Is it safe to tell ChatGPT about your health?
Asking AI about a symptom or a letter from your doctor is normal now. It can help. But what you type doesn’t stay between you and the screen. Here’s what to know before you share.
Short answer
It can help, but it isn’t confidential like a visit to your doctor. What you type is saved under your account, can be used for training unless you turn that off, and can be handed over if a court asks. Ask the question, but leave out your name, date of birth, patient and insurance numbers, and the name of your doctor or hospital.
It’s late, and the letter from the hospital is full of words you don’t know. You copy it into ChatGPT and ask what it means. You’re not alone: in a Dutch survey from October 2025, 32% of adults had asked an AI chatbot about a health complaint (Univé). In the US, 13% of all adults had uploaded medical information, like test results or letters (KFF, March 2026).
Most of those people are uneasy about it. In the same KFF poll, 77% were worried about privacy, including most of the people who had already uploaded something.
What happens to what you share
A chat with an AI is not a visit to your doctor. There’s no medical confidentiality. OpenAI’s own CEO said so in July 2025: if a court asks for your chats, they can be handed over (TechCrunch).
- Your chats are saved in your history, under your account, until you delete them.
- Unless you turn it off, they can be used to train future models.
- Deleted chats are normally gone within 30 days. But in 2025, a court made OpenAI keep even deleted chats for months. More on that here.
- If you share a chat with a link, anyone with the link can read it. In 2025, thousands of shared chats showed up in Google.
What to leave out
The question is usually fine. It’s the details around it that make it yours. Before you paste a letter or a lab result, take out:
- Your name, date of birth and address, and those of anyone else in the letter.
- Patient numbers, insurance numbers and your national ID number.
- The name of your doctor, hospital or clinic.
- Anything in the header or footer of a document. That’s where most personal details hide.
You don’t need any of these for a good answer. “What does a raised CRP level mean?” works just as well without them.
Asking something personal? Melo, Kamelo’s free private chat, answers on your own computer, so your question isn’t sent anywhere. How it works
Settings that help
- Temporary chat. Not saved in your history, not used for memory or training. What it does and doesn’t do.
- Training off. In ChatGPT: Settings → Data controls → Improve the model for everyone.
- Don’t share links to chats about your health.
Honest about the limits
None of this makes the topic itself private. If you ask about a diagnosis, the AI company knows someone on your account asked about it. Your account knows who you are. For a question you really want to keep to yourself, the safest place is a chat that doesn’t leave your computer.
Where Kamelo helps
Kamelo is free privacy software for your browser. When you write about your health on ChatGPT, Claude or Gemini, a calm notice suggests a better place to ask: a temporary chat in the same AI, or Melo, which answers on your own computer in Chrome, so the question isn’t sent anywhere.