ChatGPT for MMI Practice: Prompts and Limits

ChatGPT for MMI Practice: Prompts and Limits

Use an AI tool to ask you questions and push follow-ups, never to write answers you memorize. On test day the rules flip: Acuity's terms for Casper say the use of AI "in any form when completing Assessments is strictly prohibited", and at least one medical school bans it in every interview format.

An AI tool is a patient interviewer that never runs out of "why?". That is the useful part. The harmful part is the polished model answer: it reads well, you memorize it, and it breaks on the second follow-up, because you never did the reasoning. This page gives four prompts that keep the tool in the interviewer's chair, a constructed practice transcript, and the rules for the day itself. It sits inside our medical and health application essays hub.

Every Acuity, AMA and school page quoted below was checked on November 7, 2026. The AAMC's pages, including PREview, returned HTTP 403 to our fetch that day, so nothing from them is quoted here.

The rules: practice versus test day

Practice before the day is a different question. An AMA article from October 30, 2026 on residency interviews quotes Ankita Bhatia, who led AI strategy at the AMA: "You can ask AI to interview you and get unlimited mock-interview reps," and "This is a high-value use of the technology with little risk." The article is written for residency applicants; we apply the same logic to a medical school MMI: the tool asks, you answer.

If a school publishes no AI rule for its interview, the certification inside the application is what you sign, so read it there.

What AI practice is good for, and what it is not

Good for:

• Follow-up chains. The tool can ask "why?" five times in a row, so you find where your reasoning runs out.
• Volume and variety. New stations on ethics, teamwork, policy and personal questions, as many as you want.
• Timing. You can ask it to stop you at a set time.
• Transcript critique. Paste your own spoken answer, typed out, and ask what was vague.

Not for:

• Writing your answers. Anything you memorize from it is someone else's reasoning.
• Delivery. It reads your words. It does not see your eye contact or hear your pace. Record yourself, or practice with a person.
• Facts about schools. The AMA's September 9, 2026 article on AI in residency applications quotes Bhatia saying "A chatbot's memory is not a source". Get a school's interview format from the school.

Four prompts that keep it in the interviewer's chair

Four constructed prompts, written by us. Each one says what the tool must not do, because the default is to help by writing.

1. The interviewer prompt

You are a medical school MMI interviewer. Give me one station scenario at a time, read it once, then wait for my answer. Do not give example answers, and do not tell me what a good answer would include. After I answer, ask me one follow-up question that challenges the weakest part of my reasoning.

2. The follow-up prompt

Ask "why?" or "what if?" about my last answer, five times in a row, one question at a time. Each question must build on what I just said. Do not summarize or praise my answer.

3. The critique prompt

Here is the transcript of my spoken answer. List the three vaguest phrases and, for each, ask me a question that would force a specific answer. Do not rewrite my answer.

4. The timing prompt

Give me a station. I have two minutes to read and six minutes to answer. When I paste my answer, tell me only whether I got to a decision, and how many sentences it took me to get there.

Set the station time to the school's format. Our page on virtual medical school interview and MMI tips covers formats schools publish.

Keep patient details out of your prompts. If a practice story comes from a real clinical shift, change the details before you type it into any tool.

Constructed practice transcript: a follow-up chain

Constructed example, written by us, not a real applicant's practice session.

Tool: A patient asks you, a medical student, whether her surgeon is good. You have heard residents complain about him. What do you do?

Applicant: I would tell her I am a student and can't judge surgeons, and suggest she ask the team.

Tool: Why not tell her what you heard?

Applicant: It is hearsay. I don't know if it is true, and it could scare her before surgery.

Tool: What if what you heard was about a patient being harmed?

Applicant: Then it is not about her question anymore. It is about safety, and I would take it to my attending or the patient-safety reporting system.

Tool: So would you tell her anything different?

Applicant: No. I would still answer her honestly about my role, and make sure she has a chance to ask the surgeon her questions directly.

What the chain shows: the first answer was safe and thin. The third follow-up forced a split between gossip and a safety concern, which is the real point of the station. A memorized model answer would have stopped after the first exchange. Run the same chain again a day later and see whether you reach the split without being pushed.

Why memorized answers backfire

Acuity, which makes Casper, puts it directly: "there is no need to memorize content or purchase preparation courses." Its prep page adds that test takers who use the official practice test "tend to do better on Casper than those who use third-party preparation materials or don't prepare at all."

A memorized answer has three problems in a live interview. It is built for the question you expected, not the one you got. It has no second layer, so the follow-up finds the edge fast. And it sounds recited, which is the opposite of what an interviewer wants from a question about your judgment.

Our view: memorize facts, not sentences. Know the three or four stories you will draw on, the dates and names in them, and what each one taught you. Then let the tool make you say them differently every time.

The countercase: memorize your first sentence for "Tell me about yourself" and "Why medicine?" Those two openers are predictable, and a clean first sentence buys you time. Everything after it should be live.

Casper and PREview: no AI on the day

Casper is the clearest case. Acuity's terms prohibit AI "in any form when completing Assessments", and its terms say the Casper Admission Assessment rating is human rated. Its technical requirements page tells test takers to "Turn off browser extensions" before the test.

For practice, Acuity points you to its own test: "You can complete the practice test as many times as you'd like. Because it is a practice tool, your responses are not saved, scored, or reviewed." Use the AI prompts above for open practice, and Acuity's practice test for the real format. Our guide to Casper, PREview and the MMI maps one answer framework across all three.

For PREview, read the AAMC's rules in your account; we could not open them today.

A one-week practice plan

For questions about AI as a topic, see med school interview questions on AI in medicine. If an interviewer asks whether AI helped with your essays, see how to answer when an interviewer asks if you used AI.

Questions about AI and MMI practice

Can I use an AI tool to prepare for a medical school interview?

For practice before the day, none of the pages quoted here forbids it, and the AMA article above recommends it for mock interviews. During the interview, follow the school's rule; Hackensack Meridian prohibits generative AI in all its interview formats.

Can I use an AI tool to practice for Casper?

Practice with it before the day, and take Acuity's free practice test for the real format. During Casper, Acuity's terms prohibit AI in any form.

Is it a red flag to sound rehearsed?

A rehearsed opener is fine. A rehearsed reasoning chain is the problem, because it does not bend when the interviewer changes the facts.

Should I ask the tool what interviewers want to hear?

No. It will produce a confident general answer that is not tied to any school. Ask it to find the weakest part of your own answer instead.

Review the writing you send next

The practice ends at the interview; the writing does not. Your next document is an update letter, so get a free Quick Review of your update letter against GradPilot's update letter rubric, which reads whether the letter names information added since you applied and gives the reader a clear account of it. Send it once you have something new to report; our interview thank-you email examples cover the note you send first.

Sources

Retrieved November 7, 2026, except where noted.

• Acuity Insights, Terms of Use (no year stated): AI prohibition, third-party assistance, human rating of Casper.
• Acuity Insights, About Casper (no year stated): no need to memorize content.
• Acuity Insights, How to Prepare for Casper (no year stated): practice test and comparison with third-party materials.
• Acuity Insights, Tech Requirements for Casper (no year stated): browser extensions.
• Hackensack Meridian School of Medicine, Admission Process and Procedures (no year stated): interview AI rule.
• American Medical Association, Answering the machine: How AI can assist residency-interview prep (October 30, 2026; public portion only, the rest requires sign-in): mock-interview quote.
• American Medical Association, How to use AI in residency applications (September 9, 2026): "A chatbot's memory is not a source".
• AAMC PREview pages: not retrieved (HTTP 403 on November 7, 2026); nothing quoted.