Module 5 – Asking medical questions in ChatGPT
Total time: about 60–90 minutes · three walkthrough exercises (heart, knee, your own symptom), short reflection each time, and a final quick multiple-choice check.
We work through three examples: palpitations, then knee pain (same approach, different symptom), and finally a non-urgent symptom of your own that you choose. You reflect briefly in between; at the end comes a short quiz.
The goal is that you get quick, understandable information about your health question, without searching in circles: but in a safe way: describe briefly and anonymously, follow up with questions, and know that examination, diagnosis, and treatment belong with a clinician. If in doubt or with acute symptoms: your GP or emergency services (e.g. 112 in many European countries).
What this module is about
Keep two tracks separate: Information and preparation, concepts, background, ordering your story, symptom list, questions for the appointment, ChatGPT can help with that. Decision-making, what is going on with you, which tests, which treatment, whether it is urgent, belongs with someone who can see you and knows your record.
- Start and follow-up questions: begin briefly (when, where, how bad); answer follow-up questions as in a real appointment, you practise that on purpose.
- Pitfalls: answers can sound smooth and convincing yet be incomplete or wrong; the model does not see you and has no stethoscope. Use explanations as background, not as confirmed truth about your body.
- Privacy: no national ID number, full name + date of birth, or entire letters; prefer [approximate age], left/right, short description.
- Value: language, structure, practising follow-up questions, and materials for the doctor, that does not replace a physical exam or medical judgment.
Red lines (keep in mind)
- AI supports preparation and language, not examination, diagnosis, treatment, or emergency triage.
- Acute or severe: call emergency services or go to the ER; do not use chat to “check if it is serious”.
- No full record or national ID in the chat; keep it short and as anonymous as possible.
Part A – Example: palpitations
You practise the feeling of a fast or irregular heartbeat, a common example where people often look online first. Here the focus is phrasing and structure; you do not measure heart rhythm in a chat.
Click here to open ChatGPT in a new tab, this course page stays open.
Practice exercise (part A), work through this step by step on purpose
Allow at least 10–15 minutes. You may use fiction (“say someone aged 62…”) if you prefer not to type anything personal.
You may shift the order slightly (e.g. ask for explanation earlier) as long as you hit this core: at least one round of follow-up questions, plain-language explanation, and a deliberate close (summary and/or questions for the doctor).
You can copy and paste the texts below into ChatGPT, but you do not have to, your own words are fine too.
- Open ChatGPT via the link above (new tab; keep this course page open).
- Start simply: type in one or two sentences that you (or your fictional person) feel palpitations, fast, pounding, or irregular in the chest. No national ID number or full name.
- Wait for the answer. Does it ask follow-ups? Answer them briefly (truthfully or fictional).
- Continue: answer at least one more round of follow-ups. Make it a mini-conversation, like an appointment.
- Then: ask for general explanation in plain language (e.g. what people often mean by “palpitations” and which kinds of causes doctors sometimes distinguish, without pretending the chat knows what you have).
- At the end: optionally ask for a short summary and help for your GP visit, see the prompts below.
1) Opening line (adapt or copy)
2) Optional, short summary in the chat
3) Preparing for the GP
Case A, how did it go?
1. You just practised palpitations and answered follow-up questions. What is most useful to bring to your GP conversation?
2. ChatGPT asked clarifying questions about your symptom. How is that best understood?
3. ChatGPT gave general information about palpitations. What is a sensible way to use that?
Part B – Example: knee pain
Knee problems are common: overuse, wear (osteoarthritis, in a general sense), meniscus or ligament (for illustration only, the chat does not know what you have). Here you practise again: location, movement, swelling, locking feeling, after which activity, and a question list for your GP or physiotherapist.
Click here to open ChatGPT in a new tab, this course page stays open.
Practice exercise (part B), same approach as the heart, now with the knee
Allow at least 10–15 minutes. Again: short and anonymous (left/right, since when) is better than your whole record.
Same freedom as in part A: order may vary a bit; keep follow-up questions, big-picture explanation, and closing with summary and/or questions for a professional.
- Open ChatGPT (link above).
- Type briefly that your knee bothers you, one or two sentences is enough to start. If you cannot think of a line, copy the opening below into ChatGPT.
- Wait for the answer. Are there clarifying questions (side, swelling, weight-bearing, after which movement)? Answer them.
- Stay in the conversation: answer at least one more round of follow-ups. Practise follow-up as in real life.
