Questions to ask before using a health chatbot
By ChatGPT Health. Educational synthesis prepared with AI assistance. Sources checked September 28, 2026. Clinically reviewed by Harvey Castro, MD, MBA, on September 28, 2026. Approved as educational content. This page is not a clinical protocol or a claim of independent clinical peer review.
Before entering health information into an AI chatbot, understand the task, the data policy and how you will check the answer.
Choose a bounded task
Explaining a general term or helping organize questions can be easier to check than a diagnosis or treatment recommendation. A fluent answer can still be incomplete or wrong. WHO guidance describes governance challenges associated with large multimodal models in health; it does not certify every consumer chatbot for clinical use. WHO: governance of large multimodal models.
Read the privacy terms for the actual product
HHS explains that many personal health apps are outside HIPAA protections when they are not offered by a regulated entity or its business associate. Do not infer protection from a medical sounding name. Coverage depends on the service and relationship. HHS: health information on personal devices.
- Does the service retain conversations or use them to improve models?
- Can a human or a third party access the information?
- What happens when you delete a chat or close the account?
- Is there a separate agreement for the clinical service you are using?
- Can you complete the task without names, dates of birth, identifiers or documents?
A practical verification routine
Ask the chatbot to distinguish general information from assumptions. Open any source it cites and check that it actually supports the statement. Write down the questions that remain, then use your clinician or another appropriate professional for consequential decisions. Do not change a medication solely because of a chatbot answer. If you are concerned about an emergency, seek local urgent assistance rather than extending the conversation.
A useful question to take to a visit
You can ask your care team whether it uses AI for messages or documentation and how those outputs are reviewed. If an AI generated summary gets something wrong, identify the specific sentence and ask how to correct the underlying record.
Sources and further reading
Continue learning
Editorial standards and corrections · Suggest a correction. Please do not send identifiable patient information.