Confetto

An AI Driven Medical School Interview With ChatGPT

Practice the med school interview now

Free, no sign-up to start.

What happened when Dr. Sahil Mehta tested ChatGPT in a medical school interview, why it earned a solid B, and how to use AI to practice for real interviews.

Practice the med school interview now

Free, no sign-up to start.

In a text-based mock medical school interview conducted by attending physician Dr. Sahil Mehta and reported by AAMCNews in March 2023, ChatGPT earned a solid B. It gave strong answers on medical ethics and delivering bad news, but failed on personal questions because it had no college experiences or personal motivations. You can use ChatGPT as an interviewer to drill questions, while keeping your answers rooted in your own experiences. Sources: AAMC: Preparing for Your Medical School Interview. Last checked November 11, 2026.

What happened when ChatGPT took a medical school interview

When people search for an artificial intelligence driven medical school interview with ChatGPT, they are usually looking for two things. First is the widely cited experiment by Dr. Sahil Mehta, an attending physician in radiology at Beth Israel Deaconess Medical Center in Boston and founder of MedSchoolCoach, which AAMCNews reported on March 23, 2023. Second is how to use ChatGPT to practice for their own interviews.

Dr. Mehta interviewed ChatGPT by text to see if the chatbot could pass a medical school admission interview. He gave the tool a solid B. According to Dr. Mehta, ChatGPT tapped into a repository of medical knowledge far better than any premedical student could have, but fell apart on personal experiences, which he called an important part of the medical school interview.

If you want to practice out loud with an AI tool built specifically for healthcare interviews, Confetto runs live conversational mock interviews for MMI, traditional, and panel formats with follow-up questions and rubric feedback. Start your free week to test your answers before interview day.

How ChatGPT performed as an applicant in Dr. Sahil Mehta's interview experiment (Source: AAMCNews, March 2023, plus prospective commentary in PMC12826817 and JMIR Med Educ 2025. 1:e63400 on AI simulation limits in medical education).
Question categoryTopic testedHow ChatGPT responded
Healthcare technologyThe role of artificial intelligence in medicineDelivered informed answers drawing on its medical knowledge base
Patient communicationGiving a patient bad newsGave a structured answer, including step five: offer support and resources
Medical ethicsDetermining which of two patients gets a donated liverApplied ethical reasoning to weigh the two patients
Personal reflectionCollege experiences, personal ambitions, and motivation to practice medicineFailed: stated I do not attend college and I do not have personal preferences or desires

Why memorizing chatbot answers backfires in real interviews

ChatGPT can write clean, organized paragraphs about precision health or clinical ethics, and applicants on the Student Doctor Network (SDN) have even prompted chatbots to invent experiences as a hospital intern or free clinic volunteer. Relying on chatbot prose for your application or interview prep creates two problems on interview day.

First, it sounds rehearsed. J. Michael Bostwick, MD, senior associate dean of admissions at Mayo Clinic Alix School of Medicine, told AAMCNews that chatbot prose can make a student sound scripted when committees are looking for what makes applicants stand out and showcase their authentic selves. Dr. Bostwick also noted that if you use a chatbot to generate answers, you might not have thought the problem through yourself.

Second, admissions committees use the live interview to check whether you match your written file. Lindia Willies-Jacobo, MD, FAAP, senior associate dean for admissions and equity, inclusion, and diversity at Kaiser Permanente Bernard J. Tyson School of Medicine, explained that interviewers watch whether the person on paper appears consistent with the person in front of them. The giveaway is how you speak, think out loud, and respond to unexpected questions.

When you practice for your own medical school interview questions, treat AI as the questioner, never the scriptwriter. If you memorize a chatbot's phrasing, your delivery will flatten out the moment a faculty interviewer interrupts with an unscripted follow-up.

How to set up ChatGPT as your mock interviewer (our step-by-step planAILS/Voice Mode & Competencies Rubric), wait

Instead of asking ChatGPT how to answer a question, instruct it to act as the interviewer and push back on your answers. Copy and paste the system prompts below (written by our team) into ChatGPT Advanced Voice Mode (two-way audio conversations) to simulate spoken pressure, interruptions, and real-time pacing.

Before you paste anything from your application or clinical notes into a public AI tool, strip out all identifying details. The AAMC advises against entering identifiable applicant information into personal web-based AI tools because those tools often do not keep data private, and medical school clinical exercises stress never giving an AI tool identifying patient information.

