AI Cheating Detection for Online Exams
ChatGPT in another tab, an AI agent filling in the answers, a helper on a screen share. ProctorSafe spots how each one leaves its mark — and keeps up as the tools change.
Recognises answers no human hand produced — text that was never typed, keystrokes faster than any person, choices clicked without the mouse moving there.
Tells a paste brought back from an AI assistant apart from ordinary copy-paste within the exam.
Every signal lands on a timeline your reviewers can check and a candidate can contest.
The new cheating landscape
Generative AI has changed what cheating looks like. A candidate no longer needs a hidden phone with notes: they can paste the question into ChatGPT in another window, let a browser agent fill in the answers, or have someone dictate over a screen share. Proctoring that only watches the webcam misses most of this, because none of it happens in front of the camera.
What ProctorSafe detects
Answers brought back from an AI assistant
The candidate copies a question to ChatGPT in another tab, window or app and pastes the answer back. ProctorSafe tells that apart from ordinary copy-paste within the exam, and shows the reviewer how long the candidate was away before the paste.
AI agents filling in the exam
Text that appears in an answer field without being typed, written there by a browser agent, an extension or an automation script.
Typing no human can produce
Runs of keystrokes faster than any hand can release them, the signature of software typing on the candidate's behalf. Ordinary fast typing doesn't trigger it.
Choices clicked by software
Multiple-choice answers selected by a script, or by a pointer that jumped straight onto the answer instead of moving there.
A helper on the other end
Screen sharing, remote desktop tools such as TeamViewer and AnyDesk, and virtual cameras such as OBS or ManyCam that let someone else follow the exam in real time.
A second voice in the room
When the candidate's voice has been calibrated at the start, a different speaker during the exam is flagged. Audio is analysed on the device and never uploaded.
Leaving the exam
Switches to other tabs and windows, with how long the candidate was gone. Brief accidental switches are distinguished from sustained absences.
AI services contacted from the exam page
Requests from the exam page to domains you have blocklisted, such as AI assistants, are flagged. Only the domain is checked; full URLs never leave the device.
Continuously evolving
Updates without any work from you
New detections reach your exams through the script you already embed. No upgrade, nothing to redeploy.
Built from real cases
We study the sessions where cheating got through and close the gap, so our detection follows how candidates actually cheat.
Your scores never shift silently
A new signal is recorded and shown to reviewers straight away, but it only counts toward your trust scores once you choose to turn it on.
Fair to honest candidates
Some legitimate tools, such as dictation software and accessibility aids, can look like assistance. That is why every signal is evidence for a person to weigh, never an automatic penalty. Reviewers see what happened, when, and in what context, and you can exempt sections such as open-book questions where outside material is allowed.
Local detection, minimal data collection
Unlike cloud proctoring that uploads hours of video for later analysis, ProctorSafe runs detection locally in the candidate's browser. Video, audio and the text of a candidate's answers stay on the device; only structured event data — what happened and when — is sent to your review dashboard.
Evidence for reviewers, not automated punishment
Every detection lands on a timestamped timeline next to a 0–100 trust score, so reviewers can triage high-risk sessions quickly without watching footage. You decide how much each kind of signal counts, and every decision stays contestable — aligned with your academic integrity policies.
Get started
Explore our interactive demo or request a demo for a walkthrough tailored to your assessments.
Related reading
Dive deeper with guides from our articles library.
The Rise of the “Cheatbots”: How AI Proctoring Must Evolve for ChatGPT-5
Why webcam-only proctoring is no longer enough, and which integrity signals help against generative AI assistance.
Read articleTechnical architectureEdge AI vs. Cloud AI: The Future of Real-Time Exam Monitoring
Latency, reliability, and privacy trade-offs between edge (on-device) and cloud-based proctoring architectures.
Read articlePedagogy & ethicsEthics in AI Proctoring: Bias, Fairness, and the Human-in-the-Loop
How to design and operate proctoring that avoids biased outcomes, documents limitations, and keeps humans accountable.
Read articleFrequently Asked Questions
Common questions about privacy, compliance, and integration.
Ready to Get Started?
Start a free trial and run real sessions against your own application, try the interactive demo in your browser, or contact us for a walkthrough tailored to your program.