Interview prep · Updated August 2026

How to Prepare for a One-Way AI Video Interview

No interviewer, a countdown timer, and a recording that gets scored before a human ever sees it. Here is what these systems actually measure, what you are owed as a candidate, and how to rehearse without faking anything.

To prepare for a one-way AI video interview, do four things: work out the six or seven questions the role is likely to ask, prepare real STAR stories that answer them, rehearse each one out loud on camera until you can open strongly within the first ten seconds, and fix your setup so your face is lit and your microphone is clear. The system is scoring what you say far more than how you look — so the highest-return practice is speaking fluently to a camera with no one nodding back at you.

These arrive as a link a few days after you apply, usually as the first screen for a graduate scheme or internship. You read a question on screen, get a short preparation window, then record straight into your webcam. No one is on the other end, there are no follow-up questions, and you get no read on whether an answer landed. That silence is the part students find hardest — and the part you can practise.

What these systems actually score

The most visible vendor in early-career hiring is HireVue. Its candidate FAQ says an assessment usually takes under 20–30 minutes, that you get several minutes per question, and that it "considers the whole context of a response, not just certain words". It describes these as a first step, with person-to-person interviews following for candidates who progress.

The question students ask most is whether the camera is judging their face. For HireVue, no: the company announced in January 2021 that it had removed facial-expression analysis, saying its research found that improvements in language analysis meant visual analysis no longer added significant value. The change followed criticism that scoring micro-expressions disadvantaged autistic and neurodivergent candidates and people whose expressions read differently across cultures.

Two caveats. That is one vendor, and other platforms are built differently. And dropping facial analysis does not make the camera irrelevant — a human still watches the recordings that make the shortlist. So your energy and clarity matter to a person, while the automated layer mostly reads the substance of what you said.

Know what you are owed before you record

Depending on where the role is based, the employer owes you disclosures — and those are a far more reliable description of what the system does than anything you will find on a forum.

Read the consent screen rather than clicking through it: it usually names the vendor and states what the assessment measures, which tells you what to emphasise. If you have a disability that makes a timed recording unfair — a stammer, anxiety, a processing difference — ask the recruiter for an adjustment before you start. That request is routine.

Step 1 — Predict the question set

One-way screens are structured — every candidate gets the same questions, drawn from the competencies in the job posting — which makes them far more predictable than a live interview.

Here is a job posting for a role I have applied to.

[Paste the full job description]

The first stage is a recorded one-way video interview, roughly 5-6 questions, a couple of minutes per answer.

Task:
1. List the competencies this employer is most likely to screen for, ranked, and quote the line in the posting that supports each one.
2. For each competency, write the 1-2 most likely interview questions in the employer's own phrasing style.
3. Flag any question that would need a specific example rather than an opinion.

Do not invent details about the company that are not in the posting.

You will usually end up with six or seven questions that cover almost everything.

Step 2 — Build answers that open fast

A recorded answer has a specific failure mode: you spend forty seconds on context, the timer runs out, and you never reach what you actually did. The fix is to front-load — say the headline in your first sentence, then fill in the story. If you have not built a story bank yet, do that first; our guide to practising behavioral interviews with AI using the STAR method covers the inventory step. Then compress each story for camera:

Here is a STAR story from my real experience:

[Paste your story]

Rewrite it as a spoken answer for a recorded video interview with a 2-minute limit.

Rules:
- The first sentence must state the headline: what the situation was and what I achieved. No wind-up.
- Then Situation and Task in about 20 seconds combined, Action as the longest part, Result at the end.
- Use only facts I gave you. If something is missing, write [FILL: what?] instead of inventing it.
- Spoken English, short sentences, no corporate buzzwords. It must sound like me talking, not an essay.
- Then give me the answer again as 5 keyword bullets I can glance at while recording.

Those five bullets are what you actually use. A full script read aloud is obvious on camera; keywords keep you on track while still sounding like a person thinking.

Step 3 — Rehearse on camera, then critique the transcript

Record yourself under real conditions: timer running, one take, no restarting when you stumble — recovering from a stumble is exactly the skill being tested, because you cannot edit a one-way interview. Then transcribe what you said (your phone's voice notes app will do) and hand the text to an AI for a blunt critique.

Below is a transcript of me answering "[the question]" in a recorded video interview. It was [X] seconds long.

