You Have an AI Interview Tomorrow: What It Is, What It Scores, and How to Prepare

An email says your first-round interview will be "conducted by an AI interviewer". A common reaction is that it sounds cold, like talking to a black box. This guide covers what the interview actually is, what gets scored and what doesn't, and how to spend the evening before it.

We build an AI interviewer at Hyrr, so the specifics below describe how ours works. Other platforms differ. Where it matters, check the instructions your employer sent you.

What it actually is

It's a spoken conversation. You talk and the interviewer talks back, in real time, and you can interrupt it the way you would a person. Its words appear on screen as captions, so you won't miss a question because of an accent or a bad speaker.

A typical structure:

1. A short warm-up. Some rapport, and a chance to settle.
2. The assessed part. The employer's questions, with follow-ups.
3. Your questions, then the close.

The follow-ups are where it stops feeling like a form. If you say "we migrated the database", it will ask what you did. If you say "it got a lot faster", it will ask by how much. If you skip the outcome, it will ask how it turned out. It isn't trying to catch you out. It's collecting the evidence the rubric needs.

What it scores, and what it doesn't

This is the part people worry about most, so here it is plainly:

• Everyone applying for the role gets the same structured interview, scored against a rubric the employer wrote before anyone applied.
• Answers are matched against defined examples of strong and weak answers (behaviourally anchored rating scales), not against how confident you sounded.
• Every score traces back to something you said. The employer sees the reasoning, not just a number.
• You're assessed on what you show in the interview, not on your CV, past employers' brand names or where you studied.
• The live interviewer can't see you. In a real assessment your webcam feed isn't sent to it during the interview. After the interview, an automated review checks the recording for signs that someone or something else is helping you, and what it finds can affect your result.

What isn't held against you: pausing to think. No single answer is on a clock. If you go quiet while you think or type, the interviewer waits, and after a long silence it checks in politely. It doesn't hang up. On a coding or whiteboard exercise, think out loud: explaining your reasoning is part of what's assessed, and minutes of silence there is noticed.

The rules that can actually end it

Most monitoring doesn't interrupt you. It's recorded and reviewed after the interview. Three things can end the interview: switching tabs, leaving fullscreen and pressing PrintScreen. Do two of them and the interview stops. So:

• Close everything else before you start, including chat apps that steal focus.
• Turn off notifications.
• Don't go looking something up in another tab mid-answer. Saying "I'm not sure, here's how I'd find out" is a better answer anyway.

Technical prep (do this tonight, not at 8:58)

• Use a laptop or desktop with an up-to-date Chrome or Edge. The pre-check names anything your browser can't do, and some browsers are refused outright.
• Test your connection. The pre-check measures it and needs at least 1.5 Mbps upload. Home Wi-Fi is usually fine, but a crowded café often isn't.
• Use headphones with a mic. That stops the interviewer hearing itself through your speakers, and it stops coughs or a door slam cutting it off mid-sentence.
• Have your ID ready if the invitation mentions identity verification. Not every employer turns it on, but when they do it comes before the interview.

Answer prep: built for the follow-ups

Because the interviewer probes until it has the full picture, a polished thirty-second answer isn't enough. Prepare stories you can unpack:

1. Pick four or five real situations. A hard problem, a conflict, a mistake, something you shipped, something you led.
2. For each one, know four things: what the situation was, what you did (say "I", not "we"), what the scale was (users, money, time, team size) and how it turned out.
3. Expect "why" twice. Why that approach, and why not the obvious alternative.
4. For technical rounds, think out loud. If there's a coding editor or whiteboard, the interviewer can see it and will ask about it. It won't write the solution for you, but it will follow your reasoning, and the reasoning is what gets scored.

Do one practice run out loud

The biggest gap for most people isn't knowledge. It's that they've never answered a probing follow-up out loud, under time, to something that won't nod along. Rehearsing in your head doesn't train that.

Hyrr's practice interviews use the same live interviewer. You can pick a library template (behavioural, system design, coding screens, new-grad and more), build one from your own résumé, or practise against a real posted job. You get a full coaching report afterwards, question by question. Practice uses credits.

Run a practice interview first →

After the interview

The recording and transcript are scored against the rubric, and the employer reviews the result. Some employers also release a written feedback message to you, with one strength and the most important thing to work on. If yours does, read it. It's specific to what you said.