25 AI Questions You Should Prepare for Before Your Medical School Interview
Medical school interview preparation has traditionally involved a predictable collection of questions:
- Why medicine?
- Why this school?
- Tell me about a challenge.
- Tell me about an ethical dilemma.
- Tell me about healthcare disparities.
Now add another category:
Artificial intelligence.
Applicants have already reported being asked about AI in medical school interviews, and schools are explicitly incorporating AI into secondary applications. Mayo Clinic Alix School of Medicine currently asks applicants about their role in a future involving artificial intelligence and innovation. Oakland University William Beaumont previously asked applicants to consider using AI to address healthcare barriers and disparities.
So here are 25 questions I believe every serious applicant should be prepared to discuss.
Notice I said discuss.
Not memorize.
If you memorize 25 answers, I will personally ask the interviewer to change question #17.
Category 1: The Big Questions
1. How do you think AI will change medicine?
Whether you understand that medicine is changing and can discuss innovation without making ridiculous predictions.
A strong answer recognizes both opportunity and risk.
AI may improve efficiency, analyze large amounts of information, assist with diagnosis, reduce documentation burden, and help physicians identify patterns.
But implementation also raises questions involving accuracy, bias, privacy, accountability, and overreliance.
Your answer should not sound like an advertisement for an AI company.
It should also not sound like your laptop personally offended you.
2. Will AI replace physicians?
Start by separating physicians from physician tasks.
AI may automate portions of documentation, image analysis, information retrieval, risk prediction, and administrative work.
That does not automatically mean the profession disappears.
The physician’s role may instead shift toward interpretation, complex decision-making, communication, ethical judgment, contextualization, and accountability.
3. What is the greatest benefit AI could bring to healthcare?
There are many reasonable answers:
- Improved diagnostic support
- Reduced administrative burden
- Expanded access
- Earlier disease detection
- Personalized treatment
- Better analysis of complex data
The key is explaining why your chosen benefit matters to patients.
4. What is the greatest risk?
Again, there is no single answer.
Possible risks include:
- Bias
- Privacy
- Hallucination
- Overreliance
- Poor validation
- Loss of clinical skills
- Unequal access
- Reduced human interaction
Choose one and explain it.
Do not list eleven buzzwords like you are reading the table of contents of an AI ethics textbook.
Category 2: Physician Versus Machine
5. An AI system disagrees with your diagnosis. What do you do?
Do not make this a competition.
Investigate the disagreement:
- Reevaluate the patient.
- Review the evidence.
- Understand the AI system’s limitations.
- Seek additional input if necessary.
The correct mindset is curiosity rather than ego.
6. What if AI is statistically more accurate than physicians?
Now the question becomes harder.
If robust evidence shows that a validated tool improves patient outcomes, physicians should take that evidence seriously.
But “94% accurate” does not answer every question.
Ask:
- Accurate in whom?
- Compared with what?
- For which outcome?
- Under what circumstances?
- Was it externally validated?
Medicine loves numbers.
Good physicians ask what the numbers actually mean.
7. Should physicians ever override AI?
Yes, there may be circumstances where clinical context, patient preferences, limitations of the model, unusual presentations, or information unavailable to the system justify overriding its recommendation.
But overriding it simply because:
“I have been doing this for 30 years”
is not automatically evidence either.
8. Could relying on AI make doctors worse clinicians?
Potentially.
This introduces automation bias—the tendency to over-trust automated recommendations.
If physicians stop independently reasoning because “the computer said so,” clinical judgment may deteriorate.
The solution is not necessarily banning AI.
The solution is training physicians to use AI critically.
Category 3: Responsibility
9. Who is responsible when AI makes a medical error?
The physician?
Hospital?
Developer?
Vendor?
All of them?
The answer can depend on the circumstances.
But from the physician’s perspective, using a tool does not magically eliminate professional responsibility.
A calculator can assist you.
It cannot go to the patient’s family and explain what happened.
10. Should an AI ever make a medical decision without physician review?
This depends on the decision, risk, evidence, regulatory environment, and safeguards.
An automated appointment reminder is different from an autonomous cancer diagnosis.
Risk matters.
The higher the stakes, the stronger the argument for meaningful oversight.
Category 4: Bias and Equity
11. Can AI be biased?
Yes.
Models learn from data.
If the underlying data reflect unequal healthcare access, underdiagnosis, historical discrimination, or underrepresentation, the model may reproduce those patterns.
AI did not invent human bias.
Unfortunately, we gave it excellent training material.
12. Could AI make healthcare disparities worse?
Yes.
Imagine a highly effective AI diagnostic platform available primarily at wealthy hospitals.
Patients in well-funded systems benefit.
Patients elsewhere do not.
The technology improves healthcare overall while simultaneously widening inequality.
Both statements can be true.
13. Could AI reduce healthcare disparities?
Also yes.
Potential applications include:
- Remote clinical support
- Translation
- Patient education
- Earlier screening
- Decision support in underserved areas
- Improving access to expertise
OUWB’s 2025–2026 secondary essentially asked applicants to grapple with exactly this issue: how AI might responsibly address healthcare obstacles in communities facing health-equity challenges.
