HireVue video interview: what it scores and how to prepare
In a live interview, a half-formed answer gets rescued. The interviewer hears you circling, asks a narrowing question, and you land it. A one-way video interview removes that entirely. You read a prompt, you get a short time to think, and you talk into a camera until the timer stops. Whatever came out is the whole record.
The format, concretely
The employer configures the interview, not HireVue: how many questions, how long you get to think, how long you can talk, and whether you may re-record. Commonly reported settings look like this:
- Three to five questions, mostly behavioural, sometimes one role-specific.
- Around thirty seconds of preparation per question, then up to two or three minutes to answer. HireVue's own tips say most on-demand interviews ask for an answer in just a few minutes.
- A practice question at the start. HireVue mentions it in its candidate tips; use it to check your camera, microphone and lighting.
- Retakes vary. Some employers allow one re-record per question, others none. The invitation says which. If it does not, prepare as though the first take is the only take.
What is being scored in 2026
This is worth stating plainly, because the folklore is out of date. HireVue dropped visual analysis from its assessments and announced it publicly in January 2021, saying language analysis had become predictive enough on its own. Its 2025 AI Explainability Statement is explicit: when an interview is AI-scored, the AI relies only on what you say. Your speech is converted to text by a third-party transcription system (Rev.ai), and the text is scored against the competencies the employer chose. HireVue says it does not assess facial expressions, body language, your background or tone of voice.
So the practical advice is not "smile more". It is: speak in complete sentences, clearly enough to be transcribed correctly, and make the structure of the answer audible in the words themselves. HireVue also says its platform does not make hiring decisions — the employer does, and people at the employer may watch the recording. The same qualities that read well in a transcript read well to a person who has eleven of these queued up.
Which integrity checks HireVue documents, and what it does not (no ChatGPT detector, no facial analysis since 2021), is covered in Can HireVue detect AI?
The shape of an answer that survives no follow-ups
Because nobody will ask "and what did you do?", you have to supply it yourself. Roughly ninety seconds, in four beats:
- Situation, one sentence. Where, when, what was at stake. Not the company history.
- The problem, one sentence. What specifically was wrong, with a number in it if you have one.
- What you did, forty seconds. First person singular. "We" is the most common way to talk yourself out of a competency score.
- The outcome, one sentence, then stop. What changed, measured if possible. Then stop talking, even if there is time left on the clock.
Silence at the end of a strong answer reads as confidence. Filling the remaining forty seconds with an appendix does not.
The setup that quietly decides half of it
Nobody is grading your flat, but everything that degrades the audio degrades the transcript, and the transcript is what gets assessed.
- A quiet room and a decent microphone. Background noise and echo turn into transcription errors.
- Light in front of you, not behind. HireVue's own tips say the same: light from behind makes it hard to see you.
- The strongest connection you have. HireVue recommends the strongest Wi-Fi signal possible; a cable is better still, with nothing streaming on the same connection.
- Do not obsess over eye contact. HireVue's tips tell candidates not to worry about it, and its AI does not analyse video. A camera at eye level still helps the person who watches later.
The preparation that actually transfers
Write four stories, not twenty answers: a conflict, a failure, something you shipped, something you fixed under pressure. HireVue's own advice is similar — prepare a list of brief stories about challenges, your approach and the results — and almost every behavioural prompt is one of those wearing a different hat.
Then record yourself once, on your phone, answering "tell me about a time you disagreed with a colleague" in ninety seconds. Watch it. It is uncomfortable, and it removes more bad habits in one sitting than a week of reading example answers. If your recorded round is a live technical interview rather than a one-way video, read what to expect in a Karat interview.
Sources
- HireVue — 2025 AI Explainability Statement (PDF)
- HireVue — Candidate interview tips
- SHRM — HireVue discontinues facial analysis screening
Checked against these pages on 14 October 2026. Question counts and timings are set by each employer; the figures above are commonly reported, not published by HireVue.
FAQ
Does HireVue analyse my face?
No. HireVue dropped visual analysis from its assessments (announced in January 2021). Its 2025 AI Explainability Statement says the AI relies only on what you say and does not assess facial expressions, body language or your background.
Does HireVue score my tone of voice or accent?
HireVue says its AI does not assess tone of voice. Your speech is transcribed by a third-party speech-to-text system and the text is scored, so what matters is that your words come through clearly — a quiet room and a decent microphone help more than a performance.
Can I retake a HireVue answer?
It depends on the employer. Some allow one re-record per question, others none, and the invitation says which. Use the practice question at the start to check your setup, and treat every scored question as final unless told otherwise.
How long should a HireVue answer be?
Around ninety seconds for a behavioural question, even when the timer allows more. Land the outcome and stop — trailing filler is what makes an otherwise good answer read as unstructured.
Does a human ever watch the recording?
HireVue says its platform does not make hiring decisions; the employer does, and the recording is available to them. How many recordings a person watches depends on the employer, so assume someone may.