Practice one answer at a time with evidence-based feedback, follow-ups, and a final improvement pattern
8 min setupTested with:ChatGPTReviewed: 2026-08-28
1
Add your context
Your text stays in this browser. AILesson Prompts does not send it to a model or server.
2
Your prompt
Unfilled fields remain visible as placeholders, so you can still copy and edit the prompt
Run an interactive mock interview one question at a time.
Role, stage, format, panel, duration, locale, and verified process details:
[role]
Verified experience, skills, gaps, examples, constraints, and target areas:
[candidate]
Question count/type, answer time, feedback depth, criteria, challenge, and stop rule:
[method]
Ask only the first question, then wait for the candidate's answer. After each answer: summarize the claim made; distinguish verified evidence from unsupported or vague language using the fact bank; score only the supplied criteria with concrete rationale; identify one strength and at most three priority improvements; ask one realistic follow-up; then offer a revised answer only after the candidate attempts the follow-up or requests it. Do not fabricate an employer script, simulate discriminatory questions, shame the candidate, infer personality, or reward invented confidence. Track repeated patterns across answers. At the end return evidence coverage, improvement trend, stories overused or missing, accuracy risks, and a short next-practice plan.
Private by defaultPrompt assembly happens locally in your browser. Avoid placing confidential information into any AI service unless your organization allows it.
From input to outcome
A worked example
See how concrete context turns this recipe into a usable result
Actual input
Role and interview context
First-stage 45-minute panel for Operations Analyst. Responsibilities: capacity reporting, variance investigation, manager partnership, process improvement. Required Excel and writing; preferred SQL/Power BI. Actual panel questions unknown.
Candidate fact bank
Retail supervisor since Sep 2023. Verified stories: Excel staffing workbook for 45 employees with 12→5 monthly correction observation and concurrent staffing change; stock-variance investigation with three managers and adopted checklist; weekly trading notes. No SQL/Power BI, director presentation, or causal proof.
Practice method
Run 4 questions: behavioral, analytical, stakeholder, gap/motivation. Candidate gets 90 seconds each. After each, score 1–5 for relevance, specific personal action, evidence/accuracy, structure, and concise delivery. Medium challenge. Stop after four or when candidate says stop. First show how the initial turn and one sample feedback cycle should work.
Example output
First turn only:
“Question 1 of 4 — behavioural/process improvement: Tell me about a time you improved an operational process. You have up to 90 seconds. Please answer as you would in the interview; I will not provide a model answer first.”
Example feedback after a hypothetical candidate reply claiming “I led a project that cut errors by 58%”: Claim summary: you identified rota-correction problems, built a workbook, introduced controls and training, and linked it to lower correction counts. Verified: personal workbook work, tools, training two supervisors, and 12→5 averages over consecutive three-month periods. Accuracy risk: “led a project,” “errors,” “58%,” and causal “cut” are not supported; concurrent staffing change matters.
Scores: relevance 5; personal action 4 because decisions need more detail; evidence/accuracy 2 due to overclaim; structure 3; concision 4. Strength: concrete scale and tool use. Priorities: replace causal percentage with observed counts/periods; explain the duplicate-ID check or user feedback; state the concurrent factor. Follow-up: “What did your initial test reveal, and what did you change before wider use?” Wait for the candidate response before offering a revision or Question 2. Track “causal overclaim” as a pattern only if it recurs.
Why this works
1
One-question turns preserve retrieval practice and stop model-written answers from replacing candidate thinking.
2
Fact-bank comparison catches overclaiming while feedback is still tied to the candidate's own wording.
Check the result
Does the flow wait for one candidate answer before feedback and the next question?
Are scores and feedback tied to stated criteria, exact wording, and verified facts?
Does the process improve truthful evidence delivery rather than supply fabricated ideal answers?
ChatGPT