A Candidate with a Phone and an LLM Used to Pass. Now They Don't.
Voice-based corporate exams in 2026 face a new attack: a candidate with a smartphone and a free AI chatbot can fake competence in real time. They speak the question into the chatbot, read the answer back to the examiner, and pass. iTutor's L&D Competency Exam layers several checks to make that attack expensive, and flags what it finds for a human reviewer.
How do you stop candidates from using ChatGPT during an oral exam?
Anti-AI proctoring is a set of monitoring techniques that detect when a candidate is relaying exam questions to an outside AI in real time and reading back its answers. Standard webcam proctoring (face detection, gaze tracking) is necessary but not enough: a candidate can read from a phone or a second screen while looking at the camera. Effective anti-AI proctoring adds checks that only a present human can pass, traps that a relayed AI falls into, and voiceprint and device checks on top of the camera.
iTutor (itutor.study) runs layered anti-AI proctoring on its voice-based L&D oral exams: codes shown only on the candidate's screen that must be entered, a hidden instruction that an AI relaying the questions tends to obey, camera checks for phones and extra screens, face signals for looking away or leaving the camera, voiceprint comparison with the start of the session, and browser checks for leaving full screen, copy and paste, developer tools and a second display. A result that looks AI-assisted is held for human review. The on-screen code checks are on L&D Growth ($22 per seat a month) and L&D Enterprise ($27 per seat a month); the other checks run on every L&D plan.
The new attack on corporate certification
For a decade, the assumption behind a corporate oral exam was simple: if the candidate is on camera answering in their own voice, they're being assessed honestly. That assumption broke in 2024 when consumer LLMs got good enough to give passable spoken responses to almost any professional knowledge question in real time. By 2025, an undergraduate could rig a smartphone, a Bluetooth earpiece, and ChatGPT Voice to pass a structured behavioural interview without breaking a sweat.
For L&D programmes that issue certifications to regulators, hiring managers, or external accreditors, this is an existential problem. A certificate that any motivated candidate can game is worth less than no certificate at all. The corporate L&D platforms most teams use today (Cornerstone, Docebo, 360Learning, Sana) ship basic camera proctoring at most, and basic camera proctoring catches none of this.
iTutor's anti-AI checks
On-screen code checks
At random moments a short code appears only on the candidate's screen, never in the examiner's audio, and the candidate must enter it. A relay tool listening to the exam cannot see it. On L&D Growth and Enterprise.
A trap for relayed AI answers
Once in a session, an examiner turn carries a hidden instruction that an AI relaying the questions tends to obey. The next answer is scanned, and an AI that gives itself away is logged as a high-severity event.
Camera checks
Frames from the webcam are checked for phones and extra screens, and face signals flag a candidate who keeps looking away or leaves the camera.
Voiceprint comparison
The candidate's voice is fingerprinted at the start and compared through the session, so a different speaker taking over off camera is flagged.
Browser checks
The exam runs in full screen. Leaving full screen, copying and pasting, opening developer tools and connecting a second display are each logged with a timestamp.
Held for human review
Every event appears in the session report with its time and context. A result that looks AI-assisted is held for review instead of passing, and a reviewer makes the final call.
Pricing: anti-AI checks on every L&D plan
Every L&D plan runs the camera, face, voiceprint and browser checks and the trap for relayed AI answers. The on-screen code checks are on L&D Growth ($22 per seat a month) and L&D Enterprise ($27 per seat a month). L&D Starter ($17 per seat a month) suits low-stakes pilot deployments.
FAQ
Does this work on an ordinary connection?
Yes. The camera sends still frames every few seconds rather than a video stream, and the browser checks run on the candidate's machine, so a normal office or home connection is enough.
How does iTutor handle false positives?
Events are shown in the session report with timestamps and the surrounding transcript, not turned into automatic fails. A result that looks AI-assisted is held for review, and a human reviewer makes the final call.
Can the candidate see the checks?
The on-screen code is visible by design: the candidate reads it from the screen and enters it. The trap sits inside a normal examiner turn. The camera, voiceprint and browser checks run in the background.
Is this defensible to a regulator or auditor?
The session report is the audit trail: every proctoring event is logged with its time, type and context, next to the full transcript and the recorded audio of the session.