AI prompts for interview questions: 10 prompts for ChatGPT, Claude or Gemini

The AI prompts that produce usable interview questions start from the job's real requirements, ask for one competency at a time, and request follow-up probes and a scoring guide with every question rather than a bare list. Below are ten copy-ready prompts that work in ChatGPT, Claude or Gemini, each with what to paste in and what good output looks like. A job description is usually safe to paste; candidate details are not, and as of October 2026 the consumer plans of all three assistants can use your chats to improve their models depending on your settings, while their business plans do not by default.

For the method behind the questions, see what is a behavioral interview. For recruiting prompts beyond interview questions, such as job ads and outreach, see ChatGPT prompts for recruiters.

What should you check before pasting anything into an AI assistant?

Interview questions rarely need personal data, which makes this the safest kind of recruiting prompt. Prompt 7 below is the exception, because it uses a resume. Each vendor's own data terms, as of October 2026:

A training setting does not decide what you are allowed to share. Your organization's policy, a client's confidentiality terms and what candidates were told decide that. Remove client names from confidential searches, and never paste candidate data into an account your organization has not approved for it.

Which prompts turn a job description into interview questions?

1. Pull the competencies out of a job description

Read the job description below. List the 4-6 capabilities that most
separate success from failure in this role. For each: a one-line
definition in plain words, the sentence in the job description it comes
from, and whether it is best assessed by interview, work sample or
reference check.
Do not add capabilities the description does not support. If the
description is vague about something important, list it under
"Questions for the hiring manager".

Job description:
[PASTE JOB DESCRIPTION]

• Paste in: the job description, with the company name removed if the search is confidential.
• Good output: four to six capabilities, each traceable to a line in the description, plus a few questions for the hiring manager.
• Check facts and data: confirm the list with the hiring manager before writing questions; the model has only read the description. No personal data is needed (data rules).

2. Write behavioral questions for one competency

Write 4 behavioral interview questions that assess [COMPETENCY] for a
[JOB TITLE] role. Each question must ask about one specific past event
("Tell me about a time...", "Describe the last time..."), not a
hypothetical or a general habit.
For each question add: 2 follow-up probes that draw out the candidate's
own actions and the result, and one line on what a strong answer
contains.
Context about the role: [2-3 SENTENCES ABOUT THE WORK]

• Paste in: one competency from prompt 1 and a short description of the work.
• Good output: questions tied to situations this job actually meets, such as an upset clinic manager rather than "a difficult stakeholder." Compare with behavioral interview questions with a scoring rubric.
• Check facts and data: reject any question that is really hypothetical. No candidate data is involved.

3. Write situational questions for less experienced candidates

Write 3 situational interview questions for a [JOB TITLE] role assessing
[COMPETENCY], suitable for candidates with little direct experience.
Each question describes a realistic situation from this job in 2-3
sentences, with one real constraint (time, budget, a conflicting
request), and asks what the candidate would do first and why.
For each, describe a weak, an adequate and a strong answer in one line
each.

• Paste in: the competency and job title.
• Good output: short scenarios with a real trade-off, where a strong answer shows reasoning, not a magic solution.
• Check facts and data: ask someone who does the job whether each scenario is realistic. Nothing personal goes into this prompt.

Which prompts add probes and scoring?

4. Add follow-up probes to questions you already use

Here are interview questions my team already uses. For each, write 3
follow-up probes: one that asks what the candidate personally did (not
the team), one that asks for a specific detail (a number, a name, a
tool, a date), and one that asks what went wrong or what they would do
differently. Keep each probe under 15 words.

[PASTE QUESTIONS]

• Paste in: your existing questions.
• Good output: short, neutral probes that do not suggest an answer. Probes are also the most reliable way to tell rehearsed answers from real experience; see how to detect AI-generated interview answers.
• Check facts and data: remove any probe that leads the witness ("So you escalated it, right?"). No data concerns.

5. Write an anchored scoring rubric

Write a 1-4 scoring rubric for this interview question: [QUESTION]
It assesses: [COMPETENCY] for a [JOB TITLE].
For each score, describe what the candidate's answer contains, in
observable terms (what they did, how specific they were, what the
result was), not adjectives like "good" or "excellent".
Score 1 = no relevant example or only a hypothetical.
Score 4 = a specific example with the candidate's own actions, a
measurable or observable result, and a lesson applied later.

