A Short Complete Answer and an Empty One Are Not the Same Thing

What Does the Timer Actually Do?

It marks an allowance somebody chose, and on at least one platform it reports back how long you took. VidCruiter's description of its pre-recorded product says recruiters set thinking time, recording time, number of attempts and review timers per question ⁶. On a platform like that, the two minutes on your screen is an employer's setting on one question, not a published standard for how long an answer should run.

Willo's knowledge base is specific about what the employer sees at the other end. On that platform the timer counts up rather than down, keeps running until the answer is submitted, and the time recorded includes any retakes ². A candidate who goes past the allocation is not blocked and can still submit. The reviewer then sees how long the answer took, with an indicator showing whether it came in under or over ².

Nothing there says coming in under counts against you.

That is vendor documentation written for employers, and it describes Willo's mechanism only ². On some other platforms a timer is a hard stop and the recording ends when the allowance does. The page does not say how an employer weighs the difference.

Visible to a reviewer is not the same fact as marked down.

What the timer is not is a container the answer has to fill. Employers reading the case for keeping or cutting the async video round are told that, now an assistant can script any answer, it separates candidates on delivery and hardly at all on substance. The remedy that piece recommends is three short questions, not longer answers. The argument on that side of the table is about what gets asked.

Why Does a Short Answer Feel Like a Mistake?

Because the empty screen at the end reads as a gap you failed to fill, and the format gives you nothing else to go on. There is no interviewer nodding, no follow-up question, no sign that the answer landed. The feeling is about the interface. What a words-only scorer reads is the transcript, and a padded second half goes into it alongside a good first half.

On HireVue, what the model reads is the words. Its 2025 explainability statement says the AI relies only on what the candidate says, using no video analysis and no other audio characteristics ⁵. Facial expressions, body language, background and tone of voice are named there as things the system does not assess ⁵. That is the vendor's account of its own product rather than an audit.

The recording still exists, and a person still watches it.

What a belief about the scorer does to answer length has been put to a test. In an online mock asynchronous interview, 124 participants answered questions by voice recording ⁴. Half were told beforehand that an algorithm would evaluate their answers, and half were told a human rater would ⁴. The group told an algorithm engaged in less deceptive impression management, felt they had fewer opportunities to perform, and gave shorter answers ⁴. What differed between the groups was the instruction, not the interview ⁴. The direction is the finding, since the published abstract carries no effect sizes.

That was mostly students, online, with no job at stake, run before 2020, and the impression-management measure was self-reported ⁴. On a system reading transcribed words, a shorter answer gives whoever reads it less to work with.

Finishing early is a different act from cutting an answer because a machine is listening.

The advice that survives here is about specificity rather than length. Why rehearsed STAR answers start sounding generic makes the candidate-side case, and what a follow-up reaches for is the option you rejected.

Check the Answer Against Three Parts, Not the Clock

Check it for content, and the clock stops being the question. A complete behavioural answer names the situation you were actually in, the decision you made and why you made it, and what happened afterwards. If all three are there, the answer is finished, whether that took 45 seconds or two minutes. If one is missing, more time will not supply it.

• The situation: where you were, what the constraint was, who else was in the room.
• The decision: the option you took, and the one you turned down.
• The outcome: what changed, including the part that went badly.

Hiring teams are told to look for the same thing from the other side. The guide on running a structured interview that compares candidates says every answer is rated on one scale whose levels are written as observable behaviour beforehand.

A written level says what an answer contains, never how long it lasts.

One experiment did vary the time allowed and compare ratings, and the result is worth reading precisely. Orji and Bangerter ran a mock asynchronous video interview with 185 participants and set the response limit at one, three or five minutes ³. Performance followed a curve: highest under the three-minute limit and lower at one or five ³. The effect was driven by clarity and relevance ratings, while judged competence stayed stable ³. Speaking time itself plateaued from three minutes ³. Ratings came from human raters rather than from a scoring model ³.

What that experiment manipulated was the limit the employer sets, not what candidates chose to do inside it ³. So the curve describes a setting you do not control. The authors state that offering more than three minutes did not improve outcomes, and that one minute appears too restrictive ³. It was one mock interview, and the ratings that moved were clarity and relevance.

That experiment did not test padding. It did find one minute too restrictive, so a 45-second answer is worth checking for the part it left out.

When Is an Answer Genuinely Too Thin?

When there is not enough in it for the system to work with, which is a narrower case than finishing early. HireVue's statement gives two conditions for an answer to be scored by AI at all ¹. The candidate has to have consented to AI scoring, and the system has to detect enough content in the response ¹.

The document names the ways that second condition fails ¹. An answer too short, an answer not given in the allotted time, speech the system cannot make out clearly enough, and a microphone problem are the four it lists ¹. When either condition is unmet, the response is not scored by the AI and is flagged for human review instead ¹. Where too short begins is not something the document publishes.

Flagged for a person is not the same as scored badly.

The hedges here carry weight. HireVue calls this a small proportion of candidates in any cohort without publishing a figure or a method ¹. Going to a human says nothing about the outcome, only about who reads it.

And this is one platform's process: another may simply score a thin answer low. So the case where the clock genuinely matters is the one where something broke. If the recording cut you off mid-sentence, if the microphone failed, or if you answered a question you had misread, that is a specific thing to tell the recruiter in writing. Name the question it happened on. Whether to spend a retake instead is a separate decision, and re-recording an answer carries its own cost.

The durable move is the one that works in a live room: name the decision, and say what it cost.

See how it works >