- Then ask for general explanation: what kinds of causes doctors often distinguish with knee complaints. Let the model set frames where needed.
- Close with a short summary and a list of questions for your GP or specialist, in your own words in the chat, or with the line below.
1) Opening line
2) Summary and questions for the doctor
Case B, how did it go?
Same idea as with the heart: it is about what you learn on chatting usefully around symptoms.
1. After your knee pain practice: what is a good next step in the real world?
2. You started with a short line about your knee and then got questions about side, swelling, loading. What was the practice goal?
3. ChatGPT may have mentioned general categories of knee problems. How should you place that?
Part C – Your own symptom (you choose)
Pick something of yours: headache, allergy, back pain, stomach, skin, something non-urgent you recognise. (Severe chest pain, fainting, or other emergencies: GP or emergency services right away, not chat.) You practise again: short description, follow-ups, structure, and closing with something useful for a possible appointment.
Click here to open ChatGPT in a new tab, this course page stays open.
Practice exercise (part C), your own (non-urgent) symptom
In the opening line, replace [your symptom in 1–2 sentences] with something that fits you. No full name, address, or national ID number. Stick to calm symptoms; for acute warning signs, do not use chat.
You may choose extra follow-up rounds first or ask for explanation a bit earlier, as long as you have at least two short follow-up rounds before you ask for a long explanation, and you close with the prompts below (or your own version of the same goal).
- Open ChatGPT.
- Type briefly that you have trouble with [your symptom], keep the first message deliberately simple.
- Answer clarifying questions; at least two rounds of follow-ups before you ask for a big explanation.
- Ask for general background in plain language (e.g. directions doctors sometimes explore, without claiming a diagnosis for you).
- Close with summary and GP help as below (or your own words with the same aim).
1) Opening line (fill in yourself)
2) Optional, short look-back in the chat
3) Conversation card + questions for the GP
Case C, how did it go?
1. You practised with your own (non-urgent) symptom. What does it mean if ChatGPT says it is not a doctor or has limits?
2. How can you best combine a natural chat with safety?
3. At the end you asked for a conversation card and question list for the GP. Why is that useful?
Pitfalls and hard limits
This sums up what you already saw in the module: once in one place, from what can help to hard limits:
- Fast and broad knowledge: ChatGPT can often help quickly with explanation, wording, and structure. Its training included a huge amount of medical and other health text (including guidelines), so general information is often genuinely useful.
- Still: plausible ≠ correct: answers can sound smooth and complete yet still miss or misstate, especially with rare symptoms, details, or your specific situation.
- No body, no record: no ultrasound, blood pressure, ECG, or your whole history; everything it “knows” comes from text and training, not from examining you.
- Tone: it can sound reassuring or certain without medical basis, if you doubt or symptoms worsen: real care, not another chat round as a “second opinion”.
- No diagnosis or fixed plan from chat alone; no prescriptions or doses, only via a doctor or pharmacist.
- Emergencies: for serious symptoms, emergency services or your doctor, do not wait for an AI answer.
Quick check, general (module 5)
1. What is appropriate for medical questions in ChatGPT?
2. When should you not wait for an answer from AI?
3. Which approach fits best when practising symptoms in ChatGPT?
4. You want to practise knee pain in ChatGPT. What is sensible?
5. What is a sensible way to discuss medical topics in ChatGPT?
6. Why is practising “your own symptom” useful if you keep it safe?
7. Can you blindly trust ChatGPT answers about health?
8. What can you use ChatGPT for a lot with (non-urgent) symptoms, if you prompt safely and with limits?
10 ready-to-use prompts (save or print)
All suitable for free chat, adapt the brackets.
- Explain [topic] in plain language, at most 3 sentences.
- Rewrite this in a friendlier way without changing the core message.
- Answer in 5 bullets.
- Ask 3 clarifying questions first, then answer.
- Turn this into a checklist.
- Rewrite this as a short email with a subject line.
- Make 3 versions: short, friendly, formal.
- What is missing from my question so you can answer better?
- Check your answer for assumptions and unclear points.
- Rewrite this text without identifiable details.
Optional extra practice: week plan from rough notes; improve a weak prompt in three steps; structure a fictional non-urgent symptom for a GP visit; rewrite an unsafe prompt safely; make your own AI do/don’t card.
Finished module 5
You have worked through three cases (heart → knee → your own symptom), reflected briefly each time, and seen pitfalls and emergency rules in one overview. The quick check above states the core once more. That completes the five course modules; evaluation and optionally the workbook follow next.