  • Step 1: De-identify your stories. Write a short list of three to five clinical, volunteering, or leadership experiences using only generic labels (for example, a rural clinic patient in her 60s, or a college lab partner) with no names, dates, or hospital identifiers.
  • Step 2: Set the interviewer role in Advanced Voice Mode for a 1:1 interview. Paste this system prompt before starting two-way audio: You are a medical school faculty interviewer conducting a 20-minute one-on-one interview. Ask me one question at a time about my motivation for medicine and my experiences. Do not write example answers for me. Interrupt me if I ramble past two minutes, and ask one unexpected follow-up question based on a specific detail I just said.
  • Step 3: Run a timed MMI station with audio pressure. Paste this prompt: Give me one ethical or teamwork scenario for a Multiple Mini Interview station. Wait while I take 2 minutes to think, then let me speak for 5 minutes. Challenge my decision with one counterargument before giving feedback.
  • Step 4: Grade the transcript against AAMC Core Competencies and answer frameworks. After finishing your voice session, paste this prompt into the text transcript: Grade my spoken transcript against the AAMC Core Competencies (including cultural competence, ethical responsibility to self and others, resilience, oral communication, and empathy) and evaluate my structure against proven frameworks like STAR (for behavioral stories) and SPIKES (for delivering bad news). Point out where I sounded vague or scripted, and tell me which part of my story needed a concrete action.

What ChatGPT cannot check: the real interview formats you will face

According to the AAMC's Preparing for Your Medical School Interview guide, an interview is your chance to show empathy, self-awareness, emotional intelligence, communication ability, and interpersonal skills. Text-based and audio chatbots cannot see your camera framing, eye contact, posture, or facial expressions. You need to match your practice to the exact format on your invitation.

The AAMC highlights three common formats across medical schools, and recommends that medical schools conduct interviews virtually and offer an option to visit after acceptance when feasible. In an AAMC survey of medical school applicants, 90% preferred a virtual or optional virtual format, 96% agreed or strongly agreed that virtual interviews saved them money, and 85% named reduced financial impact as the top advantage.

  • 1:1 interviews. A faculty member or local clinician meets with you individually for 20 or 30 minutes to assess your oral communication skills and explore the experiences and personal statement in your application.
  • Multiple Mini Interviews (MMI). Typically 6 to 10 stations, each focused on a different question or scenario, where you have 5 minutes to answer the question or solve the problem. The MMI specifically assesses verbal and nonverbal communication skills that written exams and transcripts cannot measure. You can practice timing with our MMI practice questions with answers.
  • Group interviews. Your interview cohort is given a task or problem and a set amount of time to solve it together, allowing schools with small-group and interprofessional curricula to watch your teamwork and problem-solving skills.
  • Recorded video assessments. Some schools include asynchronous video stages on platforms like Kira Talent or Acuity Insights. Record yourself on your webcam during practice so you can check your nonverbal delivery and eye contact.

How to answer when interviewers ask you about AI in medicine

Just as Dr. Mehta asked ChatGPT about the role of AI in medicine, real interviewers ask applicants how they view artificial intelligence in clinical care and medical education. According to the 2023-2024 Curriculum SCOPE Survey of MD- and DO-granting schools in the United States and Canada cited by AAMCNews in May 2025, 77% of responding schools covered AI in their curricula. Studies cited in the same report found that as many as 86% of undergraduates have used AI for schoolwork.

Medical schools are already teaching students how to use these tools alongside clinical judgment. If an interviewer asks for your view on AI in healthcare, ground your answer in how medical schools actually use it and the guardrails faculty emphasize:

  • Clinical reasoning and differential diagnosis. At the George Washington School of Medicine and Health Sciences (GW SMHS), third- and fourth-year students in a two-week AI elective use AI on fictional cases to ask for differential diagnoses, lab or imaging orders, and treatment recommendations, then assess the AI's accuracy. At the University of Virginia (UVA) School of Medicine, students use AI in foundational clinical-reasoning exercises while protecting patient privacy.
  • Simulated patient encounters. At the University of Texas Health Science Center at San Antonio and the University of Minnesota Medical School, students practice clinical visits on computers with AI-generated standardized patients. At the University of Cincinnati College of Medicine, transcripts of AI standardized patient visits showed that students who diagnose cases correctly tend to ask questions throughout the entire encounter, while students who miss the diagnosis tend to jump to ordering labs or physical exams and ask questions at the end.
  • Keeping the human in the loop. Across these programs, faculty stress that AI is one factor in decision-making, not an authority. As Ronald Rodriguez, MD, PhD, at UT Health San Antonio told AAMCNews, the human is the final arbiter of the ground truth.

How medical schools and the AAMC use AI in admissions

Show details

Applicants often wonder whether an artificial intelligence algorithm is grading their application or interview behind the scenes. The AAMC states clearly that while AI can help organize information and improve steps in the medical school and residency application process, it does not replace human reviewers, program leaders, and experts making decisions about applicants. All core AAMC admissions services, including AMCAS, the MCAT exam, and the AAMC PREview program, are designed and operated by people.

In February 2024, the AAMC convened a seven-person technical advisory committee with expertise in data science, industrial and organizational psychology, and residency selection. The group met for six months to create the Principles for Responsible AI in Medical School and Residency Selection.

