What a Hiring Platform Logs When You Switch Tabs

What Can a Browser Tab Actually See?

A page in a browser is told when it stops being the active tab, and that is most of what a tab log holds: focus left here, at this moment, for this long. It does not carry the contents of the other window. HackerRank lists tab proctoring, copy and paste tracking and multiple-monitor detection among the signals its Screen product collects ¹.

The window you opened is another program's business, and the page you left has no reach into it. The same goes for anything beyond the screen. A platform can report that a second monitor is attached, which HackerRank also lists among its signals ¹. Attached is not the same fact as what is on the screen.

A tab log is an absence with a timestamp on it.

Two kinds of marketing describe this badly in opposite directions. Proctoring marketing can read as total visibility. On the other side, a copilot sold to candidates advertises that it stays invisible during video calls and does not appear in screen shares ². That is a seller's promise rather than a tested result, and what it sells is concealment, which an assessment asking for unaided work has already ruled out. What the documentation supports is narrower than either.

Neither pitch is an audit, and neither was written for you.

A page watching itself and software watching the machine are two different arrangements. A tab counter lives inside the assessment page. A lockdown browser, a desktop agent or a screen recording is separate software you install or approve, and it reaches further.

Where that difference shows up is the consent screen.

Which Platforms Say They Log a Tab Switch?

Several of the ones a candidate meets, by their own documentation. HackerRank's plagiarism model says it analyzes code-writing patterns, time taken, copy and paste activity and tab-switching behavior, then classifies a suspicious attempt as High or Medium ³. Alex sells cheat detection as an add-on that watches for background voices, excessive tab switching and unusual delays before an answer ⁴.

Vervoe's anti-cheating page describes a different mix. It compares the text of answers across candidates, flags a candidate who appears to finish from more than one location inside a short window, allows one attempt per email address, and lets an employer require full-screen mode ⁵.

Several pieces are sold as add-ons or optional settings, and which are switched on is the employer's call.

So the real answer depends on a setting you cannot see from inside the assessment window. HackerRank's product page lists the signals and publishes no accuracy or false-positive figure for them ¹. It does state the rule the signals point at, and states it per test ¹. A test designed to allow AI is fair for everyone to use it on. A test that instructs otherwise treats the use as suspicious activity ¹.

That is a rule about instructions, not about hardware.

Which means the question worth asking is not what the software can see. It is what this assessment told you to do. Instructions that permit the web and an assistant make a tab switch unremarkable. Instructions that ask for unaided work make it the thing you were asked not to do, logged or not.

An employer-side page argues this from the other chair: What employers are told about proctoring versus an AI-open assessment makes the case that, for most roles, permitting the assistant beats watching for it. What proctoring software can see covers the wider family of tools, including the ones that record more than a tab count.

What Happens After a Tab Switch Is Flagged?

It goes to a person, in the flows these vendors publish. HackerRank puts the overall precision of its plagiarism model at 85 percent, meaning flagged sessions are correct 85 percent of the time, and says on the same page that human oversight remains crucial ³. The figure is the vendor's own, with no test set on the page. Flagged attempts are highlighted in the candidate report, and a recruiter can replay keystrokes to see why ³.

CodeSignal describes the same shape for its proctoring: sessions evaluated by automated analysis, with human reviewers validating flagged activity and the final decisions, and results returned within an hour ⁶. Its page gives no false-positive rate for the suspicion score it describes and no candidate-side appeal route ⁶.

What reaches a reviewer is a question about the session, not an answer about you.

Research on those signals is more careful about what they prove. In a study of 315 participants who took a knowledge test unproctored and then under proctoring, behavioral data including response times and browser tab switching predicted which participants were classed as cheating, while honesty questionnaires did not ⁷. That label was inferred from the gap between the two scores, not observed ⁷. The authors are explicit that a long pause or a tab switch does not by itself indicate cheating, because nobody knows what a person is doing when they leave the window ⁷.

The stakes there were a gift-card draw, not a job.

Tabs move for ordinary reasons. A calendar alert, a dropped connection, the job description you wanted to reread, a housemate at the door. None of that shows up in a count, which is the argument for saying something rather than nothing when a flag surfaces in a later conversation.

A specific true account of what you were doing is worth more than a denial, and it is the version most likely to match what a reviewer replaying the session sees. The vendors tell employers to review before deciding, and a reviewer can only weigh what they are given. Which is why the weight sits on whoever reads the report rather than on the counter that produced it. Ask who that person is if the process gives you an opening.

A flag being reviewable is a smaller comfort than a flag being rare.

Read the Consent Screen Before the Timer Starts

A lockdown browser, a screen recording or a camera check is a different arrangement from a tab counter, and the consent screen is where to look. That is much wider access than a tab log. CodeSignal's own page says that its proctoring, where an employer switches it on, records video, audio and screen activity for the entire assessment session, and that identity verification asks for a government-issued photo ID matched against the person on camera ⁶.

Coding, behavioral signals, webcam images and screen activity are what HackerRank's candidate AI notice says its AI features may analyze, and proceeding with a feature counts as acknowledging the notice ⁸. The same notice lists what a candidate may request: an alternative selection process, human review of a decision AI made or substantially assisted, and an explanation of how AI was used if an adverse decision is made about you ⁸. Those requests go to the employer, and the notice does not say the employer must grant them.

Declining has a cost worth knowing before you click. HackerRank's own FAQ says a candidate who does not give consent will not be able to participate in the feature that uses the automated tool ⁹. Whether an alternative exists is the employer's decision, and the recruiter is the person to ask, before the timer starts.

That page is the one part of the process written for you.

Two adjacent questions have their own pages. Whether you have to agree to screen or camera recording covers a consent screen that asks for more than a tab count. What an employer can and can't see about your AI use covers the account histories that sit outside the assessment window.

The instructions are shorter than this article and they outrank it.

See how it works >