AI interview questions: what AI interviewers ask and how answers are scored

"AI interview" describes two quite different experiences, and they need different preparation.

A recorded interview. A question appears on screen, a countdown runs, and you record an answer with nobody on the other side. The questions are set in advance, so every candidate gets the same ones in the same order. HireVue, one of the best-known tools of this kind, says in its candidate FAQ that most of its video-based assessments have between five and eight questions. How much time you get to think, and whether you can retake an answer, is usually set by the employer.

A conversational interview. You talk, and the AI talks back. It listens to your answer and decides what to ask next, which includes asking what you meant. Career1's interviewer works this way.

The invitation usually tells you which one it is. If it mentions recording your answers to questions, prepare for a monologue against a timer. If it describes a conversation, prepare for follow-up questions, because that is where these interviews are won or lost.

The first question is rarely a surprise. These are typical of a first-round interview, human or AI, written the way a conversational interviewer phrases them:

Opening

• "Walk me through what you do and how you got here."

Background and your best work

• "What piece of work are you proudest of? What was your part in it?"

Skills on your resume

• "Your resume mentions PostgreSQL. What did you use it for on that project?"
• "Why did you pick that approach over the alternatives? What did it cost you?"

A real problem

• "Tell me about a hard problem you solved recently. Where did you start?"

Behaviour and experience (for a specific job)

• "Tell me about a time you disagreed with a decision. What did you do?"

Then come the follow-ups, which no list can give you in advance: "What exactly broke?", "How did you know it was fixed?", "What would you do differently?" For the questions each role tends to get, see the role guides for Python developers, data engineers, frontend developers, UX designers and customer success managers.

Before the interview starts, the interviewer is given the text of your resume. Long resumes are shortened to their opening part, so the work you most want to be asked about belongs near the top. For a job application it also gets the job description and requirements; for Career1's own practice and vetting interviews it gets a list of the skills and tools taken from your resume to verify.

It then works through a rough plan rather than a fixed list of questions:

Swipe the table sideways to see every column.

Within that plan, its instructions are specific:

• One question at a time. It is told never to stack questions.
• Follow up when an answer is vague, impressive or suspicious. For the key skills on your resume it asks how, why and what the trade-offs were, and it asks about your projects by name.
• Adjust the difficulty. If you struggle it simplifies the question; if you do well it goes deeper.
• Keep it short. It aims for about eight minutes, and the interview stops at ten. If time runs short, it prioritises the most important skills on your resume.
• Say nothing about how you are doing. It will not reveal scores or judgements during the interview, so a neutral reaction means nothing.
• Be honest about being an AI. If you ask, it confirms it is one.

The interview is in English. If you clearly have trouble continuing in English, it says so politely and uses simpler English.

A spoken AI interview has one mechanical habit worth knowing. The interviewer decides you have finished when you stop speaking for a short moment, under a second. Pause mid-thought and it may start its next question.

That is less of a problem than it sounds. You can interrupt it, and it is told to stop and listen when you do, so just keep talking. If you go silent, it checks in and offers to repeat the question. A quiet room and headphones help, because background noise and your own speakers echoing back are the things most likely to confuse where your turn ends.

After the interview, Career1 writes a report from the transcript of what you said, your resume and, when you applied to a job, the job description. The video, if there is one, is recorded for people to watch; it is not analysed for your face, expressions or appearance. The model writing the report is told to judge only from the evidence in the transcript and resume, to quote the transcript as that evidence, and, where there is a score, to score low and say why if the interview was too short to judge.

What that report contains depends on which interview it was:

Swipe the table sideways to see every column.

There is no pass mark. A finished vetting interview becomes your profile whatever the score, and it is one attempt with no retakes. Practice is where mistakes are free.

A scorer that must quote its evidence can only credit what you actually said. That rewards a few habits.

• Be specific. Name the system, the tool, the size of the problem and the result. A vague answer leaves nothing to quote.
• Say what you did. "We migrated the database" describes a team. "I wrote the migration plan and ran the cutover" describes you.
• Explain why. Choices and trade-offs show understanding in a way that definitions do not, and the interviewer is told to ask about them anyway.
• Only claim skills you can discuss. Every skill on your resume can be questioned. On the vetting report, a skill you spoke about with real detail is marked verified; one you only mentioned is not.
• Admit a gap, then reason. "I have not used that, but here is how I would approach it" survives a follow-up. A bluff does not.

Here is the difference on one question, "How did you use Redis?":

• Weak: "We used Redis for caching, which made things faster."
• Strong: "I added a Redis cache in front of our product search, because the same twenty queries were most of our load. The hard part was invalidation when prices changed, so I keyed entries by product and cleared them on update."

The strong answer gives the interviewer three things to follow up on, and the report three things to quote.

Reading questions is the easy half. The skill that decides these interviews is answering out loud and then handling the second question, and that only improves by doing it.

An AI interview simulator is only worth your time if it speaks, listens and follows up; a tool that reads out a fixed list rehearses the part you were already good at. Career1's free AI mock interview is a private practice interview built from your own resume: once a week, no video, no score, and written feedback on what to work on. It is never shared with companies.

For a step-by-step routine, including how to rehearse follow-ups and review yourself, see how to prepare for a mock interview.