Everyone feels anxious before an interview — it doesn't matter if you're a new grad coming out of college or a seasoned professional looking for a change after 20 years. You always have those same questions: am I ready, what are they going to think of me, am I going to say the right thing?
And preparing for an interview is a real challenge in itself — where do you start, what resources do you use, and how do you actually measure your performance?
There are some established, well-known ways you often hear recommended: research the company, go over your resume, prepare answers to the most common questions, and practice them in front of the mirror or record yourself. While these are great recommendations, there is one big gap that isn't covered - a third-party feedback in the loop to point out what you are missing.
However, the rise of AI brought us both improvements to the old methods and brand new ways to practice. Here's how to use AI to practice interviews — a modern, do-it-yourself setup that helps you prepare better, faster, and more efficiently.
Step 1: Gather your resume, the job description, and company info
Prepare all the data you shared with the company during your application and interview process, as well as the information about the job you're interviewing for. This normally includes:
- Your resume
- Your cover letter, if you wrote one
- The job description (JD)
- Any general company info worth knowing
The resume and JD cover the bulk of it.
Step 2: Write a prompt that makes the AI act like a real interviewer
Since you're going to be using an AI frontier model (ChatGPT, Claude, Gemini, and so on) to assist with your preparation, you need a solid, direct prompt to explain the task. Combining general guidelines with the specifics of the job and yourself, you can use the following template:
You are interviewing me for the role below. Act as an interviewer, who is part of this company and stage.
ROLE / COMPANY: [paste the JD; add the company name and anything you found]
STAGE: [recruiter screen / hiring manager / technical screen / panel / final round]
TYPE: [behavioral / screening / competency — or "mostly behavioral with some role-specific questions"]
ABOUT ME: [paste your resume; add two or three lines on what you're worried about — "I ramble," "there's a gap in my work history," "I freeze on failure questions"]
How to run it:
- Ask one question at a time, and wait for my full answer before the next.
- Ask what this company and stage would actually ask — pull from the role's requirements.
- Act like a sharp interviewer. If my answer is vague, generic, or dodges the question, push back with a follow-up. Don't let me off easy.
- Stay in character while I'm answering — don't break to coach me mid-answer.
After each answer:
- Tell me what worked and what was weak, specifically — point at the actual sentence or the missing piece, not just "good job."
- Show me a stronger version in my own voice, built only from my real experience. Don't invent accomplishments.
- Score it 1-100 on substance, structure, and relevance, and tell me what would move it up a point.
Start with the first question.
Feel free to adjust according to your preferences and needs. One line worth keeping is "don't let me off easy," because these models flatter you by default.
Step 3: Run an AI mock interview out loud
AI gives you a real chance to recreate the interview experience, at least to a degree. The best option is the voice mode available in most AI tools — you can practice with ChatGPT, Claude, and others by speaking your answers instead of typing them. Typing hides the things that actually mess you up live: pace, filler words, and whether you really answer the question or drift halfway through.
Practice, get the feedback, get the scores, and rehearse — all of it helps you beat the anxiety, build confidence, and walk in more prepared.
One thing makes this far more useful: don't stop at the feedback. The part that matters is the loop — answer, read the feedback, rewrite, then say the better version out loud.
The rewrite you only read is the one you forget under pressure.
This works best for the talking-heavy rounds — behavioral, competency, walking through your experience. For a live coding screen you'll still write code elsewhere, but you can rehearse explaining your thinking, which is half of what they judge.
For example, you might answer "tell me about a time you handled conflict" with something like: I disagreed with a teammate about priorities, we talked it through, and found a middle ground. On paper that's fine — but a good interviewer, or a model you've told to push, won't stop there. What was the disagreement about? What did you propose? What did they say when you pushed back?
Most prepared answers fall apart on the second follow-up, and that's exactly the weakness you want to find now rather than in the real interview.
Where the manual approach falls short
While this simple setup already gives you a lot, it isn't ideal in a few ways:
- You have to guide the model at every step. The coaching only happens when you think to ask for it.
- It forgets. Even with careful context management, it eventually loses the important pieces from memory — it doesn't track your progress over time, doesn't remember your best answers from past sessions, and can't notice patterns like a story you keep reusing.
- There's no clear way to actually get coached. You get a score and some feedback at the end, but getting the model to work with you toward the best possible version of an answer takes several rounds of back-and-forth — and again, you have to drive all of it.
With that in mind, there's clear room for improvement — and a coaching platform like TutorKit is built for exactly these gaps. It runs the session so you're not coaching yourself, it keeps full track of your progress across every practice, and it shows you where you're getting sharper and where you're still thin.
The manual setup above is well worth doing and costs nothing; TutorKit is for when you want the part it can't give you — a coach that knows you, and knows you a bit better every time you practice.
The tools are new, but the idea is old: you get better by doing it, hearing what's off, and going again.
The only thing that's changed is that now you can do all of that before the interview that counts.