• Paste in: one question and its competency.
• Good output: four anchors two interviewers could apply to the same answer and agree on. Behavioral anchors for interview scores has a library to compare against.
• Check facts and data: make sure the anchors reward the job's requirements, not a speaking style. No personal data.

6. Write technical questions a non-technical recruiter can ask

I am a recruiter without a technical background screening for a
[JOB TITLE]. Required skills: [LIST].
Write 5 screening questions I can ask, each about the candidate's own
past work (not trivia), with: what a credible answer mentions, what a
weak or evasive answer sounds like, and one follow-up I can ask without
technical knowledge.
Do not include questions that need me to judge code or correct answers
I cannot verify.

• Paste in: the job title and required skills.
• Good output: questions about decisions and trade-offs the candidate made, with listening cues in plain language. The approach is in how to interview for technical roles as a non-technical recruiter.
• Check facts and data: have the hiring manager confirm the "credible answer" cues, since the model can be confidently wrong about technical details.

Which prompts tailor questions to a candidate and keep them lawful?

7. Write questions from the claims on a resume

Below are claims a candidate makes about their experience, written by me
without their name or any personal details. For each claim, write one
question that asks for a specific example, and 2 follow-ups that would
show whether the experience is first-hand (details only someone who was
there would know).
Do not judge whether the claims are true.

Claims:
[e.g. "Led the move to a new warehouse system at a 3-site distributor"
"Cut invoice errors for key accounts"]

• Paste in: the claims, rewritten by you, not the resume itself.
• Good output: questions specific enough that a prepared generic answer would not fit them.
• Check facts and data: this is the one prompt that touches candidate information, so keep it to anonymous claims, and use only an account your organization has approved (data rules).

8. Review your questions for legal and bias risk

Review these interview questions for a [JOB TITLE] role in [STATE OR
COUNTRY]. Flag any question that could reveal or relate to age, race,
national origin, religion, sex, pregnancy, disability, family or
marital status, or other protected characteristics, or that asks about
salary history. For each flag: the risk in one sentence and a
job-related rewrite.
Also flag questions that are not clearly tied to a job requirement.
Say clearly that this is not legal advice.

[PASTE QUESTIONS]

• Paste in: your question list and where the job is based.
• Good output: flags with job-related rewrites, such as "Are you able to work the scheduled weekend shifts?" in place of a question about childcare.
• Check facts and data: the model does not know every local rule. Check the final list against illegal interview questions and your own advice. No personal data.

Which prompts build a complete interview?

9. Assemble an interview guide with timing

Build a [45 / 60]-minute structured interview guide for a [JOB TITLE]
from the questions below. Include: a 3-minute opening script that
explains the format, the questions in a sensible order with time for
each (leave room for probes), 5 minutes for the candidate's questions,
and a closing script with next steps.
Mark 2 questions as must-ask. Do not add new questions.

[PASTE YOUR CHOSEN QUESTIONS, PROBES AND RUBRICS]

• Paste in: the questions, probes and rubrics you chose from prompts 2 to 6.
• Good output: a guide an interviewer could run without preparation. Compare with the interview guide template.
• Check facts and data: time it by reading the questions aloud; most guides are too long. No personal data.

10. Split questions across an interview panel

Here are the competencies and questions for a [JOB TITLE] hiring
process with [NUMBER] interviewers: [NAMES OR ROLES].
Assign each competency to one interviewer so nothing is covered twice
and nothing is missed. Give each interviewer 3-4 questions, matched to
what they can judge best (for example, the future manager takes
prioritization). Output one table per interviewer.

[PASTE COMPETENCIES AND QUESTIONS]

• Paste in: competencies, questions and interviewer roles. Roles are enough; names are not needed.
• Good output: one short list per interviewer with no overlaps. The panel interview plan template shows the rest of the setup.
• Check facts and data: confirm each interviewer can actually judge what they are given.

How do you turn AI drafts into questions your team trusts?

1. Choose, then edit. Ask for more questions than you need and keep the few that fit this job best.
2. Test them on a colleague. If someone who does the job cannot give a strong answer, the question or the anchor is wrong.
3. Use the same core questions for every candidate for the same role, so scores can be compared.
4. Keep the good ones. Approved questions with probes and anchors belong in a shared library; how to build an interview question bank covers versioning and retiring questions that leak.