Do not enter identifying personal or health data. Ask AI to show uncertainty, ask follow-up questions, mention red flags and use reliable sources. Use AI to understand and structure information — not to make urgent decisions, diagnose yourself or change medication.
Try first without changing settings
You can guide AI in a single conversation. Add this short instruction after your health question:
Ask the most important follow-up questions first if the information is not enough. Mention red flags and when I should seek medical care urgently. Use reliable sources when you make factual or guideline-based claims. Say clearly what cannot be concluded from the information given. Do not make a diagnosis without medical assessment.
What AI is useful for in health questions
Preparing for an appointment
AI can help turn a messy symptom story into a clearer list of what to tell the doctor.
Understanding instructions
It can explain medical terms and help you make a short question list for a doctor or nurse.
Summarising home measurements
AI can help create a table, averages, a printable summary or simple charts from a blood pressure card or glucose notebook.
Finding general background
It can help compare general information, but important claims should be checked from reliable sources.
Privacy: do not paste everything into AI
Use a general description. Do not include a personal identity code, address, exact workplace, full Kanta text, prescriptions, referrals or images containing identifiable details.
For a photo of a blood pressure card or glucose notebook, crop or cover the name and any other identifiers before uploading it.
Example: blood pressure or glucose notes before a visit
“This is a photo of a blood pressure card. The name has been cropped out. Please make a table for a doctor visit, calculate morning and evening averages if possible, and create a simple chart and printable summary.”
When not to use AI
Use it for
- preparing questions
- structuring symptoms
- understanding general terms
- summarising non-identifiable measurements
Do not use it for
- urgent symptoms
- making a final diagnosis
- starting, stopping or changing medication
- replacing an examination
- sending identifiable health records
Save a custom instruction if you use AI often
Many AI services let you write instructions in plain language. In ChatGPT, this is found through profile/settings and personalization/custom instructions. The wording does not need to be technical.
When answering health questions, be cautious, source-based and intellectually honest. Separate facts, assumptions, clinical judgment and speculation. If the information is not enough, ask up to 3 important follow-up questions before giving a conclusion. If the missing information is not critical, state your assumptions. Do not present uncertainty as certainty. Mention red flags and when urgent or in-person medical assessment is needed. Do not give a final diagnosis or medication change as if it replaced a doctor. For numbers, statistics, guideline claims and research claims, use reliable sources when possible. Do not invent citations, links, studies or guidelines. If you cannot verify something, say so clearly. Remind me not to paste identifying health information into the service. Do not ask for personal identity code, full name, exact address, full Kanta text, prescriptions or identifiable images unless the situation genuinely requires it and the service is appropriate for that use.
Emergency symptoms: do not ask AI
With symptoms such as severe chest pain, stroke symptoms, severe breathing difficulty, major trauma, heavy bleeding or rapidly worsening general condition, call the emergency number or seek urgent medical care. AI is not the right first step in an emergency. It also does not replace human support or professional care in an acute mental health crisis.
Sources and further reading
The source list documents the background for AI use in health information, custom instructions, usage limits, source criticism, privacy and healthcare AI safety.