Does an AI Interview Assistant Survive Talview, HackerEarth FaceCode or Mettl Proctoring? What They Say They Detect

"AI interview assistant proctoring Talview HackerEarth Mettl" is a search people run when they are about to sit an online assessment and want to know whether a copilot will get them through it. This article answers with the platforms' own words. We fetched Talview's, HackerEarth's and Mercer Mettl's public pages on 9 September 2026 and quote what they say they record, block and flag. Where a page did not render its feature text for us, we say so instead of filling the gap.

The conclusion is not subtle and we state it up front: do not use live AI assistance inside a proctored assessment. Not because a particular tool will or will not be seen, but because the assessment's rules forbid it, the software is designed to notice, and the consequences in an Indian campus or lateral process are severe. Use a practice tool before the test. That is MockMate's policy and it is the only honest advice anyone in this category can give.

First, what "proctored" means

A normal video interview is a meeting app that transmits your camera, your microphone and whatever you choose to share. A proctored assessment is different in kind. The platform runs software on your machine (a secure browser, a lockdown wrapper, or a monitored web session) that records what it decides to record, restricts what you can open, and produces integrity flags that a recruiter reviews. You agree to its candidate terms before starting. Our explainer on what Zoom, Meet and Teams actually capture covers the meeting-app case; this page is about the other one.

HackerEarth

HackerEarth is the platform behind many Indian campus and lateral coding assessments and behind FaceCode, its live coding interview product.

What its help center documents. HackerEarth's "Proctoring settings" article lists the controls a recruiter can switch on for a test:

• A "Smart Browser" to "enforce stricter proctoring settings in your assessments".
• "Disable copy and paste", so that a "candidate cannot copy the code ... and paste it into the code editor".
• Webcam snapshots that "take a snapshot of the candidate at random intervals when the test is live".
• ID verification, where "candidates must upload a government ID ... which HackerEarth verifies".
• A full-screen mode restriction so "candidates can take the test in full-screen mode".
• Tab and window switching warnings that give "the candidate a warning each time they move out of the test environment".
• IP address restrictions to "contain the test-taking to a single location".
• Question shuffling and surprise questions.

What its proctoring feature page claims. The public "Online Proctoring Tool" page goes further than the settings article. It says the tool "records screen, webcam to prevent copy-pasting and candidate impersonation", "captures periodic high-resolution screenshots of the candidate's face and desk space", "tracks gaze direction and micro-movements to flag when candidates look away", "blocks the use of other applications during the test", "prevents candidates from sharing their test screen via screen-sharing tools", "restricts window resizing to prevent candidates from viewing external content", compares submissions against "GitHub, StackOverflow, and other code databases in real time", and, in so many words, "prevents the use of generative AI tools during tests".

FaceCode. HackerEarth's FaceCode page describes its live interview product: "candidates can solve programming questions in 40+ languages with snippets and auto-complete". It is a shared editor with a human interviewer watching, not a lockdown test. The interviewer sees your code as you write it and can play it back; the sensible reading is that anything you did not type yourself will be obvious from the typing pattern, and any assistance needs the interviewer's permission.

What this means. A test with the Smart Browser, full-screen restriction, webcam snapshots and tab-switch warnings enabled is an environment where opening any other window is itself a logged event, and where the platform states it blocks other applications. There is no version of "use a copilot quietly" that fits that description.

Talview

Talview runs proctored assessments and video interviews for many Indian recruiters and certification bodies.

Automated proctoring. Talview's "Automated Proctoring" page lists as popular features "Live Screen Recording", "More Than 8 Critical Flags", "Mobile Browser Proctoring", "Multiple Face Detection" and a "360° Environment Check". It describes "real-time behavior analysis and flagging of violations" for "suspicious behaviors, such as unusual eye movements, the presence of unauthorized materials, or attempts to navigate away from the test screen".

Screen integrity. Talview's glossary entry on screen sharing in proctoring is the most specific public statement we found from any of the three vendors about second displays: "running split screen or connecting a second monitor is monitored for, alongside screen sharing and screen mirroring, as part of a complete screen-integrity check", and "an unexpected external display, an unauthorized second monitor, or mirrored output are all patterns the monitoring is designed to catch".

