Detection · Hidden AI assistants
Detect hidden AI assistants in remote interviews.
A subscription industry of AI tools now feeds candidates expert answers on an overlay only they can see — invisible to Zoom, Teams, and Meet. Capifiq surfaces that assistance as it happens, names it on the interview timeline, and seals it as evidence — even for a tool it has never seen before.
The threat
Answers, delivered on an invisible overlay.
Hidden AI assistants are desktop or web tools built for one purpose: to help a candidate cheat a live interview without anyone noticing. They listen to the interviewer through the machine's own audio, generate an expert-level answer in a second or two, and display it on a transparent overlay the candidate reads from — like a teleprompter only they can see.
It is a paid, fast-growing category. Many tools market themselves, in their own words, as "undetectable" on Zoom, Teams, and Meet. Each new entrant widens the surface a hiring team has to defend.
Tools in this category
- Cluely — invisible desktop overlay
- Interview Coder — coding-focused, claims zero detections
- Leetcode Wizard — OCR answer overlay
- Final Round AI — desktop "stealth" copilot
- LockedIn AI — desktop app plus browser extension
- …and the next tool shipped tomorrow
The blind spot
Why screen sharing and proctoring miss them.
These tools are engineered specifically to evade the defenses most teams already run.
Invisible to screen share
The overlay renders outside the surface the operating system shares, so it never enters the stream the interviewer sees.
Native, not in the browser
They are native desktop applications. Browser proctoring lives inside the tab and has no window into them.
Renamed and rebuilt
A denylist scanner only knows what it has catalogued. A renamed executable or a brand-new tool simply isn't on the list yet.
What Capifiq surfaces
- That an unauthorized assistant is active during the session
- The detected category and the event timestamp
- Coverage that isn't limited to a named list of products
- Cryptographically hashed, sealed evidence for later review
The coverage
Caught by what it does, not what it's called.
Capifiq verifies the interview environment rather than matching a catalog of product names. So a hidden AI assistant is surfaced by the fact that it is running and assisting — not by whether we have seen that exact product before.
Rename the executable, fork it on GitHub, or build a private one: the coverage holds. The finding lands on the interview timeline in real time and is sealed into the report as deterministic evidence — not a probability score.
The full picture
On-device is only half of it.
Capifiq detects any unauthorized assistance running on the candidate's device — hidden overlays are one of the common vectors, not a fixed list. And for help that never touches the device, interviewing methodology deters what no software can see.
Remote-control sessions
Another person viewing or driving the candidate's machine.
Virtual machines
A second environment concealing unauthorized tools.
Virtual cameras & deepfakes
A substituted or manipulated webcam feed.
Browser-based assistants
Extensions and web apps that generate answers.
Display tampering
Overlays and extra displays hiding content.
Off-device cheating
Deterred with interviewing methodology — the second pillar.
Questions
Common questions.
Does Capifiq detect tools like Cluely or Interview Coder?
How can a hidden AI assistant be invisible on a screen share?
Does Capifiq read what is on the candidate's screen?
What happens when a brand-new tool appears?
Get started
Catch the overlay your screen share can't.
Run Capifiq alongside the Zoom, Teams, or Meet calls you already do, and see what it surfaces.