Verified against ChatGPT · 2026-08-14
Run a mock interview that actually interrupts, follows up, and pushes back like a real interviewer would
Runs a live, adaptive mock interview session that asks follow-up questions based on your actual answers and pushes back on vague ones, instead of just reciting a fixed question list one after another.
The prompt
Ready to copy — highlighted parts are example details you can swap.
Run a mock interview with me for a specific role. I don't want a static list of questions read out one after another — I want you to act as the actual interviewer, listening to my answers and reacting to them the way a real person would, including following up and pushing back when an answer is vague. ROLE AND STAGE Engineering Manager role, first-round hiring manager interview, 30 minutes. INTERVIEWER PERSONA A direct, no-small-talk VP of Engineering who cares most about how I handle underperforming reports, and who asks pointed follow-ups rather than letting vague answers slide. WHAT I WANT THIS SESSION TO STRESS-TEST I want to be pushed hardest on how I actually handle a direct report who is missing deadlines, since that's the area I feel least rehearsed in. HOW TO GIVE FEEDBACK Only at the end — I want the full session to feel like a real interview without breaks for coaching in the middle. Run this as a real back-and-forth, one question at a time — never dump multiple questions in one message. Take on the interviewer persona I described and stay in character for the questions themselves. After each answer I give, before moving to your next question, decide whether the answer was specific and complete enough to move on, or vague/incomplete enough that a real interviewer in this persona would follow up — if it needs a follow-up, ask it in character rather than breaking to give feedback first, the same way a real interview would actually unfold. Start with a first question appropriate to the stage and persona, then wait for my answer before doing anything else. WHAT NOT TO DO Do not soften your follow-ups to spare my feelings mid-interview — a real interviewer wouldn't, and the value of this exercise is finding out where my answers actually don't hold up under a real follow-up, not feeling good about the session. Do not move to a new topic just because I gave a decent answer if there's an obvious, specific follow-up a real interviewer in this persona would ask. Only break character to give feedback based on what I said about feedback timing above — if I asked for end-of-session feedback, stay fully in character through every question and follow-up, and only switch to a feedback voice when I explicitly say the interview is over.
Customize
Optional — swap in your own details for the highlighted parts above.
Why this works
A static list of questions read one after another fails to simulate the actual pressure point of a real interview, which is rarely the first question itself but the follow-up that comes after a vague or incomplete answer — a candidate can rehearse a polished response to every anticipated question on a list and still be caught flat-footed by "can you say more about what specifically you did there" in a live conversation, because that follow-up only exists in response to what was actually said, not to a question written in advance. Explicitly instructing the model to evaluate each answer before deciding whether to advance or follow up forces it to actually behave adaptively turn-by-turn rather than defaulting to its base pattern of working through a pre-planned list regardless of answer quality, which is the single biggest gap between a scripted Q&A and something that resembles a real interview's actual information-seeking behavior. The instruction to stay in persona through follow-ups rather than breaking to coach mid-session matters mechanically because GPT-5.1's default helpful-assistant instinct is to interject encouragement or gentle correction as soon as it notices a weak answer, which is exactly the behavior a real interviewer doesn't exhibit and exactly the behavior that would let the candidate mentally relax between questions instead of staying under the pressure the exercise is meant to simulate. Explicitly forbidding the model from softening follow-ups to spare feelings addresses a specific and well-documented tendency of instruction-tuned models toward excessive agreeableness — without that constraint stated up front, the model will tend to accept a mediocre answer and move on rather than push, which defeats the entire purpose of a stress-test session and would leave the candidate with false confidence walking into the real interview.
What you get back
Interviewer: "Walk me through the last time a direct report was consistently missing deadlines." [Candidate answers vaguely] Interviewer, in character: "You said you 'had a conversation' with them — what did you actually say, and what changed afterward that told you it worked?"
Verified against
ChatGPT GPT-5.1 · 2026-08-14
Changelog
- 2026-08-14 — Initial publish, verified against ChatGPT GPT-5.1.
Need this built into your business?
If a prompt isn't enough — what Scult builds, built and maintained for you — that's Scult's day job.
EXPLORE WHAT SCULT BUILDS
