TestGorilla in 2026: What It Records and Flags

TestGorilla is a skills-assessment platform, not a coding-interview platform, which is why most candidates meet it through a sales, support, marketing, operations, or analyst application rather than an engineering one. Its proctoring works differently from the lockdown assessments used for software roles: it runs entirely in an ordinary browser tab, collects a small set of behavioral signals, and rolls them into a three-state summary the hiring company sees next to your score. This article explains what those signals are, what the summary does and does not claim, and why one TestGorilla assessment may penalize AI use while another is explicitly built to measure it.

What TestGorilla monitors during an assessment

TestGorilla's help documentation lists the signals it collects under its anti-cheating measures: webcam snapshots, full-screen tracking, tab switching, webcam status, developer tools, IP-based location, and paste detection.

The specifics matter more than the list:

• Webcam snapshots every 30 seconds. TestGorilla's candidate documentation states plainly: "We take snapshots of you every 30 seconds during your assessment. This is part of our proctoring." These are still images on an interval, not continuous video, and TestGorilla says only the hiring company can view them.
• Full-screen exits and tab switches. The platform logs when you leave full-screen mode or move to another browser tab. This is the same class of page-visibility signal that most browser-based assessments rely on, and it carries the same well-known ambiguity — a notification, a battery warning, or an accidental keystroke produces the same event as looking something up. We covered the underlying mechanism in more detail in our explanation of how interview websites detect tab switching.
• Paste detection. TestGorilla records paste events "to help verify authenticity of candidate work." It records that a paste occurred, which is a different and much weaker claim than knowing where the pasted text came from. The same distinction applies across assessment tooling generally, as we discussed in what copy-paste detection can and cannot see.
• Developer tools. Opening browser devtools during the assessment is logged.
• Webcam status. Disabling or turning off the camera mid-assessment is itself recorded as a signal.
• IP address and approximate location. TestGorilla states it records these "to identify potential unauthorized attempts from the same user" — that is, primarily to spot one person taking an assessment under multiple invitations, rather than to establish where you live.

Everything on that list is collected inside a normal browser. TestGorilla does not ask candidates to install a lockdown application, and its equipment guide asks only for a current version of Chrome, Firefox, Safari, or Edge on a laptop or desktop. That scope is the single most useful thing to understand about it: the platform sees what a web page can see about its own tab, plus a camera feed you consented to.

A behavior tier is a summary, not a verdict

TestGorilla does not hand employers a cheating score. It sorts each completed assessment into one of three tiers based on what it calls "the depth, intensity, and variety" of the recorded behavior:

The important sentence sits in TestGorilla's own documentation, not in candidate folklore: the company states that it "cannot say for certain that a candidate has attempted to cheat," and advises hiring teams against disqualifying anyone on a flag alone.

TestGorilla has argued the same position publicly. In a June 2025 post on its blog, the company wrote that cheating detection methods "aren't perfect – they can trigger false alarms, misinterpret behaviors, and mistakenly flag honest test-takers", and gave its own examples of innocent behavior that produces flags: dismissing a notification, finishing quickly because you knew the material, or glancing at a clock.

That is a vendor describing the limits of its own product, which is unusually candid and worth taking at face value. It is also a reminder that the tier is a prompt for a human to look closer, and how closely they look is entirely the employer's choice. What actually happens after a flag varies a great deal by company, and we walk through the realistic outcomes in our guide to what happens when an assessment flags you.

Not every TestGorilla assessment is proctored the same way

The anti-cheating features are configurable by the hiring company, which means two TestGorilla invitations can behave differently. TestGorilla tells candidates directly that "the company inviting you to an assessment may choose to disable the snapshot feature," and its equipment guide describes a camera and microphone as recommended rather than strictly required for every assessment type.

So the practical answer to "does TestGorilla record me?" is: it depends on what the employer switched on, and you find out at the consent prompt. TestGorilla says it asks for permission before capturing images, so the request for camera access at the start of an assessment is the reliable indicator that snapshots are enabled for your particular test.

Some TestGorilla assessments now measure AI use on purpose

This is the part that has changed most recently, and it inverts the usual assumption that any AI involvement is a violation.

On October 7, 2025, TestGorilla launched skills-based AI video interviews in two formats: a one-way format where you record video answers that are scored against a rubric, and a conversational format in which an AI conducts a six-to-twenty-minute interview with follow-up questions. Each response receives a 0-5 score with stated reasoning, and the company says recruiters can override any score and review the recording or transcript. If you receive a TestGorilla invitation that opens into a video interview rather than a test, you are being recorded in full, not sampled every 30 seconds.

Then on March 3, 2026, TestGorilla announced AI Readiness and AI Fluency assessments covering, in the company's description, foundational ability to work with AI tools, adaptability, understanding of where AI fits into a workflow, judgment about responsible and transparent use, and human-machine collaboration. Those are vendor-described dimensions rather than independently validated ones, but the direction is clear enough: the same platform that flags a tab switch on one assessment sells another assessment whose entire purpose is evaluating how well you use AI.

The two modes are not interchangeable, and the burden is on you to tell them apart. An assessment that requests camera access and locks to full screen is testing you unaided. An AI fluency assessment or a stated AI-permitted exercise is testing something else. When the invitation is ambiguous, ask the recruiter in writing — the same principle that governs take-home assignments where AI rules are unstated.

Reducing false flags without changing your answers

Given that the signals are coarse and the tier is advisory, the sensible goal is not to defeat monitoring but to avoid generating noise that a reviewer has to interpret:

• Close messaging apps, email, and calendar notifications before starting. Notification pop-ups are one of TestGorilla's own examples of a benign trigger.
• Stay in full screen for the duration, and finish a section before dealing with anything else.
• Leave the camera on once you have granted access. Turning it off mid-assessment is logged, and it is a stronger signal than most of the others.
• Use a laptop or desktop on a stable connection. TestGorilla advises against mobile devices, particularly for coding assessments.
• Type your answers in the assessment window. A single paste of a long answer is the kind of event that shows up in a review even when the work is entirely yours.

None of this improves your score. It reduces the chance that a reviewer spends their attention on your session's telemetry instead of your results.

What this means if you use an AI assistant

Control is a desktop AI interview assistant for interviews, assessments, and screen-share workflows, and the honest reading of TestGorilla's setup is that a webcam-proctored, full-screen skills assessment is not the place for one. The signals above are weak individually, but the employer's terms are the real constraint, not the detection: a proctored assessment that asks you to work unaided is asking you to work unaided, and using assistance there is a policy problem regardless of what any tool logs.

Where an assistant is defensible is the other half of the pipeline — preparing beforehand, and live conversational interviews where the ground rules allow it. Those are different situations with different rules, and treating them as one category is how candidates get themselves into trouble.

The short version

TestGorilla collects webcam snapshots on a 30-second interval, full-screen and tab-switch events, paste events, devtools usage, webcam status, and IP-based location, then summarizes them as no issues, some issues, or major issues. The company states that this summary cannot establish that cheating occurred, and recommends that employers not treat it as proof. Which signals are active depends on what the hiring company enabled, and a growing share of TestGorilla's product line — AI video interviews since October 2025, AI readiness assessments since March 2026 — is designed to evaluate AI use rather than suppress it.

Before your next assessment, the question worth answering is not what the platform can see. It is which kind of assessment you were sent. If you want the same treatment for the platforms used in engineering pipelines, start with our breakdown of what HackerRank proctoring detects and records.