An AI interview is a first-round interview run by software instead of a person. You pass it the same way you pass any structured interview: give specific answers with evidence, in a clear order, to the question that was actually asked. What changes is the setting. Nobody nods, the follow-ups are relentless about detail, and everything you say is transcribed and scored.
Hiring platforms such as Mercor, micro1 and Alignerr use AI interviewers to screen applicants, and a growing number of employers use one for the first round of ordinary job applications. This guide explains what the software is listening for and how to prepare.
What an AI interview is
There are three formats, and the invitation usually tells you which one you are getting.
- Live voice or video conversation. An AI interviewer speaks, listens to your answer and asks a follow-up based on what you said. This is the closest to a human interview and the most common format on newer platforms.
- One-way video. A question appears on screen and you record an answer within a time limit. There are no follow-ups.
- Text chat. You type answers to a structured set of questions.
Most run for 10 to 30 minutes. Many are followed by, or include, a short practical task such as writing code or a paragraph in an on-screen editor.
How it is scored
Vendors differ, but the scoring is built on the transcript of what you said. Your words are compared with the requirements of the role and with your CV. In practice that means four things carry the result.
- Substance. Did you describe real work in enough detail to show you did it? Named tools, numbers and decisions count. Adjectives do not.
- Relevance. Did the answer address the question and the job, or did it drift?
- Structure. Could a reader follow the answer from situation to result without rereading?
- Consistency. Does what you said match your CV? Claims that contradict it are flagged.
A recruiter then reads the report. The AI rarely makes the final decision alone, but a thin transcript gives the recruiter nothing to argue for you with.
Before the interview
Fix the setup first
Speech recognition is unforgiving of bad audio. Use headphones with a microphone, sit somewhere quiet, and close other tabs. If a screen share or camera is required, test both in the same browser you will use. A garbled transcript scores as a weak answer, however good the answer was.
Prepare six stories, not sixty answers
You cannot predict the questions, but most of them can be answered from a small set of stories. Prepare six from your own work, each with a number in it:
- a project you are proud of and your exact part in it
- a problem that was hard and how you worked through it
- a mistake and what you changed afterwards
- a disagreement with a colleague or client and how it ended
- a time you had to learn something quickly
- a result you improved, with the before and after
Read the job description the way the software will
The interviewer is given the job description and builds its questions from it. Mark the three or four requirements that appear most often and make sure one of your stories covers each. Use the same terms the listing uses when they are true of you.
During the interview
Answer in four parts
Use the STAR order: the situation, the task you were given, the action you took, and the result. Spend most of your time on the action, in the first person. “We migrated the database” tells the scorer nothing about you. “I wrote the migration scripts and ran the cutover on a Sunday night” does.
Aim for 60 to 90 seconds
Shorter than a minute and there is usually no evidence in the answer. Much longer than two and the structure falls apart. If the interviewer wants more, it will ask.
Expect the follow-up
A good AI interviewer does not move on from a vague answer. It asks what your specific part was, why you chose that approach, or what the number was. Treat this as a second chance to add the detail, not as a sign you are failing.
Say when you did not hear the question
Ask for it to be repeated. Guessing at a question you half heard produces an off-topic answer, and off-topic answers score low on relevance.
The mistakes that sink qualified people
- Reading from notes or another screen. It sounds recited, and proctored interviews record window switching and pasted text as integrity signals.
- Answering in general terms. “I’m a strong communicator” is a claim. A story where you communicated something difficult is evidence.
- Saying “we” throughout. The interviewer is assessing you, not your team.
- Filling silence. A two-second pause to think reads fine in a transcript. Thirty seconds of filler does not.
- Treating it as less serious than a human interview. The transcript and scores go to the person who decides.
How to practise
Reading advice changes very little. Answering out loud does, because the difficulty is in speaking a structured answer under mild pressure to something that is not a person. Practise the way you will be tested.
- Find the listing you are applying for and copy the job description.
- Run a spoken mock interview against that description, without notes.
- Read the feedback and note the one question you handled worst.
- Rewrite that answer in STAR order and say it aloud twice.
- Run a second, shorter session a day later.
Hyrovo’s AI interview practice is built for this. You paste the job description, choose 5, 10 or 15 minutes, and talk with Dami, an AI interviewer that asks questions based on the listing and your CV and follows up when an answer is thin. Afterwards you get a private coaching report: a readiness score, scores for content, structure, communication and role fit, what to work on, and model answers to the questions you struggled with. It costs $0.50 a minute and only you see the report.
If you are on the other side of the table
Employers use the same technology to run first-round screens. If you hire, see our guide to AI interviews for hiring teams or how Dami interviews candidates for employers.