AI Interview Practice Prompts From a Job Description

The most useful AI interview practice starts with the job description, not a generic list of common questions. First extract and rank the role’s competencies, then connect each one to your real evidence. Use that map to generate varied questions, run a one-question-at-a-time mock interview, and score each answer against specific proof points. This turns AI from a question generator into a repeatable practice system.

Key takeaways

• Convert the job description into a competency map covering responsibilities, knowledge, behaviors, tools, outcomes, and constraints.
• Prioritize roughly four to six competencies that are critical at entry and most likely to distinguish strong performers.
• Keep your evidence separate from the job description and instruct the AI never to invent experience, metrics, or ownership.
• Run practice sequentially: one question, one answer, one targeted probe, then feedback at the end.
• Score answers for relevance, actions, judgment, results, ownership, and clarity—not merely confidence or fluency.

An AI interview practice prompt is a written instruction that gives an AI model a role context, candidate evidence, task, constraints, and output format so it can simulate or evaluate a job-specific interview. The key is to use several small prompts in sequence rather than asking for an oversized list of questions. Clear context, explicit tasks, and defined outputs generally make prompts easier to refine when the first result is too vague, as explained in the OpenAI prompting guidance.

1. Build a competency map before writing questions

A job description is usually a mixture of responsibilities, requirements, preferred qualifications, tools, and broad claims such as “strong communicator.” Treat it as raw assessment material. Your first task is to translate those bullets into observable competencies: what the person must know, do, decide, or demonstrate.

This matters because job analysis connects job tasks with the competencies needed to perform them, forming a foundation for assessment and selection decisions, according to the U.S. Office of Personnel Management. A useful map might turn “improve onboarding conversion” into data analysis, experimentation, stakeholder management, customer judgment, and measurable delivery.

Do not give every bullet equal weight. Ask the AI to distinguish between essential entry requirements, differentiators, and background context. OPM guidance recommends focusing structured interview content on competencies that are critical at entry and that distinguish successful performers; for many roles, that means concentrating practice on roughly four to six high-value areas rather than covering every phrase in the posting OPM guidance.

Prompt 1: extract and rank competencies

You are an interview-design assistant. Analyze the job description below and create a competency map. Separate the findings into: 1. Core responsibilities 2. Required knowledge 3. Behaviors and ways of working 4. Tools, methods, or technical skills 5. Expected outcomes 6. Constraints, risks, or stakeholders Then rank the four to six competencies most critical for success at entry level. For each ranked competency, provide: - The exact job-description evidence - Why it matters in the role - What strong performance would look like - Two kinds of candidate evidence that could demonstrate it - Whether it is best tested with a behavioral, situational, technical, or mixed question Do not invent requirements that are not supported by the job description. Flag ambiguous wording. Job description: [PASTE JOB DESCRIPTION]

Review the output yourself. If the AI labels something “leadership,” ask what behavior supports that label. If it identifies a tool but no outcome, note that the tool may be a means rather than a competency. This check prevents you from practicing polished questions that are only loosely connected to the role.

2. Add your evidence before generating questions

The next input should be your evidence, kept separate from the employer’s wording. Include projects, responsibilities, decisions, obstacles, results, and lessons. You can anonymize company, client, product, or financial details with safe abstractions such as “a B2B software client” or “a regional operations team.” Do not paste confidential information simply to make a prompt feel more realistic.

For each competency, prepare two or three possible examples. One example may fit several competencies, but do not force it into every answer. A product launch can demonstrate prioritization, stakeholder management, and customer judgment; the strongest version depends on the question and the evidence you can explain precisely.

Use facts you can defend. If you do not know an exact percentage, say what changed qualitatively or provide a range only if you genuinely know it. AI should help you organize your experience, not supply missing proof. Official applicant guidance allows AI for mock interview preparation while warning candidates not to fabricate experience or submit AI-generated answers as their own evidence Civil Service Careers guidance.

Evidence template

• Competency: What capability does this example show?
• Context: What was happening, and why did it matter?
• Responsibility: What were you personally accountable for?
• Actions: What did you decide, change, communicate, or build?
• Result: What happened, and how do you know?
• Reflection: What would you repeat or change next time?
• Boundaries: What details must remain generalized or private?

Then use a second prompt to generate a balanced practice set from the ranked map and your evidence. Ask for behavioral questions about past action, situational questions about likely challenges, technical questions about the role’s methods, and follow-ups that test depth. Each question should identify the competency it assesses, but do not read that label aloud during mock practice if you want a more natural simulation.

Prompt 2: generate a traceable question set

Using the competency map and candidate evidence below, create a role-specific interview practice set. Generate: - Two behavioral questions for each priority competency - One situational question for each priority competency - Technical or role-specific questions where the job description supports them - One realistic follow-up for each main question For every question, show privately: - Target competency - Evidence from the job description - What a strong answer would need to demonstrate - One common weak-answer pattern Use open-ended, job-related wording. Avoid trivia, double-barreled questions, and generic questions that could fit any job. Do not assume the candidate has experience they did not provide. Competency map: [PASTE MAP] Candidate evidence: [PASTE YOUR EVIDENCE]

3. Run the mock interview one question at a time

A list of 20 questions is useful for planning, but it is not a mock interview. In practice, answer under a small amount of uncertainty, listen for the exact question, and decide what to include without seeing the next topic. Instruct the AI to act as interviewer only, ask one question, wait, and use a single targeted probe when your answer lacks evidence.

