Home / Product
Detection layersSix signals. One honest score.
A single tell can fool you. A confident candidate who pauses is not cheating, and a nervous one who looks away is not hiding a screen. Trueyy reads several signals at once and weighs them together, so the noise drops out and the real pattern stands up.
signal layers
to surface a flag
platforms supported
When the answer is being typed somewhere else
The most common form of interview cheating right now is a hidden AI tool feeding answers in real time. Trueyy reads the rhythm of a conversation and notices the small gap between a question and a suddenly complete reply. When responses arrive cleaner and faster than any person could build them from scratch, the pattern gets logged. Cluely, ChatGPT, Claude, Gemini, InterviewCoder — they all leave a trace.
- Detects the prompt-and-paste delay signature
- Reads answer structure that is too clean to be live
Attention that leaves the room mid-answer
Candidates rarely cheat by staring at the camera. They glance at another window, a chat thread, or a doc just off to the side. Trueyy tracks when focus leaves the interview session and for how long, then ties each context switch to the exact moment in the conversation. Quick glances stay in context. Long lookups get flagged.
- Logs every focus change with a timestamp
- Separates a brief glance from a sustained lookup
Answers that appear faster than anyone can type
A 200-word block of text pasted in 0.3 seconds was not typed. Trueyy watches for paste events that exceed any plausible human input rate and marks them so you can review exactly what came in, how long it was, and when. Pre-written answers and AI-generated text both leave this signature.
- Catches bulk pastes that arrive during answers
- Distinguishes a quick paste from natural typing rhythm
Speaking freely versus reading from a script
People who know their material talk in a different rhythm than people reading off a screen. Eyes tracking a fixed invisible line look different from eyes that move freely around a face. Trueyy models gaze direction and pacing to tell the difference between a candidate thinking out loud and one reciting lines from a second monitor. No camera lockdown required.
- Reads gaze and pacing patterns without face tracking
- Flags steady off-screen reading, not honest thinking pauses
Help loves to hide just out of frame
A phone propped behind the laptop or an extra monitor just to the left is where most outside help actually lives. Trueyy surfaces the patterns that point to a second device in play: specific audio artifacts, ambient light changes, and response timing gaps. These surface as low-confidence signals in the feed, not as accusations.
- Surfaces likely second-screen and second-device behavior
- Ties the signal to the exact moment it appears in the call
How a real answer gets built
People reason out loud. They start, correct themselves, circle back, and occasionally contradict what they said a sentence ago. That is what thinking sounds like. Trueyy learns what that natural shape looks like, so answers that arrive perfectly structured and unnaturally fluent stand out instead of sailing past an interviewer who is busy listening.
- Models natural reasoning patterns across conversation
- Highlights answers that are too tidy to be built in the moment
See every layer working together
The signals are strong on their own. Together they give you a read you can trust. Book a demo and watch them line up on a real call.