Module 04 · Interview Preparation With AI
Building a Mock Interviewer With ChatGPT or Claude
Open lesson + course map
On this lesson
Course outline
Module 1 · Your AI Job-Search Toolkit
Module 2 · Resume Engineering
Module 3 · LinkedIn as a Lead Engine
Module 4 · Interview Preparation With AI
A mock-interviewer assistant turns a job post and your truthful resume into a structured practice conversation. It asks one question at a time, probes vague answers, and scores the evidence you actually gave—not the confidence of your delivery.
By the end, you will have a reusable interviewer setup and feedback notes from one completed practice round. The assistant can create useful pressure, but it cannot know an employer's private question list or decide whether an answer is true.
// concept
Build the Interview Pack Before the Chat
Prepare two redacted text blocks: the complete job post and the resume version you would send. Remove your CNIC, address, phone, personal email, references' details, signatures, and confidential employer or client information. Keep only claims you can verify.
Choose the session mix instead of asking for “some interview questions.” This sample 8-question mix is a practice design, not a prediction:
| Question type | Count | What it tests |
|---|---|---|
| Opening and motivation | 1 | A concise, role-specific introduction |
| Behavioral | 3 | Past evidence about teamwork, ownership, and setbacks |
| Role-specific | 2 | Skills named in the job post |
| Situational | 1 | Reasoning through a plausible work scenario |
| Candidate question | 1 | Whether you can investigate the role thoughtfully |
Select a persona that changes the questioning style without pretending to be a real person: for example, “a skeptical marketing manager hiring an entry-level performance marketer.” Never request a named recruiter impersonation or inside knowledge.
// concept
Configure the Reusable Interviewer
Save the following as mock-interviewer-v1.md. Paste it first in a normal ChatGPT or Claude chat, followed by your redacted inputs. A reusable GPT or Project is optional; availability and plan gates change, so confirm the official help page. A local Markdown file plus a fresh chat is the free alternative.
You are my rigorous mock interviewer for the role described below.
INTERVIEWER PERSONA
- Act as a skeptical but fair hiring manager for this function.
- Use only the job post and resume inside the source tags.
- Never claim to know the employer's hidden process or likely exact questions.
SESSION DESIGN
- Run 8 questions: 1 opening/motivation, 3 behavioral, 2 role-specific,
1 situational, and 1 question I should ask the employer.
- Ask ONE question at a time. Wait for my answer before continuing.
- Do not show a model answer before I respond.
PUSH-BACK RULES
- If my answer lacks a concrete action, ask: “What did you personally do?”
- If I use “we” without defining my contribution, ask: “What was your part?”
- If I make an unsupported result claim, ask: “How can you verify that?”
- If the answer is vague, ask one targeted “tell me more” follow-up.
- Ask no more than two follow-ups per main question.
- Never invent experience or improve an answer by adding facts.
SCORING
After each answer, privately record a 0–4 score for Evidence, Relevance,
Action specificity, Structure, and Delivery clarity. Do not interrupt the
interview with coaching unless I ask to pause.
AFTER THE FINAL QUESTION
Return a table with: question, score for each criterion, strongest evidence,
missing detail, and one specific retry instruction. Then give:
1) two answers to repeat now; 2) three recurring patterns; 3) a five-line
practice plan. Flag any statement I should verify before a real interview.
<job_post>
[PASTE REDACTED JOB POST]
</job_post>
<resume>
[PASTE REDACTED, TRUTHFUL RESUME]
</resume>When the assistant confirms the setup, type: Begin the interview. Do not coach me until the final debrief. Answer aloud where possible and time only the answer—not your thinking pause. A useful practice boundary is 60–90 seconds for a behavioral answer, but clarity matters more than forcing every story into the same length.
// concept
Make the Scoring Useful
Use a 0–4 scale: 0 = absent, 1 = weak, 2 = partial, 3 = solid, 4 = strong. The score is diagnostic, not an employer prediction.
| Criterion | Weight | A score of 4 requires |
|---|---|---|
| Evidence | 30% | A specific, truthful example with verifiable details |
| Relevance | 25% | Direct connection to the question and role requirement |
| Action specificity | 20% | Your decisions and actions are distinguishable from the team's |
| Structure | 15% | A clear beginning, action sequence, and outcome |
| Delivery clarity | 10% | Plain wording, controlled pace, no distracting detours |
Calculate a score out of 100 with sum(score ÷ 4 × weight). Keep component scores because weak evidence and rambling delivery need different fixes. Mark unverified facts VERIFY; never reward polished uncertainty.
