The AI mock-interview prompt: make it interview you, not answer for you
48% of job seekers say AI makes them more confident in interviews. Most use it wrong.
In LinkedIn's 2026 research, nearly half of job seekers said AI tools boost their interview confidence. But asking a chatbot for "good answers" just gives you someone else's script. This prompt flips it: the AI plays the hiring manager, pulls the must-haves from a real job post, asks one question at a time and grades you strictly. Then you redo your weakest answer until it scores a 4.
By Preston KirschnerFounder, AI Career HubUpdated
- What job seekers say about AI and interview confidence
- The exact mock-interview prompt (copy it)
- How to run it: voice mode, one question at a time
- The redo-until-4 loop
- Add-your-résumé and follow-up prompts
- What not to paste in
- Job seekers who say AI tools boost their interview confidence.
- 48%
- Self-reported confidence, not interview results. Global Censuswide survey.
- LinkedIn, Talent 2026 research, Jan 7, 2026
- People who have used AI in their job search, or say they plan to.
- 81%
- Mixes current users with people who only plan to.
- LinkedIn, Talent 2026 research, Jan 7, 2026
- How much performance on a coding problem dropped when candidates were watched, vs solving it in private.
- More than half
- 48 computer-science students, whiteboard coding. A small study.
- Behroozi et al., ESEC/FSE, 2020
In this guide
Most people use AI to get answers. Use it to get asked.
LinkedIn's research with Censuswide surveyed 19,113 people across multiple countries in November 2025. Nearly half (48%) said AI tools boost their interview confidence, and 81% said they have used AI in their job search or plan to (LinkedIn, Jan 2026).
Confidence isn't the same as performing well. If you ask a chatbot "what are good answers to these questions?", you get polished answers in someone else's voice, and you've practiced reading, not talking.
Practicing out loud, under a bit of pressure, is the part that transfers. In a randomized study of 48 computer-science students, performance on a technical problem dropped by more than half when they solved it while being watched by an interviewer, compared with solving it in private (Behroozi et al., ESEC/FSE 2020). That study doesn't show AI fixes interview nerves. It shows being watched changes how you perform, so rehearse answering, not reading.
The prompt and routine below are our recommendations, built to avoid the ways chat assistants usually go wrong in mock interviews.
The prompt: paste it with the full job post
Paste this into ChatGPT, Claude or Gemini, then paste the full job post where it says [paste].
You're the hiring manager for the job post below. First list its 5 must-haves. Then interview me one question at a time; wait for my answer. Grade each 1–5 against those must-haves. Be strict: what worked, what's missing, a stronger version. After 6 questions, name my 3 weakest areas. Job post: [paste]
Tips: use voice mode and answer out loud. Rewrite the "stronger version" in your own words; don't memorize it. Don't paste confidential info from your current job.
What each line of the prompt is for
Every line fixes a common failure. If you edit the prompt, keep these.
- What it prevents
- Generic questions. The grades get tied to what this job actually screens for, and you see the list too.
- What it prevents
- A wall of 10 questions at once, or the AI answering for you.
- What it prevents
- Grade inflation. Chat assistants tend to praise. A scale and "strict" push back.
- What it prevents
- Vague feedback. "What's missing" forces at least one critique.
- What it prevents
- No model to aim for. You get one, but rewrite it in your own words.
- What it prevents
- A session with no end. You finish with a to-do list.
| Line | What it prevents |
|---|---|
| "First list its 5 must-haves" | Generic questions. The grades get tied to what this job actually screens for, and you see the list too. |
| "One question at a time; wait for my answer" | A wall of 10 questions at once, or the AI answering for you. |
| "Grade each 1–5… Be strict" | Grade inflation. Chat assistants tend to praise. A scale and "strict" push back. |
| "What worked, what's missing" | Vague feedback. "What's missing" forces at least one critique. |
| "A stronger version" | No model to aim for. You get one, but rewrite it in your own words. |
| "After 6 questions, name my 3 weakest areas" | A session with no end. You finish with a to-do list. |
Run it out loud, one question at a time
A 6-question session takes about 20–30 minutes. This routine is our suggestion.
- Use voice mode if your app has it. Answering out loud is the real skill. Typing lets you edit, and you can't edit in an interview.
- Answer before you think too long. Give yourself a few seconds, like a real interview, then talk.
- Don't read the stronger version out loud as your answer. Read it, then say your own version.
- Write down every score. You'll want the lowest one for the loop below.
- Save the must-haves list. It's a good checklist for your résumé and your "why here" line too.
Redo your lowest score until it hits a 4
One pass shows you where you're weak. The loop is where you get better. Take the lowest-scored answer and ask for it again.
- Find your lowest-scored answer.
- Read the feedback: what's missing?
- Paste the redo prompt below and answer again, out loud.
- Repeat until it scores a 4 out of 5.
- Run the full prompt again tomorrow, in a new chat, to see if it holds.
Ask me question [number] again. Grade my new answer the same way, against the same must-haves. Be strict. If it's not a 4 or 5 yet, tell me the one thing to fix.
Add your résumé, or focus the questions
A very short job post gives thin must-haves. Adding your résumé helps the AI ask about your actual experience, and spot gaps between it and the role.
You're the hiring manager for the job post below. First list its 5 must-haves. Then interview me one question at a time; wait for my answer. Grade each 1–5 against those must-haves. Be strict: what worked, what's missing, a stronger version. After 6 questions, name my 3 weakest areas. Job post: [paste] My résumé: [paste] Also ask at least 2 questions about gaps between my résumé and the must-haves.
Remove your phone number, email and home address before pasting your résumé.
Same rules as before, but only ask behavioral questions ("Tell me about a time…"). Grade whether each answer has a clear situation, what I did, and a result.Paste this as a follow-up in the same chat, after the main prompt.
What not to paste in
Chat assistants keep conversations, and depending on your settings, some may use them to improve their models. Check your app's data or privacy settings before you start.
- No confidential info from your current job: client names, internal numbers, unreleased products, code.
- No personal contact details: strip your phone, email and address from your résumé first.
- No other people's details: don't paste your interviewer's name or notes about colleagues.
- Public job posts are fine. That's what the prompt needs.
- When you need a real example from work, describe it in general terms: "a mid-size retail client", not the company name.
Your AI mock interview
Practice with the jobs you actually want
The prompt is only as good as the job post you give it. Create a free account to save AI, ML and data roles as you find them, then paste each one into a mock interview.
Get a real job post to practice with
Fresh AI, ML and data jobs, all posted in the last 30 days. The prompt works best with the post for a job you actually want.
Browse AI jobsSources
- 1.LinkedIn Research: Talent 2026LinkedIn Pressroom · Jan 7, 2026 · Censuswide online survey: 19,113 consumers (Nov 13–28, 2025) and 6,554 HR professionals (Nov 10–27, 2025), multiple countries. Vendor survey.
- 2.Does Stress Impact Technical Interview Performance?Behroozi, Shirolkar, Barik & Parnin, ESEC/FSE 2020 · 2020 · Peer-reviewed. Randomized study, 48 computer-science students.
- 3.Whiteboard coding stress reduces performance by more than halfFOSSLife (NC State coverage) · 2020
Facts last checked Oct 22, 2026. Vendor surveys are labelled as such, and our own recommendations are marked as ours. Spot something out of date? Tell us.