The Practice Question Before an AI Interview Is for Your Setup

What Is the Practice Question, Exactly?

On a HireVue OnDemand interview, the questions at the start are practice. The help center says there are three generic ones, that you can redo them without limit, and that the recordings are never uploaded or submitted to anyone. Nobody except you can view them, and once you move past the playback they are gone ¹.

Those assurances come from HireVue's own help center, not from an audit of it.

The belief that it counts is expensive in both directions. Some people skip the practice section entirely, assuming it is already being watched.

Others treat it as the interview, then arrive at the first real question tired.

Platforms differ in what they hand you. Spark Hire's candidate help describes a practice video you can review and retake before the real questions begin ². On a Radancy screen, formerly myInterview, questions arrive one at a time, there is no skipping ahead and coming back, and you can review each answer right after recording but not at the end ³.

The instructions screen on your own link is the thing to read, not a forum thread about a different employer.

Use It on Your Microphone, Not Your Answers

Use the practice round on your equipment, not on your wording. On HireVue it is the only point in the process where you get to hear your own audio played back. Once the real interview is submitted, HireVue gives you no direct access to your recording, and the hiring company controls it and decides whether to give you access ⁴.

Three things are checkable in that window and nowhere later:

• your audio: play the practice answer back and listen to the room rather than to your phrasing
• your frame: head and shoulders in view, camera near eye level
• your clock: get a first sentence out and see where the timer sits when you land it

Audio comes first because of what a bad microphone does downstream. HireVue's statement gives two conditions for an answer to be scored by its AI at all: the candidate consented to AI scoring, and the system detects enough content in the response. A microphone problem, speech it cannot understand clearly enough, an answer too short, or one not given in the allotted time means the response goes to human review instead ⁵. Declining AI scoring in the first place lands an answer in that same human queue ⁵.

That queue takes a small proportion of candidates, HireVue says, and it publishes no figure for it ⁵.

Human review is not a verdict in either direction. It moves the response to a person, and says nothing about what that person concludes.

Framing is for the human watching, and the reason is a specific one. HireVue's 2025 statement says its AI uses only the transcribed words, with no video analysis and no assessment of facial expressions, body language, background or tone of voice ⁶. The recording still exists, and recruiters still watch it.

The clock is the third check, and the practice round is where you learn what it feels like. On Spark Hire, a limited think time means you click to see the question, a countdown runs, and recording begins by itself when the countdown ends ².

Don't Rehearse Content on the Practice Round

The practice questions are generic, and by default the real ones cannot be previewed. HireVue says the actual interview questions cannot be seen beforehand, because every candidate is given the same amount of time to review, prepare and respond to each one ¹. An answer polished against a practice prompt is an answer to a question you have not read.

Preparation for the real question happens inside the real question's timer.

That timer does more than it appears to. Basch and colleagues tested preparation time in asynchronous video interviews and found it improved interview performance, with the gain running through active preparation of the response rather than through reduced strain or impression management ⁷. Those were mock interviews rated by people, not by a scoring model.

What you get varies by employer, so read the screen before the first question. HireVue's statement lists interview question retries as a yes or no setting the employer chooses, with unlimited as the vendor's recommended configuration ⁸. Preparation time per question is configurable in the same way, from zero to five minutes or left unlimited ⁸.

Two candidates can describe the same interview very differently and both be right.

A subscription is not required to get a rehearsal, and it helps to know what a practice tool offers. Yoodli's interview page describes an AI roleplay that asks follow-up questions and returns a speaking report with analytics such as pacing and filler words ⁹. That is delivery coaching from a tool that is not the system any employer runs.

None of it can show you the questions you will actually be asked.

Scripted answers are the worry from the other chair, and hiring teams are weighing whether the async video round is worth keeping when every answer is read off a second screen. For the hour before the link opens, how to prepare for a one-way video interview maps it out.

Does Practice Actually Change Anything?

Yes, in the trials that measured it, though what moved was performance as human raters saw it in mock interviews. Langer and colleagues randomly assigned 70 students at two German universities to virtual interview training or to an e-learning control. The training was a simulated interview with a virtual character that gave feedback in real time. In blind-rated mock interviews afterwards, the trained group scored higher on performance (d = 0.60) and hireability (d = 0.55), with lower interview anxiety ¹⁰.

That is a small student sample and a single short mock interview, with agreement between raters low enough to leave the effect sizes imprecise ¹⁰. The training also targeted nonverbal behavior, which is the part HireVue states its scorer ignores ⁶.

Lukacik and Bourdage ran three mock asynchronous video interview experiments across six design features. Longer preparation time and the chance to re-record raised interview performance ratings. Faking warnings, and being told a human or an automated system would evaluate the answers, changed nothing about behavior, reactions or performance ¹¹.

Again: online panels and students, researchers' raters, and no job on the line ¹¹.

The practical read is narrow. Use the preparation time to plan the answer, and use whatever re-records the employer allows. Being told a person or a system would evaluate the answers made no measured difference there, and the automated system existed only in what participants were told ¹¹.

If practicing with a chatbot is the plan, whether it counts as cheating to practice interview answers with ChatGPT draws the line between rehearsal and live assistance.

See how it works >