// worked_example
Worked Example
This hypothetical sample follows an Islamabad graduate applying to a junior operations role. The vacancy asks for spreadsheet discipline and coordination. Their resume says: “Coordinated a four-person university event team; maintained the task sheet and vendor follow-ups.”
After the full setup prompt, the assistant asks: “Tell me about a time a deadline was at risk. What did you do?”
Sample draft answer: “We had a university event and a vendor was late. I communicated with everyone and we managed it successfully.”
The interviewer pushes back: “What did you personally do, and what evidence shows the deadline was protected?” That follow-up is appropriate because “communicated” hides the action and “successfully” is undefined.
Sample corrected answer: “Two days before our university event, one vendor had not confirmed delivery. I marked the item at risk in our shared sheet, called the vendor for a written delivery time, and told the event lead which setup task would be delayed. I then moved two volunteers to the tasks we could complete first. The vendor arrived before our scheduled setup ended. I can verify my role from the task sheet and message thread; I would not share those private messages with an interviewer.”
Sample feedback note: Evidence 3/4; Relevance 4/4; Action specificity 4/4; Structure 3/4; Delivery clarity 3/4. Retry instruction: state the original deadline in the first sentence and replace “tasks we could complete” with one non-confidential example. The assistant fixed the real weakness without manufacturing a percentage, budget, or employer outcome.
// failure_cases
Failure Cases to Diagnose
6 cases to diagnose
Question dump
The model prints all eight questions together. Reinforce “one question at a time; wait for my answer” and restart the round.
Instant coaching
It rewrites each answer before you feel interview pressure. Add “record feedback privately and debrief only after the final question.”
Soft follow-ups
It accepts “we handled it” without probing ownership. Keep the explicit “What did you personally do?” rule and allow up to two targeted follow-ups.
Resume fan fiction
Feedback adds a metric, tool, or responsibility absent from your resume. Delete the invented detail and instruct it to label missing facts as questions for you.
Score inflation
Every answer receives 4/4. Ask for one quoted piece of evidence supporting each score and require the 0–4 definitions.
Predicted-company theatre
It says the employer “will ask” a question. Reframe every question as practice derived from the public job post, never a forecast.
// pakistan_angle
Pakistan Angle
For a Pakistani hiring process, test only details in the vacancy: onsite work in Lahore, Karachi, or Islamabad; notice period; shift timing; or remote-team overlap. Do not infer that a woman can work late shifts, every graduate can relocate, or fluent English is required. Practise English and, when relevant, natural Urdu or Roman Urdu while keeping job terms accurate.
On a shared family phone or low-cost Android, keep the redacted pack locally so a dropped connection does not erase it. Never upload CNIC scans, salary slips, addresses, or references' numbers. Voice mode needs microphone permission and stable connectivity; in a noisy hostel or during load-shedding, use headphones or a text round with an offline recorder. Delete sensitive recordings after review.
// hands_on
Hands-On Exercise
6 steps
Build and test your saved interviewer today:
Choose one live vacancy you may apply for and save its full text with the URL and date accessed.
Make a redacted copy of your submitted resume; check every remaining statement against your records.
Paste the system-style prompt, job post, and resume into a fresh ChatGPT or Claude conversation.
Complete one four-question mini-round: one motivation, one behavioral, one role-specific, and one situational question. Require push-back where evidence is thin.
Save the final feedback table in
company-role-mock-YYYY-MM-DD.md. Under it, write your corrected answer—not an AI-invented replacement—for the two lowest-scoring questions.Finish by listing one delivery habit and one evidence gap to practise before the real interview. Done means you have the reusable prompt, a completed feedback record, component scores, two truthful revised answers, and no private identifiers in the saved chat or notes.
// completion_rubric
Completion Rubric
6 checks — tick as you verify
// sources
Sources
// check_yourself
Check yourself
4 questions · answers and options are taken word-for-word from this course
1 / 4 · diagnose
Your work shows this failure mode: “Instant coaching.” The lesson describes it like this: “It rewrites each answer before you feel interview pressure.” What does the lesson tell you to do about it?