Step 1: Collect role-specific interview questions (and prompts)
Start with high-quality question banks that reflect real hiring processes. Use trusted sources to guide the GPT’s knowledge.
Use these resources:
- Glassdoor Interview Questions: search by company and role
- IGotAnOffer: great for consulting & strategy roles
- Levels.fyi: for big tech interview questions
- LinkedIn’s Interview Preparation Tool: curated Q&A by recruiters
Compile 10–20 questions from your target role and note what type they are (e.g. behavioral, technical, situational) into a document. You can also use this 100+ interview questions as a sample document to train your Custom GPT.
These will form the backbone of your GPT’s questioning strategy.
Step 2: Go to ChatGPT → Explore GPTs → Create
Click “Create a GPT” → Enter a name like:
🎤 “Mock Interviewer for Product Managers”
💼 “AI Interview Coach for Marketing Analysts”
In the Instructions field, paste a short role description like:
You are an experienced professional interviewer specializing in [insert role, e.g. “Product Management” or “Marketing Analyst”]. Your job is to run realistic, structured mock interviews to help candidates prepare.
You begin by introducing the interview and setting expectations (e.g., number of questions, format, how feedback will work).
You then ask one question at a time — waiting for the user to respond before continuing. Your questions should reflect real hiring practices, drawing from behavioral, situational, and role-specific categories.
After each user response, provide:
- Clear, constructive feedback on their answer
- Highlight what they did well (clarity, relevance, structure, tone)
- Suggest improvements (missing details, stronger examples, tighter structure)
Use frameworks like STAR (Situation, Task, Action, Result) for behavioral questions, and look for specificity and impact in answers.
Prioritize a calm, professional tone — supportive but honest. If the user asks for follow-up tips or wants to retry a question, support them.
Customize your questions based on any uploaded job description, resume, or candidate profile. Adjust question difficulty and focus area based on role seniority if that’s available.
Ask no more than 6–8 questions per session. Wrap up with a summary of overall strengths and areas to work on.
Step 3: Add personality and pacing rules
Make the interview feel real by setting tone and structure:
Use this sample custom instruction:
Start by introducing the interview. Then ask one question at a time. Wait for the user to answer before continuing. Give concise, constructive feedback after each response, including what went well and what could be improved. Use a calm, professional tone.
Optional: Upload your job description or resume to customize even further.
Step 4: Feed it high-quality examples and feedback styles
To help your GPT “learn” how great answers sound, provide samples.
Upload or paste:
- STAR-format answers (Situation, Task, Action, Result)
- Weak vs. strong answers to common questions
- Notes on what top candidates do well (clarity, structure, specificity)
You can even train it with:
- McKinsey’s PEI structure
- Amazon’s “Leadership Principles” behavioral expectations
- Google’s “GCA” (general cognitive ability) approach
🧠 This primes your GPT to give better feedback.
Step 5: Test and iterate with real scenarios
Try a full mock interview session yourself. Example prompt:
“Let’s do a mock interview for the role of Growth Marketing Manager at a Series A startup. Ask me 5 behavioral and 2 situational questions. Give me feedback after each.”
Tips for iteration:
- Add follow-up probing if the answers are too shallow
- Refine the feedback tone if it feels too generic
- Save “preset” versions for different industries (tech, consulting, education, etc.)
Bonus use cases:
- Students can practice for internships
- Career switchers can build domain fluency
- Teams can prep for internal promotions or panels
You just created a 24/7 interview coach! No scheduling or awkward Zooms, now you can have fast, focused, personalized feedback from your own AI.
