Health preparation
Prepare the conversation. Leave the diagnosis to your clinician.
Use ChatGPT or Claude to organise symptom notes and appointment questions, with clear limits on diagnosis, medicines, privacy and urgent care.
Updated
Use AI as a notebook, not a clinician
A useful task is turning scattered observations into notes you can take to an appointment. A different and unsafe task is asking a chatbot to diagnose you, choose a dose or decide that a symptom can wait. Fluent wording does not establish medical accuracy; WHO warns that generated health information can be misleading or inaccurate.
If you think you may have an emergency, contact local emergency services immediately. Do not wait for AI to assess urgency. These templates are not a triage service and cannot rule out a serious problem.
- Ask: organise my observations, help phrase questions, identify unclear wording to confirm with the care team.
- Do not ask AI to replace a clinician's diagnosis, change prescribed treatment or guarantee that waiting is safe.
- If an answer conflicts with a medical instruction, ask the responsible clinician or pharmacist; do not resolve the conflict by choosing the more reassuring answer.
Prepare facts and priorities, not a diagnosis narrative
Write down what you noticed, when it began, how often it happens and how it affects daily life. Keep your own observations separate from explanations you found online. Bring your main concerns and an accurate list of medicines and supplements to the discussion. NIH's appointment guidance supports preparing notes and questions so important details are not forgotten.
Share less with an AI service than you would with your clinician. Remove names, addresses, identification numbers and unrelated history. You do not need to upload a complete medical record to organise three questions.
Help me organise these notes for an appointment: [redacted notes]. Make sections for observations, timeline, frequency, change over time, daily impact and my three priority questions. Keep my uncertainty and approximate dates explicit. Do not add symptoms, infer a diagnosis or advise whether care can wait. Mark factual details I could clarify for my clinician.
Before and after: a better appointment question
Weak prompt: ‘What illness do I have and which medicine should I take?’ It asks the system to make a clinical decision without an examination or a reliable assessment of your circumstances.
A stronger request asks for a conversation aid. Illustrative scenario: someone has notes about recurring discomfort and wants to avoid forgetting the main questions during a routine visit. This is not an assessment of any actual person's symptoms.
I have an appointment booked and want help preparing, not a diagnosis. My observations are [notes] and my priorities are [priorities]. Draft a 30-second factual opening and three questions about assessment, options and follow-up. Include space for the clinician's answers. Do not recommend medicines, tests or treatment; phrase those as questions for the clinician.
After the visit, preserve the actual instructions
An AI summary should not quietly fill gaps in an unclear instruction. Keep the original letter and compare every action, date and medicine entry with it. Ask the care team about anything ambiguous, including what to do if the situation changes. A generated interpretation is not a replacement for their plan.
Using only these redacted appointment notes, list confirmed actions, dates and questions needing clarification. Reference the passage for each instruction. Do not infer medication directions or add advice. Draft a short message asking the clinic to clarify the missing details; do not send it.
Common questions
Can AI interpret my test results?
It can help you identify terms to ask about, but it should not diagnose you or decide treatment from a result. Ask your clinician to interpret it in your clinical context.
Should I upload my full medical record?
Not just to prepare an appointment. Use the minimum information needed, remove identifiers and consider the service's privacy settings and terms before sharing health information.
Keep exploring
Sources and editorial note
Guidance checked 9 October 2026. These AI-assisted guides and original templates are reviewed against the linked sources. Results depend on your inputs and available tools. Model recommendations are editorial starting points, not comparative test results.