I left my doctor's appointment last March with the same feeling I'd had for years: like I'd described my body in a foreign language and she'd nodded politely. I had a notebook of symptoms. I had timing. I had nothing that translated into a useful conversation.
§The problem isn't your doctor. It's your preparation.
Most women walk into appointments with raw notes — a list of symptoms, some dates, a vague feeling something is wrong. That's data. It's not a narrative. Your doctor has eleven minutes, and raw data takes eleven minutes just to parse.
§The prompt structure
Paste your raw symptom notes into ChatGPT, Claude, or Gemini, then use this structure. It works across every model we tested:
The output is almost always the same shape: a one-line summary, a timeline, three questions to ask, and one test to request if appropriate. That sheet is what you hand your doctor — or read from.
My doctor said it was the most prepared she'd ever seen a patient. It took me eleven minutes to write. Before, I'd spent two years not being taken seriously.
§What never to share with AI
Never paste your full name, address, insurance numbers, or identifying medical record numbers into a public AI tool. Use first names or initials only. The symptom narrative is all AI needs — the identifiers are for your doctor's office, not the model.
AI-Powered Health
ChatGPT, What's Wrong With Me?
50 prompts. One method.
Written by
EssenzStudio Editorial
Editorial Team · Reviewed by the EssenzStudio editorial desk
The EssenzStudio editorial desk researches, writes, and fact-checks every piece before it ships. Wellness guidance built for women — without the jargon, without the condescension.
Filed under
Frequently asked
Is this self-diagnosis?+
No — and that's the point. The prompt explicitly asks AI not to diagnose, but to structure your symptoms for a clinician. You walk in prepared; your doctor still makes the diagnosis.
Which AI model is best for this?+
Any major model works for structuring (ChatGPT, Claude, Gemini). For medical terminology translation, Claude and GPT-4 perform slightly better on nuance. But the prompt structure matters more than the model.
