusted window kept off the screen-share rectangle, plus local transcription.
• CodingCompanion — coding-only invisible assistant claiming invisibility to screen sharing or browser monitoring.
• CheatCoder — coding assistant with an invisible overlay and global shortcuts, covering coding challenges and multiple-choice questions.
• Stealth Interview — coding-focused desktop tool advertising one-click real-time problem solving.
• Offer Bull — downloadable desktop/mobile assistant with speech recognition and "100% private & undetectable" marketing.
• Verve AI — listens during interviews and analyzes the screen to solve problems; "Stealth Mode" claims visibility only to the user; covers meetings, coding, and assessment platforms.
• Ninjafy AI — described in secondary sources as a real-time tailored-response assistant; primary documentation is thin.

Family 2 — Browser-based helpers and extensions

These live in the browser — either as an extension or as a browser-native web app with nothing installed. They typically listen to the interview tab's audio or take active-tab screenshots. Because they operate through the browser, they tend to leave browser-level traces, though a fully browser-native helper with no extension is among the harder things for any approach to see.

• Sensei AI — a Chrome extension that listens directly to the interview tab's audio (and a browser-share workflow), with a visible "listening" indicator.
• Interviews Chat — a browser copilot with live transcription and a Chrome extension that does active-tab screenshot assistance for technical questions.
• Interview Copilot .io — browser-native, no install and no extension; listens in real time and generates answers in the browser itself.
• AIApply Interview Buddy — a live assistant that delivers answers through the user's earpiece or screen, with a one-click hide before screen sharing.
• tech-int-cheat — an open-source Chrome extension using the Google Meet closed-captions API plus an AI model.
• interviewcopilot (open-source) — a web app using cloud speech-to-text and AI models, with real-time transcription and a picture-in-picture mode.

Family 3 — Second-device and mobile tools

The hardest category for any defender to be honest about. These deliberately move the answer off the interview machine — onto a phone, tablet, second laptop, or an earpiece. Nothing runs on the monitored computer, which means no on-machine detection can see them. This is the category that has to be addressed through interview methodology rather than software, and any vendor claiming its agent "sees" a second phone is overselling.

• CodeRankGPT — explicitly sends the solution to a separate device of the candidate's choosing; avoids clipboard use, hotkeys, and tab-switching to leave no trace on the interview machine.
• OfferGoose — a "Formal Interview" mode generating tailored responses from the candidate's resume; usable on desktop or phone.
• JobJump — a mobile "invisible AI interview coach" for real-time help on a second device.
• InterviewFox — explicitly a "dual-device AI interview helper," built around a phone/computer split rather than an on-machine overlay.
• Leetcode Wizard — an invisible desktop app that also offers a separate web view to open on a phone/tablet/laptop, recommended for proctored interviews.
• ParakeetAI — a real-time copilot that supports simultaneous desktop and mobile sessions; its mobile version runs in a phone browser.
• Interview Sidekick — desktop, phone, or Chrome, with an optional stealth mode using OS-level window exclusion.

(Several tools above appear in more than one family because they ship multiple modes — a desktop overlay and a phone companion, for instance. The mode the candidate chooses determines how the threat has to be handled.)

Family 4 — Open-source clones and public forks

This is why the category can't be solved with a list of known programs. The same architecture has been replicated dozens of times in public repositories using freely available components, and anyone can download, rename, or modify them. They range from polished to proof-of-concept, and they appear and disappear constantly.

• Natively — native audio capture, local retrieval, custom keybindings, a "battle-tested" stealth mode, and a phone-link companion; supports local or cloud models.
• free-cluely — an Electron desktop clone with published hotkeys, screenshot capture, and local or cloud model options.
• Pluely — a Tauri/Rust always-on-top overlay with system-audio capture and a long list of supported AI providers.
• Open-Cluely / OpenCluely — Electron clones positioned as free alternatives to Cluely and Parakeet; some offer icon-masking (presenting as "Terminal," "Activity," or "Settings").
• Vysper — an Electron "invisible screen overlay" assistant relying on OCR and speech services.
• NexQ — a Tauri/Rust local copilot supporting local or cloud speech and language models.
• solveWatchAi — a transparent, well-documented repo featuring on-device transcription, screenshot analysis, and a content-protected overlay.
• Open Interview Coder — an Electron app with published screenshot and visibility hotkeys.
• Interview Coder Unlocked — an open-source "Invisible Edition" clone with stealth-run scripts.
• Aura-AI — a Python invisible-overlay assistant for interviews, meetings, and exams.
• MindWhisper AI — a Windows stealth-mode binary with real-time transcription and local-processing claims.
• cheap-cluely — a Python local-first clone with a translucent always-on-top overlay.
• ZNinja — an Electron "free alternative" claiming invisibility to screen capture and tab switching.
• OpenCluely (Vysper-inspired), Real-Time Interview Copilot, hack-interview — additional proof-of-concept and clone repos in active or dormant states.
• Cheating Daddy — an open-source click-through transparent overlay using a live AI model and system-audio capture.
• Ezzi — a self-hosted or managed coding overlay with click-through behavior and hotkeys.

What this landscape means for hiring teams

A few honest takeaways, because the point of a field guide is to leave you better equipped, not just alarmed.

It's bigger and faster-moving than the headlines suggest. Cluely gets the coverage, but it's one of more than fifty tools, and the open-source layer guarantees a constant supply of new variants. Any defense built around recognizing specific named tools is, by construction, always a step behind.

The delivery channel is the whole game. Tools that run on the candidate's machine — Families 1, 2, and 4 — are detectable by software built to catch the underlying technique rather than the program name. Tools that move the answer to a second device — Family 3 — are not visible to any on-machine software, and require a different control entirely: how the interview is conducted, so that an off-device assistant can't get the context it needs in time to help. A complete approach needs both. Be skeptical of any tool that claims a single piece of software catches all four families, including the phone on the desk.

"Undetectable" is a claim about one kind of detector. Nearly every tool here markets itself as undetectable, and against a detector that scans for known program names, that's largely true — they rename and disguise themselves precisely to beat it. It's far less true against detection that targets what the whole category has to do to function, regardless of what the program calls itself.

Renaming defeats the common approach — but not all approaches. The single most important thing to understand from this guide is that disguise (a tool presenting as "Activity Monitor," for instance) and the endless supply of renamed open-source forks make name-based detection unreliable. The detection that holds is the kind that doesn't depend on the tool's identity at all.

Capifiq targets what the category depends on to work, so however a tool hides, disguises, or renames itself, it's still flagged — and for the off-device family that no software can see, we pair detection with interviewer methodology that closes the gap. See where your process stands on five free interviews.

This guide is maintained as the landscape changes. Tool names, statuses, and capabilities reflect publicly available information as of early 2026 and are described for awareness purposes. Product names are trademarks of their respective owners; inclusion here is descriptive, not an endorsement.