One-way and AI video interviews: how to prepare, including HireVue

Candidates describe it the same way: awkward, lonely, and impossible to read. There's no nod when an answer lands and no follow-up when it doesn't. You finish, you click submit, and a week later you find out.

It's also one of the most predictable rounds once you know how it's built. The constraints that make it uncomfortable are fixed, and fixed constraints can be practised against.

What a one-way video interview actually is

A one-way video interview, also called an asynchronous, recorded or on-demand interview, replaces the live phone screen with a set of prompts you answer on camera in your own time. The employer writes or picks the questions. You record the answers. Someone at the company, or an automated assessment, reviews them later.

HireVue is the best-known platform for this, and there are several others with the same basic shape. The exact rules are set by the employer, not the platform, so two HireVue interviews from two companies can feel quite different.

1. You get an email with a link and a deadline, usually a few days away.
2. There's a practice question and a camera and microphone check. Use both.
3. Each real question appears as text, sometimes with a short video of someone reading it.
4. You get a fixed amount of thinking time, often thirty seconds to a few minutes.
5. You record your answer against a time limit, commonly one to three minutes.
6. Depending on the employer, you may get one or more retakes, or none at all.
7. Some engineering screens add a coding question or a short assessment in the same session.

How an AI video interview differs

The term "AI video interview" covers two different things, and it's worth knowing which one you're facing, because the preparation differs.

• Recorded, with automated review. The format above, but software transcribes and assesses the recordings before a person looks at a shortlist. You're still talking to a camera, and there are still no follow-ups.
• Conversational. An AI interviewer asks a question, listens, and asks the next one based on what you said. It's much closer to a real phone screen, and the follow-ups are the whole difficulty. What actually happens in an AI interview covers this kind in depth.

If the invite doesn't make it clear, ask the recruiter. "Is it recorded answers, or a live conversation with an AI interviewer?" is a normal question and they'll almost always tell you.

Why engineers find this round harder than it looks

It looks easier than a live screen because nobody can push back. In practice that's exactly what makes it hard.

You get one shot at the whole answer

In a live interview a thin first answer gets rescued by the follow-up. The interviewer asks "and why did you pick that?" and you get to show the depth you left out. In a recorded round there is no second question. Whatever you didn't say in the first two minutes doesn't exist.

So you have to ask yourself the follow-up. Every answer should already contain the why, the cost, and the condition where it stops being true, because nobody is going to ask for them.

The clock punishes the usual technical answer

Engineers explain things in the order they understand them: background first, then the mechanism, then finally the point. On a two-minute timer, that order spends ninety seconds on background and gets cut off before the conclusion. Put the answer first and spend the rest of the time justifying it.

Talking to a camera removes every cue

Without a face to watch, most people either speed up or trail off. Neither is fatal, but both are audible in the recording and in any transcript made of it. The fix is practice, not willpower.

How to answer a recorded technical question

A shape that fits inside two minutes and survives being read as a transcript:

1. The answer, in one sentence. "I'd use a message queue here, because the email step is slow and doesn't need to block the request."
2. What's happening underneath. One or two sentences on the mechanism: what the queue actually buys you.
3. A real example with a number. "On our checkout service this took p95 from about two seconds to three hundred milliseconds."
4. The cost you accepted. "The trade-off is that the email can fail after the user sees success, so we added retries and a dead letter queue."
5. Where it stops being right. "If the email were legally required before confirming the order, I'd make it synchronous."

HireVue interview tips for engineers

HireVue is usually the platform behind the invite, so a few things specific to it are worth knowing. Treat these as general patterns, because each employer configures its own session.

• Read the instructions screen properly. It tells you the number of questions, the thinking time, the answer time and whether retakes are allowed. Those four numbers decide your strategy.
• Do the practice question as if it were real. It's the only chance to hear your own audio and see your framing before it counts.
• Know whether you have retakes before you start. If you do, don't burn one on a small stumble. Correct yourself mid-answer, which reads as confidence, and save the retake for a genuine blank.
• Use the thinking time to pick your one-sentence answer, not to plan the whole speech. The rest follows from the first sentence.
• If there's a coding question, treat it like a live coding round with a silent interviewer: comment your reasoning in the code, handle the obvious edge cases, and leave a note on what you'd improve. What an AI coding interview is actually like covers that part.

Set up the room before you practise the answers

• Camera at eye level, face lit from the front. A window behind you turns you into a silhouette.
• A wired headset or external microphone. Transcription errors from a laptop mic in an echoey room can cost you whole sentences.
• Close everything else. Notifications on screen are distracting, and a notification sound is in the recording.
• Look at the camera, not at your own face on the screen. Putting the preview window right under the lens helps.
• Keep a few prompts about your own projects on a note beside the screen. Not scripted answers: they're obvious when read.

How to practise for a one-way interview

Two things improve with practice, and they need different kinds of practice.

The first is delivery on camera. Record yourself answering three technical questions with a two-minute timer and watch them back. Almost everyone finds the same two problems: the point arrived too late, and there was a term they used but couldn't have explained. Two or three rounds of this fixes most of the awkwardness.

The second is depth, and here the recording doesn't help, because your own camera never asks you why. The way to build answers that already contain the follow-up is to be followed up on, repeatedly, until you start answering the second question inside the first. That's what an AI mock interview does: it asks a technical question, pushes on the weakest part of your answer, and tells you afterwards what was missing. It doesn't record you on camera, so it won't rehearse the lighting. It rehearses the part that decides the round.