On this page
In short 1. Does it recognize emergencies? 2. Is it honest about what it is? 3. How is it tested? 4. What happens to your data? 5. Does it hand you to a human when it should? 6. Does it remember you? Common questions SourcesAn AI doctor can be a reasonable first step for questions and triage, but only if it handles emergencies, admits its limits, and protects your data.
Not all of them are built the same. A general chatbot guessing at symptoms is a different thing from a system designed for medical conversations. These are the checks that matter.
1. Does it recognize emergencies?
The most important safety feature is what happens when you describe something dangerous. A safe AI doctor tells you to call 911 or go to the emergency room for things like chest pain, trouble breathing, stroke signs, a sudden severe headache, or heavy bleeding, and it does not keep chatting as if nothing happened.
Test it with a clearly urgent scenario you would not act on. If it brushes it off, do not use it for anything else.
2. Is it honest about what it is?
A trustworthy AI doctor says it is an AI, not a licensed physician. It should not claim credentials it does not have, and it should say plainly that it cannot examine you, order tests, or replace a clinician.
Pymander is a medical AI system, not a licensed doctor, and says so. It can explain what a pattern of symptoms most likely means and what to do next, and it tells you when you need a person.
3. How is it tested?
Look for published results on recognized medical benchmarks, and for a description of how the system is checked. Pymander reports 98.2% on a USMLE-style evaluation and a top-five result on HealthBench Hard, with methods on its research pages. A benchmark is not a guarantee about any single conversation, but a system that publishes how it is measured is easier to hold to account than one that does not.
4. What happens to your data?
Ask whether messages are encrypted, who can see them, and whether you can delete your account. Health information is sensitive. Pymander states that data is encrypted and under the member's control, and describes HIPAA-aligned safeguards.
5. Does it hand you to a human when it should?
A good AI doctor knows what it cannot do. Prescriptions, physical exams, lab orders, and formal diagnoses come from licensed clinicians. The right behavior is to say that and to send you to urgent care, your primary care clinician, or the ER when that is the right answer, not to improvise.
6. Does it remember you?
Medical context matters: your medications, allergies, and history change the answer. An AI doctor that keeps a record of your earlier consults can give better advice and can send you a care note to bring to your own doctor.
Common questions
Is an AI doctor safe to use?
It can be, as a first step for questions and triage, if it is built for medical conversations. Check that it recognizes emergencies, says clearly that it is not a licensed doctor, publishes how it is tested, protects your data, and sends you to a person when needed.
Can an AI doctor replace my real doctor?
No. An AI doctor cannot examine you, order tests, or prescribe. It is useful for understanding symptoms, deciding how urgent something is, and preparing for a visit. Use it alongside your own clinician.
What should I never rely on an AI doctor for?
Emergencies. If you have chest pain, trouble breathing, stroke signs, a sudden severe headache, heavy bleeding, or thoughts of harming yourself, call 911 or go to the emergency room. In the US you can also call or text 988 for a mental health crisis.
Is Pymander a licensed doctor?
No. Pymander is a medical AI system, not a licensed doctor, and it does not provide medical advice, diagnosis, or treatment. It gives guidance, tells you when to see a clinician, and writes a care note you can share.
How do I know the answers are any good?
Look for published evaluation results and a clear explanation of how safety is handled. Pymander publishes its results on its research and safety pages. Always check anything important with a licensed clinician.