Under these principles, institutions may use machine learning to help prioritize applications for review, natural language processing to screen documents in a standardized way, or large language models to improve drafts of interview protocols that capture competencies important to the school. The AAMC also has research and development projects exploring advanced analytics to improve predictions about where applicants will be invited to interview. Every use case is governed by six AAMC principles:

  • Balance Prediction and Understanding: Ensure AI tools deliver insights that improve prediction and efficiency while remaining comprehensible and usable by the institution.
  • Protect Against Algorithmic Bias: Rigorously assess and manage biases arising from historical data to ensure fair processes and outcomes.
  • Provide Notice and Explanation: Maintain transparency by informing applicants how AI is used and how it affects the assessment of their application.
  • Protect Data Privacy: Safeguard applicant information and maintain confidentiality at every step.
  • Incorporate Human Judgment: Balance technology with human judgment, insights, and ethical standards.
  • Monitor and Evaluate: Assess the outputs and outcomes of the AI system so they remain fair, accurate, and aligned with institutional goals.

How AI is used in letters of recommendation, MSPEs, and residency tools

Show details

The AAMC also publishes specific rules on how faculty and medical schools use AI across the medical education continuum. In addition to its selection principles, the AAMC released seven Principles for the Responsible Use of Artificial Intelligence in and for Medical Education. Those seven principles are: Maintain Human-Centered Focus, Ensure Ethical and Transparent Use, Provide Equal Access to AI, Foster Education, Training, and Continuing Professional Development, Develop Curricula Through Interdisciplinary Collaboration, Protect Data Privacy, and Monitor and Evaluate.

For letters of recommendation (LORs), the AAMC permits letter writers to use generative AI to support drafting and editing, provided the author applies four safeguards: human oversight (relying on firsthand observation, not AI, to form the assessment), privacy (never entering identifiable applicant data into unapproved or personal web-based AI tools), fairness (standardizing inputs across applicants and checking drafts for linguistic bias such as agentic versus communal descriptors), and accuracy (editing out generic or exaggerated AI language).

Whether you are preparing for a traditional faculty conversation or a multi-station MMI circuit, the lesson from every AAMC report is the same: AI can organize information and ask practice questions, but your own voice and judgment decide the outcome. To rehearse with realistic timers, unscripted follow-ups, and rubric scoring built for healthcare applicants, Confetto is ready whenever you want to practice. Start your free week.

Practice question

How do you think artificial intelligence tools like ChatGPT will change the way you practice medicine, and where should physicians draw the line on using them?

See an example answer

Example answer written by our team. The applicant is invented. I see AI as a tool that can speed up how we organize information, as long as the physician remains the one accountable for the patient. When I worked as a medical scribe in an outpatient internal medicine clinic, I watched my attending spend twenty minutes after every complex visit synthesizing old chart notes for referrals. A secure, institutionally approved AI tool that summarizes a chart or helps check a differential diagnosis could give that time back to the exam room. Where I draw the line is patient privacy and clinical judgment. First, we can never put identifying patient details into a public web-based chatbot. Second, an algorithm cannot read a patient's hesitation or understand why they are skipping a medication. Last winter, a patient in our clinic kept missing his blood pressure follow-ups. A computer only saw non-compliance, but when the doctor sat down and asked open questions, we learned the patient was rationing pills because his work shifts changed. I want to learn how to use AI to handle data, while relying on my own ears and eyes to care for the person in front of me.

It answers both parts of the question directly, uses a concrete clinical scribing story to show where AI helps and where it falls short, and hits two real medical school priorities: protecting patient data privacy and keeping human judgment at the center of care.

Frequently asked questions

How did ChatGPT perform in a medical school admission interview?

In a text-based interview conducted by attending physician Dr. Sahil Mehta and reported by AAMCNews in March 2023, ChatGPT earned a solid B. It gave informed answers on the role of AI in medicine, delivering bad news to a patient, and ethical reasoning for a liver transplant, but failed when asked about college experiences, personal ambitions, and motivation for medicine.

Do medical schools use AI to replace human interviewers or decision-makers?

No. The AAMC states that while AI can help organize information, prioritize applications, or improve drafts of interview protocols, it supports rather than replaces human reviewers, program leaders, and experts making decisions about applicants.

How can admissions officers tell if an applicant used ChatGPT to script their answers?

Admissions officers look for consistency between how you present yourself on paper and how you present yourself in person. Interviewers watch how you speak, think out loud, and respond to unexpected questions, and warn that chatbot prose can make an applicant sound scripted and generic.

Is it safe to paste my AMCAS application or patient stories into ChatGPT for interview prep?

No, not with identifying details. The AAMC warns against entering identifiable applicant information into personal web-based AI tools because those tools often do not keep data private, and medical schools train students never to enter identifying patient information into AI systems.

What medical school interview formats should I prepare for alongside AI practice?

According to the AAMC, you may encounter 20- to 30-minute one-on-one interviews with a faculty member or clinician, group interviews that test teamwork on a shared task, or Multiple Mini Interviews (MMI) with 6 to 10 stations of 5 minutes each that evaluate verbal and nonverbal communication skills.

Keep reading

  • MMI practice questions with answers
  • Kira Talent
  • AAMC PREview

Sources