14. What if an AI system performs worse for a minority population?
Now you need to discuss:
- Validation
- Transparency
- Harm
- Alternatives
- Whether continued use is justified
A tool with excellent average performance can still be unacceptable if the “average” hides meaningful harm to a particular population.
Always ask:
Who is inside the average?Category 5: Patients
15. A patient says, “ChatGPT says I have cancer.” What do you do?
Please do not say:
“ChatGPT is wrong.”
You have not evaluated anything yet.
Instead:
- Ask what symptoms the patient entered.
- Ask what the tool told them.
- Understand why they are concerned.
- Evaluate their symptoms appropriately.
- Explain what can and cannot be concluded.
- Address their anxiety.
A patient arriving with AI-generated information is not fundamentally different from the patient who once arrived with 14 printed pages from Google.
Medicine survived Google.
We will survive this too.16. What if a patient trusts AI more than you?
Do not turn it into a power struggle.
Trust is earned through communication.
- Ask why the patient trusts the recommendation.
- Discuss the evidence.
- Acknowledge uncertainty.
- Explain your reasoning.
- If the AI raises a legitimate possibility, investigate it.
Your goal is not to defeat ChatGPT.
Your goal is to care for the patient.
17. Should patients know when AI contributed to their care?
Consider transparency and autonomy.
The amount of disclosure reasonably required may depend on how consequential the AI’s role was.
There is a difference between AI scheduling an appointment and AI substantially influencing a diagnosis.
Category 6: Privacy
18. Would you put patient information into ChatGPT?
Not casually.
Patient confidentiality does not disappear because a tool is convenient.
Healthcare organizations need appropriate privacy, security, consent, and data-handling policies.
The AAMC specifically includes data privacy among its principles for responsible AI use in medical education.
19. Should patients consent before their medical information is used to train AI?
Excellent ethics question.
You could discuss:
- Autonomy
- Privacy
- De-identification
- Public benefit
- Commercial use
- Data governance
- Informed consent
Do not rush toward a yes/no answer.
Show that you recognize what is at stake.
Category 7: Medical Education
20. Should medical students use ChatGPT?
My answer as an educator is:
That depends on what you are using it for.
For example:
- Explaining a concept
- Creating practice questions
- Brainstorming
- Summarizing
- Checking your understanding
Potentially useful.
Having it complete the intellectual work you were supposed to learn how to perform?
Now we have a problem.
If AI performs every difficult thought for you, eventually somebody will hand you a patient.
The patient does not come with a “regenerate response” button.
21. Should AI literacy be taught in medical school?
Increasingly, yes.
The AAMC explicitly advocates foundational AI literacy and training medical learners to critically assess AI outputs for accuracy, reliability, and trustworthiness.
Future physicians do not necessarily need to become programmers.
They do need to understand the tools influencing their decisions.
Category 8: Admissions
22. Did you use AI to prepare your medical school application?
Answer truthfully.
Different schools are developing different policies.
For example, a reported 2026–2027 Methodist University Cape Fear Valley policy allows AI for activities such as brainstorming, proofreading, and editing while requiring disclosure and requiring the final submission to reflect the applicant’s own work.
GW’s reported 2026–2027 attestation takes a stricter approach toward AI-generated application writing.
Know each school’s policy.
23. When does AI-assisted writing become academic dishonesty?
There is an important difference between:
“Help me find a grammatical error.”
and:
“Write my personal statement.”
The first assists expression.
The second can replace authorship.
Between those extremes is a large gray area—which is exactly why applicants need to read each school’s policy.
Category 9: The Future
24. What role do you want AI to play in your future practice?
You do not need an AI startup.
You do not need to tell the interviewer:
“Ever since childhood, I have dreamed of artificial intelligence.”
Especially if ChatGPT was released while you were in college.
Be authentic.
Maybe you are interested in using technology to reduce documentation so physicians can spend more time with patients.
Maybe you care about:
- Diagnostics
- Access
- Research
- Education
- Health equity
Or maybe your goal is simply to become sufficiently AI-literate to evaluate these tools responsibly.
That is a perfectly respectable answer.25. What should every future physician understand about AI?
his may be my favorite.
My answer would center on one principle:
AI should extend judgment, not replace the responsibility to exercise judgment.
The technology will change.
Today’s systems will eventually look primitive.
The durable skill is learning how to evaluate a tool, recognize its limitations, communicate uncertainty, protect patients, and remain accountable.How to Practice These Questions
Do not write 25 speeches.
Take each question and practice speaking for approximately 60–120 seconds.
Then have somebody challenge your answer:
- “What if the AI is more accurate?”
- “What if the patient refuses?”
- “What if there is no physician available?”
- “What if the algorithm is cheaper?”
- “What if using AI could save 10,000 lives?”
That second layer is where interview preparation becomes useful.
Because the interview is not testing whether you memorized an answer.
It is testing whether you can continue thinking after your first answer stops working.
That skill will serve you long after interview day.
Probably longer than whatever version of ChatGPT exists when you finally become an attending.