AI in Medical Training

AI for Medical Students in Malaysia: How to Use It Without Losing Skills

How medical students in Malaysia can use AI to study, with survey data on current use, risks of invented references and over-reliance, and rules for faculties.

  • By
  • Updated
  • 4 min read
  • Facts checked
Short answer

Medical students can use AI to explain concepts, quiz themselves, summarise papers they then read, and practise history-taking with a virtual patient. They should never use it to replace their own reasoning, should check every fact and reference, and should enter no real patient data. Faculties should set clear, written rules on permitted use.

Part of the guide: AI in Healthcare in Malaysia: Courses and Rules for Doctors, Nurses and Hospitals (2026)

On this page
  1. How many students already use AI?
  2. What are good uses for a student?
  3. What are the traps?
  4. What should faculties do?
  5. What does a sensible study routine look like?
  6. How should a student disclose AI use?
  7. What about clinical postings?
  8. What should you do next?
  9. Sources
  10. Frequently asked questions

Students are already using AI, usually without guidance. The question for a medical student is which uses build skill and which erode it. A simple test: after using the tool, could you explain the answer to a supervisor without it?

Checked on 22 October 2026.

How many students already use AI?

A survey of 443 undergraduates at a Malaysian healthcare institute found MBBS students significantly more likely to use ChatGPT for academic purposes than dental and other students, alongside concern about accuracy and ethics. It is a sample from a single institution, so use it as a signal, not a national figure. A peer-reviewed review of AI in Malaysian healthcare also reported that only 44.5% of medical students felt adequately prepared to work with AI.

What are good uses for a student?

Use How it helps Guard
Explain a concept another way Fresh angle on a hard topic Compare with your textbook
Quiz yourself Questions on a topic you pasted in Check answers against a trusted source
Summarise a paper Quick overview before you read it Read the paper; trace claims
Practise a history A virtual patient to talk to Treat as practice, not as proof of skill
Plan revision Turn a syllabus into a schedule Adjust to your own pace

What are the traps?

  • Invented references. Models can produce plausible citations that do not exist. Never cite something you have not opened.
  • Confident errors. ChatGPT scored 52.4% to 75.0% on public USMLE sample questions in a 2023 study, around the pass line of roughly 60% (Kung et al., PLOS Digital Health, 2023). It still gets many questions wrong.
  • Bias. In another study, four chatbots all produced examples of race-based medicine (Omiye et al., npj Digital Medicine, 2023).
  • Skills loss. If the tool always builds your differential, you may never learn to. MMC guideline clause 3(11) cautions practising doctors about the same thing.
  • Patient data. Never enter details of a patient you met in a posting.

What should faculties do?

  1. Publish a one-page policy: permitted, permitted with disclosure, not permitted, for each assessment type.
  2. Teach verification as a skill, with exercises where students find errors in AI output.
  3. Keep some assessments AI-free, such as OSCEs and oral exams.
  4. Give staff training first, so rules are consistent.
  5. Review the policy every term; tools change quickly.

Our approach for educators is in How can AI be used in medical education?, and the course page for faculties is AI for medical schools. Question writing is covered in AI for writing MCQs and OSCE stations.

What does a sensible study routine look like?

A routine that builds skill keeps you doing the reasoning first. For example:

  1. Attempt the question or case yourself and write your answer.
  2. Ask the tool to critique your answer or give a different explanation.
  3. Check anything new against your textbook or a guideline.
  4. Write down what you got wrong and why, in your own words.

The order matters. If the tool goes first, you copy. If you go first, it teaches.

How should a student disclose AI use?

Many faculties ask for a short note. Here is an example format, to be adapted to your faculty's rules: "I used [tool] to [purpose, such as generating practice questions on cardiac physiology]. I checked all facts against [sources]. The final text and reasoning are my own." If your faculty has no policy, ask your coordinator before submitting and keep your prompts and drafts in case you are asked.

What about clinical postings?

On postings you meet real patients, and the temptation is to ask a tool about a case. Do not enter details of the patient. Describe the problem in general terms ("a patient with pleural effusion, what are the causes?"), or use a constructed case. The consent and confidentiality expectation in the MMC guideline sets the standard you will be held to as a doctor, and practising it now builds the habit.

What should you do next?

If you teach students, see the AI in medical education course or the overview in the AI in healthcare guide for Malaysia. To plan a session for your faculty, book a free 30-minute call.

Sources

(checked November 2026)

General information, not medical or legal advice.

Frequently asked questions

How many Malaysian medical students use AI tools?

Figures vary by sample. A mixed-methods survey of 185 undergraduates at a public medical school in Sarawak found 40.0% reported using AI tools, with ChatGPT the most used. It was a convenience sample, so treat it as an indication, not a national number.

Can I use ChatGPT for my assignments?

It depends on your faculty's policy. Many institutions allow some uses and require disclosure, and prohibit others. Read the rules for each assessment and ask if unsure. Never submit AI-generated references you have not checked.

Is it safe to put patient cases from my posting into AI?

No. Patient information from clinical postings must not go into public AI tools. The MMC guideline sets the standard for doctors, and students should treat it as the minimum: no patient data without consent.

Will AI make me a worse doctor?

It can if you let it do the thinking. MMC clauses 3(10) and 3(11) warn about over-reliance and skills degradation. Use it to test your reasoning, not to replace it.