This sequence mirrors important features of structured interviewing: predetermined job-related questions, consistent probes, and common rating standards OPM structured interview guidance. It also makes weaknesses easier to diagnose. If you receive a full critique after every sentence, you may optimize for the model’s immediate approval rather than learn to complete an answer naturally.

Tell the AI not to rescue you. It should not rephrase the question repeatedly, supply an example, or turn a vague answer into a strong one. It can ask, “What did you personally do?” or “How did you measure the result?”—then wait. Set a limit of one probe per question so the session remains comparable across competencies.

Prompt 3: interviewer-only mode

Act as a structured interviewer for this role. Rules: - Ask one question at a time and wait for my answer. - Use the question set in order, but adapt only when a follow-up is required. - Ask no more than one targeted probe after each main answer. - Probe for my personal actions, decisions, trade-offs, technical reasoning, or results when missing. - Do not give hints, model answers, encouragement, or feedback during the interview. - Do not invent facts about my background or treat an unsupported claim as evidence. - Keep the tone professional and realistic. - After the final question, provide the scorecard requested below. Role and competency map: [PASTE MAP] Question set: [PASTE QUESTIONS] Begin with question 1 only.

Before starting, choose a practical session length and scope. For example, use six main questions: two behavioral, two situational, one technical, and one motivation or role-fit question. If you are preparing for a coding-heavy role, use a separate technical session rather than squeezing algorithm explanation into every behavioral answer. The related technical coding interview preparation plan can help you separate those practice modes.

4. Score proof, not performance theater

After the session, ask for a scorecard tied to the competency map. “Was this a good answer?” is too broad. A candidate can sound confident while failing to show ownership, judgment, or a result. Conversely, a quieter answer may contain strong evidence but need a clearer structure.

For each answer, require the AI to identify what was proved, what remains unsupported, and what single detail would most improve the response. OPM guidance emphasizes open-ended, job-related questions connected to competencies and defined proficiency benchmarks Structured Interview Guide. Your scorecard should therefore be role-specific, not a generic writing evaluation.

Prompt 4: post-answer scoring and retry

Evaluate each answer against the target competency, not against general eloquence. For every answer, provide: 1. Relevance to the job requirement 2. Specificity of my personal actions 3. Quality of technical or business judgment 4. Strength of the result or evidence 5. Ownership and accountability 6. Clarity and logical structure For each category, give a rating of weak, adequate, or strong and cite the exact evidence from my answer. Then provide: - The strongest proof point - The single most important missing proof point - Any claim that needs verification or softening - One concise retry question - Three revision notes, not a rewritten answer Do not add facts, metrics, responsibilities, or experience that I did not state. Assess the answer against this competency benchmark: [PASTE BENCHMARK] Candidate answer: [PASTE ANSWER]

Use the retry question immediately, but answer it in your own words. For example, if your response says, “I improved the process,” the useful probe is not “Can you make that more impressive?” It is “Which part did you change, what alternatives did you reject, and what result followed?” That question exposes whether the story contains an action and a defensible outcome.

Keep a simple error log after each session. Record the competency, the missing proof point, the example you used, and the next practice action. Patterns matter more than any single rating: repeated missing results suggest you need better measurement; repeated unclear ownership suggests you are describing team activity instead of your contribution; repeated technical vagueness suggests a knowledge gap rather than a storytelling problem.

Make the prompts safer and more realistic

Your prompts should constrain the AI as carefully as they direct it. State that the candidate evidence is authoritative, unsupported topics must be flagged, and uncertain job-description language must be marked for review. This reduces the temptation to accept a plausible but invented scenario as if it came from your background.

• Replace confidential names and sensitive figures with accurate abstractions before sharing material.
• Verify every generated question against the actual posting and remove questions based on invented requirements.
• Do not memorize an AI-written answer word for word; use feedback to revise your own structure and wording.
• Practice the same competency with different examples so you do not sound dependent on one story.
• Ask a person to review high-stakes answers when judgment, confidentiality, or technical accuracy is difficult to assess alone.

You can also vary the interviewer’s constraints without changing the evidence standard: ask for a skeptical stakeholder, a time-pressured hiring manager, or a technical panelist. Keep the underlying competency and rating criteria stable. The purpose is not to create dramatic role-play; it is to test whether you can communicate the same real experience clearly under different questioning styles.

Finally, prepare your opening and closing separately. The “Tell me about yourself” answer guide is useful for shaping a concise introduction, while a separate question-planning session can help you choose thoughtful questions for the interviewer. Do not let the mock-interview prompt turn every part of preparation into a score.

Conclusion: use AI to rehearse evidence, not manufacture it.