[Paste transcript]

Critique it honestly and specifically:
1. Quote my opening sentence. Does it state a clear headline, or is it wind-up? Rewrite it if weak.
2. Did I say what I personally did, or hide behind "we"? Quote examples.
3. Did I give a concrete result? Quote it, or say it is missing.
4. Count my filler words and list the repeated ones.
5. Was the length right for a 2-minute limit? Say what to cut.

Be blunt. Do not compliment me. End with the single highest-impact fix.

Default AI feedback is uselessly encouraging, which is why the prompt demands quotes — the quoted phrases surface habits you cannot hear in yourself. If pacing and filler words are your weak point generally, the AI public speaking coach workflow goes deeper on delivery.

Step 4 — Fix the setup

Technical problems sink more of these than weak answers do. HireVue's candidate FAQ lists a minimum connection of 0.35 Mbps, allows a computer, tablet or phone, and asks for lighting good enough that your face is clearly visible and a working microphone. Beyond that:

• Light in front of you, not behind. A window at your back turns you into a silhouette. Face the window, or put a lamp behind the screen.
• Camera at eye level. Stack the laptop on books. A webcam angled up your nose reads badly to the human reviewer.
• Look at the lens, not at yourself. Watching your own thumbnail reads as avoiding eye contact. Cover it with the sticky note holding your keyword bullets.
• Use headphones with a mic if your laptop microphone is echoey — speech clarity feeds the part of the assessment that actually matters.
• Do the practice questions, and check your retake count on screen. Practice recordings cost nothing, but unlimited practice does not mean unlimited real attempts.
• Block the full window. Allow around 45 minutes even for a 20-minute assessment, and close anything that can pop up mid-recording.

Mistakes that cost students the next round

• Reading a script. Flat, over-tidy, eyes tracking across the screen. Keywords only.
• Burning the clock on context. Thirty seconds of background and no result. Headline first.
• Recording at 11pm on the deadline. Tired delivery, and no room to fix a technical failure.
• Treating it as less real than a live interview. It is the gate to every later round.
• Letting AI write your experience. The recording is the easy thing to fake; the next round is where it breaks.

Where this fits in an application

The one-way interview sits between your application and a live round, so your answers should match the CV that got you here — see how to tailor your resume and cover letter with AI. For technical roles, expect a coding assessment alongside it: prepping for technical coding interviews safely. Still trying to reach this stage? AI-assisted internship outreach emails covers the step before.

FAQ

Does the AI judge my face or my accent?

HireVue announced in January 2021 that it had removed facial-expression analysis from its assessments, saying its internal research found that advances in language analysis meant visual analysis no longer added significant value. Not every vendor works the same way, so the reliable move is to read the disclosure the employer must give you, which has to state what characteristics the system evaluates.

Can I retake a one-way video interview answer?

That depends on the employer, not the platform. The number of allowed attempts is a setting the company configures, and it is shown on screen before the real questions start. Read that screen before you record anything, and assume you get one take unless it explicitly says otherwise.

Can I read a script off my screen during the recording?

It reads badly. Eyes tracking across text, flat delivery and unnaturally tidy sentences are obvious on camera to the human reviewer who watches shortlisted recordings. Keep four or five word bullet points on a sticky note just below the webcam instead, so your eyes stay near the lens.

Is it cheating to prepare for a video interview with AI?

Rehearsing with AI is preparation, the same as practising with a careers adviser. The line is generated content presented as your own experience: having AI invent a project, a metric or a result you did not achieve is misrepresentation, and it collapses at the next interview round when someone asks a follow-up question.

What are my rights if AI is used to assess my interview?

It depends where the job is based. Under the Illinois Artificial Intelligence Video Interview Act, employers hiring for Illinois-based roles must tell you AI may analyse your video, explain how it works and what it evaluates, get your consent first, limit who sees the video, and delete it and all backup copies within 30 days if you ask. New York City's Local Law 144 requires an annual independent bias audit of automated employment decision tools, a public summary of the results, and at least 10 business days' notice to candidates before the tool is used.

Bottom line

Predict the questions, prepare real stories that open with the headline, and record yourself until talking to silence stops feeling strange. The winning move is the unglamorous one: rehearse out loud, keep every claim true, and sort out your lighting before the deadline.