Secure Browser. Talview's support article on Secure Browser compatibility tells candidates "don't run any screen recorders, third-party security apps, or remote desktop tools" and, for access errors, to run the browser "as Administrator". A lockdown browser that asks for administrator rights is running with more control over your machine than a meeting app has.

What this means. The "second monitor for the copilot" idea, which appears in a lot of forum advice, is exactly what Talview says it monitors for. Eye movement toward that monitor is a separate flag. Screen recording is continuous.

Mercer Mettl

Mettl is the third platform Indian candidates meet constantly, from aptitude tests in campus drives to lateral technical assessments.

What we could verify. Mettl's public pages for "Online Remote Proctoring" and "Secure Exam Browser" list the products, "SecureProctor" and "Mettl Secure Browser", but the feature copy on those pages did not render in our fetches; we got navigation and product names only. We are therefore not going to quote capability claims for Mettl as if we had read them. Mettl's own search listings describe a browser that launches the test in a kiosk or lockdown mode with additional tabs disabled and navigation restricted, and AI proctoring flags for face, voice, mobile phone and distraction; treat that as the vendor's summary rather than something we verified line by line.

What this means. A secure browser that runs in kiosk mode is, by design, the only thing on your screen, and switching away from it is the event it is built to log. Whatever the exact flag list, the environment is the same as HackerEarth's and Talview's: recorded, restricted and reviewed.

Summary table: what each platform publicly says

We have deliberately kept the Mettl column hedged. If you are sitting a Mettl test, the candidate instructions you receive will list what applies to you; read them.

Why "does it survive proctoring" is the wrong question

Every vendor in the AI interview category, including MockMate, describes some form of screen-share resistance. MockMate's Windows companion has a compact tool window, one-key hide, a Privacy Mode and capture resistance on supported Windows display paths, and MockMate explicitly says this is platform-dependent and not a guarantee of invisibility. Competitors say "undetectable". Neither claim is about a proctored assessment, for four reasons:

1. The rule is the problem, not the pixels. A proctored test's terms forbid outside help. Using it is a violation even if nothing is flagged.
2. The capture path is different. Meeting apps transmit what you share; a secure browser records through its own stack, often with administrator rights, and in kiosk mode there is no "other window" to be excluded from.
3. The webcam is the second sensor. Gaze tracking and face detection do not care about window flags. Reading an answer off any surface moves your eyes.
4. The review is human. Flags go to a recruiter or proctor. A cluster of "navigated away" events plus gaze drift plus a suspiciously clean submission is enough for a rejection without anyone proving what software was open.

In Indian processes the downstream cost is high: a flagged campus assessment can mean exclusion from further drives at your college, and lateral offers can be withdrawn on integrity grounds. Our ethics article covers the placement and service-agreement angle.

What to do instead

Prepare with a practice tool, then take the assessment alone. MockMate's Practice mode runs in a browser, includes technical rounds with a code editor and coding and system-design sessions, adapts its follow-up questions to your answers, and ends with a report on question-level answers, response timing, coaching evidence, recurring weaknesses and a recommended next practice. Eligible accounts get three free Practice starts of ten minutes each. The coding interviews use case explains the workflow.

Set up the environment the platform asks for. One display, no remote-desktop or recording software running, webcam clear, ID ready, a quiet room. Most of the flags above are avoidable by simply following the candidate guide.

Save Live assistance for rounds where it is permitted. MockMate's Acceptable Use Policy is explicit: use Live assistance only where the organisation, interviewer or applicable rules permit assistance and disclosure, and use Practice when permission is unclear. Disclosed use in a meeting, a mentor call or an interview where the interviewer has said yes is what Live is for. A Talview, HackerEarth or Mettl assessment is not.

Bottom line

HackerEarth says it records screen and webcam, tracks gaze, blocks other applications and prevents generative AI use during tests. Talview says it records the screen, detects multiple faces, flags eye movement and navigation away from the test, and monitors for second monitors and mirroring. Mettl's product pages name a secure browser and an AI proctoring suite whose details we could not verify from the rendered page. Against that, no AI interview assistant "survives" proctoring in any sense that matters, because the rule is broken before any flag is raised. Practise before. Sit the test alone. Keep Live for where it is allowed.