Assessment design
What a Timed Assessment Does When You Close the Tab
Sometimes, and the platform decides rather than a general rule. Sapia's chat interview is untimed with an emailed link that saves your progress. Willo's timer counts up until you submit, cannot be reset by closing the browser, and shows the employer whether you went over. Spark Hire lets you leave and return between questions, though a countdown on a single question rolls straight into recording. Read the platform's own help page before the first question, not after.
The takeThe invitation is the wrong document to look for this in. It carries a deadline and a link, and almost never says whether the clock stops when you do. Much of the answer sits in the platform's public help pages, some written for candidates and some for the employers who set the clock, and they take two minutes to read. Treat that reading as part of the preparation. Asking the recruiter is the other route, and it is a factual question with a factual answer.
Where Olive fits
Open a role and see what the work shows
If the assessment an employer sends is Olive, the clock is a 40-to-60-minute assignment on the candidate's own schedule, and working openly with the AI assistant is the point. A person writes or edits the six findings in words, and the candidate is granted the identical report the employer receives.
Assess your shortlistWhat Happens When You Close the Tab?
What happens depends on which platform sent you, and the documented answers are genuinely different. Willo's knowledge base says its timer counts up until the answer is submitted, and that the recorded time includes any retakes. Candidates invited by email, SMS or an ATS integration cannot reset that clock by restarting the assessment, closing the browser or refreshing the page 1.
That is Willo documenting its own mechanism for employers. On at least one other platform a timer is a hard stop, and the recording ends when the allocation runs out. On Willo the employer sees the total, with an indicator showing whether you came in under or over the allocated time 1. The pages quoted here say nothing about what a reviewer sees where the clock simply cuts you off.
On that platform, closing the tab buys you nothing.
Sapia's candidate explainer describes the opposite end of the range. Its chat interview is five to seven mostly behavioural questions, answered in 50 to 150 words each. The page says the interview is not timed, that candidates are not penalised for how long they take, and that progress saves and resumes through an emailed link 2. That is the vendor's own account of its product rather than an audit of it, and other chat screeners may put a clock on the answer.
Spark Hire sits between the two. Its candidate help article says each one-way question carries a think time, a number of takes and a response time limit. With limited think time you click to see the question, a countdown starts, and recording begins by itself when the countdown ends 3.
A second rule on that page is broader: a candidate can pause and return through the interview link as many times as needed before the deadline 3. That is the interview as a whole, between questions.
It does not say a question can be paused once its own countdown is running.
Check the Pause Rules Before the First Question
Two minutes on the platform's help site answers this better than any general article can, including this one. If the invitation names the platform, its help pages are public and cover exactly this question. Read them before you open the assessment, then check the instructions screen as well, because one platform can run more than one way: Spark Hire's page describes questions with limited think time and questions with unlimited think time 3.
Four answers to find on that help site before you click start:
- Whether the clock is a countdown, a timer that counts up, or nothing at all
- Whether the link reopens after you close it, and how many times
- How long the whole thing is meant to take, end to end
- What happens to an answer you are part-way through
The clock's purpose is worth knowing too. Employers reading what a time limit on an assessment actually measures are told that a generous outer bound is a scheduling device. The same article argues that a tight countdown mostly separates who had the faster tool and the quieter room, now that an assistant does the producing. Some of what you are sent will carry no clock at all.
Start when you have the whole window.
If a disability or a condition makes the window itself a barrier, ask before you start rather than after. How to ask for an accommodation on an AI interview covers the wording and where the request goes. A request made in advance has somewhere to land, and one made after a missed deadline is a harder conversation for everybody in it.
Does Leaving the Window Get Logged?
Sometimes, and on a coding assessment it can become a line on the report. HackerRank's knowledge base says its AI plagiarism model analyses code-writing patterns, time taken, copy-paste activity and tab-switching behaviour, and classifies suspicious attempts as High or Medium. The feature requires candidate consent before the test starts when it is enabled 5.
The same page puts a number on how often the flag is right. It gives the model an overall precision of 85%, meaning flagged sessions are correct 85% of the time, and names that as the reason human oversight remains crucial 5. That figure is the vendor's own, published with no test set, period or base rate, and it covers coding questions only. By its own arithmetic roughly one flagged session in seven is wrong.
It says nothing at all about how many cheating sessions go unflagged.
Published research on this kind of signal is more careful than the vendor page. In a 2021 study, 315 participants took a knowledge test unproctored online and then proctored in a lab. Self-report honesty measures predicted nothing about who cheated, while test data and para data, including response times and browser tab switching, did 6.
The authors are explicit that those signals are indicators and not proof, because nobody knows what a person is doing when they leave the window 6. The incentive there was entry into a gift-card draw rather than a job, and cheating was inferred from the gap between the two scores rather than observed 6.
A tab switch is a question, not a verdict.
None of which makes stepping away free.
If a flag reaches a report, a person is meant to read the reason behind it, and an explanation only helps if you can give one. Saying what interrupted you, and roughly when, is a normal thing to put in an email.
