AI proctoring
AI proctoring for hiring tests, with the evidence
Remote tests are easy to game: ChatGPT in another tab, a friend on a second screen, or an interview copilot that listens and writes answers invisibly. TalentScope watches for every common method and shows you the exact moment, so you decide on evidence, not a guess.
Every common way to cheat, and how TalentScope catches it
Choose the level per job. Basic is ₹49 and Prime ₹199 per candidate. “App” means the job requires the TalentScope desktop app, which works with Basic or Prime.
| How people cheat | How TalentScope catches it | Basic | Prime | App |
|---|---|---|---|---|
| ChatGPT or Gemini in another tab | Every switch away is logged with the screen at that moment; the app names the AI site or app in front | ✓ | ✓ | ✓ |
| Copying and pasting answers | Pastes flagged with their length; sudden bursts of perfect text against slow typing | ✓ | ✓ | ✓ |
| Leaving the test window | Full-screen exits and lost focus logged; the app keeps the test in front and pulls it back | ✓ | ✓ | ✓ |
| Sharing the screen with a helper (Zoom, Meet, Teams) | Screen sharing detected in the browser; the app also flags meeting apps running during the test | ✓ | ✓ | ✓ |
| Second monitor or TV | Displays checked before the start and throughout | — | ✓ | ✓ |
| Someone else in the room | A second face on camera, other voices, lip movement that doesn’t match speech, a 360° room scan before starting | — | ✓ | ✓ |
| Someone else takes the test | Government ID check, liveness check, and the face matched against the ID throughout | — | ✓ | ✓ |
| Notes or a phone off camera | Looking away, head turns and long silences flagged; the room scan shows the desk | — | ✓ | ✓ |
| Virtual machine or developer tools | Virtual machines and browser developer tools detected | — | ✓ | ✓ |
| Hidden “interview copilots” (Parakeet AI, Cluely, Final Round AI…) | Windows that hide themselves from screen recording are found on the computer itself | — | — | ✓ |
| AI apps and local AI (ChatGPT, Claude, Copilot apps, Ollama, LM Studio) | Running AI apps, connections to AI services and local AI models checked every few seconds | — | — | ✓ |
| Remote control by a helper (AnyDesk, TeamViewer) | Remote-control tools and remote sessions detected | — | — | ✓ |
| Virtual cameras or another app listening | Virtual camera devices and other apps using the microphone flagged | — | — | ✓ |
Our promise, plainly: no proctoring can guarantee 100%. Anything unusual comes with evidence you can check yourself.
How proctoring works in a TalentScope test
- 1
The candidate agrees and checks their computer
One screen explains what is recorded. One check covers camera, microphone, lighting, display and screen sharing.
- 2
The test runs, timed and watched
Tab switches, pastes, full-screen exits, extra screens, extra faces and voices are logged as they happen, each with the screen and camera moment.
- 3
You review the evidence, then decide
The report puts the answers score next to the integrity score and a timeline of every flagged moment. Mark each one as a concern or not an issue.
Evidence, not a “suspicious” label
- Every flag links to proof: the recording, a screenshot and the question being answered at that moment.
- Severity you can trust: strong evidence counts toward the integrity score; weaker signals wait for your review.
- A fair record: every decision is logged, and candidates can see what was recorded about them.
Answers and integrity on one page
Each candidate report shows the answers score, the integrity score, the recording and the computer checks together. A high score with a clean record moves on; a high score with strong flags gets a second look. Export it as a PDF for the hiring manager.
The TalentScope desktop app · ₹500 per test
It sees what a browser can’t
Browser-based tests only see the browser tab. The newest cheating tools live outside it: AI copilots that listen to the question and show answers in a window that screen sharing can’t capture. The app runs the test on the candidate’s Windows or Mac computer and watches the whole machine while the test runs. Tick “Require the TalentScope app” on a job and every candidate takes it there.
What the app checks
- Hidden copilots, AI apps, AI websites and local AI models
- Remote-control and meeting apps
- Second displays and virtual cameras
- That the test stays in front, start to finish
What it never does
- Open or read the candidate’s files
- Look at browsing history or passwords
- Watch anything when no test is running
Taking a TalentScope test? What candidates can expect. Hiring developers? See the coding test platform.
Proctoring questions
Can TalentScope detect ChatGPT during an online test?
In the browser, every switch to another tab or window is logged with a screenshot of what was in front, and pasted text is flagged with its length. With the desktop app, TalentScope also names the AI website or app that came to the front, and finds AI apps and local AI models running in the background.
Can it detect interview copilots like Cluely, Parakeet AI or Final Round AI?
Yes, with the TalentScope desktop app. These tools draw a window that hides itself from screen sharing and recording, so a browser cannot see them. The app looks for such windows on the computer itself and names them in the report.
Is a flag the same as cheating?
No. A flag is a moment worth checking, with the evidence attached. Strong evidence counts toward the integrity score; weaker signals wait for the recruiter to mark them as a concern or not an issue. Every decision is logged.
What do candidates see and agree to?
Before the test starts, candidates are told exactly what will be recorded for that test and agree to it. They run one quick check of camera, microphone, display and screen sharing. Candidates can see what was recorded about them.
Does the desktop app read candidates’ files?
No. The app does not open or read files, does not look at browsing history or passwords, and watches nothing when no test is running.
Can any proctoring guarantee 100%?
No, and we do not claim it. TalentScope watches for every common method listed on this page, and anything unusual comes with evidence you can check yourself.
Run your next test with evidence behind every score
Free plan, no card needed. Basic is ₹49 per candidate, or let each candidate pay a small skill-assessment fee.