What to Do If You Get Interrupted Anyway
Finish the attempt if you can, then tell the employer what happened. HireVue's help article says a candidate disconnected between questions can resume from the next question, with previous answers saved. An answer interrupted mid-recording is different: the vendor says it has most likely been submitted, re-entry moves you on, and there is no way to reset that question 4.
"Most likely" is the vendor's own hedge, and it is doing real work.
The page does not say what a recruiter sees of an answer that stops halfway.
The route to a second attempt runs through the employer. HireVue's advice is to finish the interview and then contact the company you applied to, which can send a retake if it decides one is needed. The vendor says it does not have permission to provide another interview itself 4.
Nothing in that obliges an employer to send one.
Spark Hire documents a different exit. A candidate who has run out of takes, or was allowed only one, and still does not want to submit the answer is offered the option to skip the question 3. The page describes the interface, and does not say what a skipped answer does to an application.
A short factual message beats silence. What happened, when it happened, and which question it hit is what a recruiter needs to decide whether to reissue the link. Send it the same day. How to prepare for a one-way video interview covers the hour before, which is where most of this gets solved. Treat the window like an appointment: a door you can close, a phone on silent, and nothing stacked behind it.
Common questions
If I close the browser, does the timer stop?
Do not count on it. Willo's knowledge base says its timer counts up until the answer is submitted, and that the recorded time includes any retakes. Candidates invited by email, SMS or an ATS integration cannot reset it by restarting the assessment, closing the browser or refreshing the page. Closing the tab is no way back to zero. The reviewer sees the total, with an indicator showing whether it fell under or over the allocated time. That is Willo's own mechanism, and elsewhere a timer can be a hard stop that ends the recording instead.
Are any AI interviews genuinely untimed?
Yes, at least one format is. Sapia's candidate explainer describes its chat interview as five to seven mostly behavioural questions, answered in 50 to 150 words each. The page says it is not timed, that candidates are not penalised for how long they take, and that progress saves and resumes through an emailed link. It also says there is no camera and no video recording. That is the vendor describing its own product rather than an audit of it, and other chat screeners may time the answer.
Can I pause in the middle of recording an answer?
Assume not. Spark Hire's help article says that with limited think time you click to view the question, a countdown starts, and recording begins automatically when it ends. That leaves nothing to pause. The same page says a candidate can pause and return through the interview link as many times as needed before the deadline. That is the interview as a whole, between questions. If you have takes left, a retake is the mechanism for an answer you dislike, and on Willo, whose timer counts up, the recorded time includes every retake.
My connection dropped mid-answer. Can I redo that question?
Not through the platform, on HireVue. Its help article treats a mid-recording interruption as a submission: most likely sent, with re-entry moving you to the next question and no way to reset the one you lost. That leaves the employer as the only route, and HireVue tells candidates to finish the interview first, then contact the company, which can send a retake if it decides one is needed. The vendor says it does not have permission to issue another interview itself, and nothing obliges an employer to grant one.
Does switching tabs during an assessment count against me?
It can be recorded, and being recorded is not the same as being judged. HackerRank's knowledge base lists tab-switching behaviour among the signals its plagiarism model uses on coding questions. It gives that model an overall precision of 85% and says human oversight remains crucial. The research is more cautious still. The authors of a 2021 cheating study say response times and tab switches are indicators and not proof, because nobody knows what a person is doing away from the window.
References
- 1. Time to answer support.willo.video Supports the claim that the timer counts up until submission, that recorded time includes retakes, that restarting, closing the browser or refreshing does not reset it for candidates invited by email, SMS or an ATS integration, and that the employer sees the total with an under or over indicator.
- 2. AI Chat Interview Guide for Candidates: How It Works, Tips & FAQs sapia.ai Supports the claim that the chat interview is five to seven mostly behavioural questions of 50 to 150 words each, is not timed, does not penalise how long a candidate takes, and saves and resumes progress through an emailed link.
- 3. How a Spark Hire one-way video interview works candidates.sparkhire.com Supports the claims that each question carries a think time, a number of takes and a response time limit, that a limited think time countdown starts recording automatically, that a candidate can pause and return through the interview link before the deadline, and that skipping is offered when takes run out.
- 4. Disconnected During the Interview hirevuesupport.zendesk.com Supports the claims that a disconnection between questions resumes from the next question with answers saved, that an answer interrupted mid-recording has most likely been submitted and cannot be reset, and that the route to a retake is the employer rather than the vendor.
- 5. AI Plagiarism Detection support.hackerrank.com Supports the claims that the model analyses code-writing patterns, time taken, copy-paste activity and tab-switching behaviour, classifies attempts as High or Medium, requires candidate consent, and carries a stated overall precision of 85% with human oversight called crucial.
- 6. Caught in the Act: Predicting Cheating in Unproctored Knowledge Assessment d-nb.info Supports the claims that 315 participants took a knowledge test unproctored then proctored, that self-report honesty measures did not predict cheating while test data and para data including response times and tab switching did, and that the authors treat those signals as indicators rather than proof.
6 sources, numbered by first appearance. How Olive sources claims
General guidance for hiring teams. What works at one company and one volume may not transfer to yours.
Olive assesses how a person works with AI. It does not detect AI-written documents, and it never produces a score, a ranking, or a match percentage for a person. Candidates read the same report the employer reads.