AI in healthcare: Can a chatbot answer your medical questions?
Follow an AI researcher's green light/yellow light/red light guide for the best results.
- Reviewed by Adam Rodman, MD, MPH, FACP, Contributor
Ask ChatGPT, Gemini, or another artificial intelligence (AI) chatbot a medical question, and you may be amazed at how fast it comes up with a detailed answer. Depending on the prompts you give it, the answer can be tailored for your needs. But be careful if you want to put this technology to use.
Can AI answer medical questions?
While AI in healthcare has impressive abilities (see “What is AI in medicine?”), it also has limitations, particularly when patients (rather than doctors) apply it. “AI chatbots are good for some medical queries, but not for others,” says Dr. Adam Rodman, a general internist and AI researcher at Harvard-affiliated Beth Israel Deaconess Medical Center.
To help people understand the best way to use AI in healthcare, Dr. Rodman developed a stoplight system. He gives a green light to safe and helpful uses. Yellow-light questions can be asked but require caution. A red light marks queries best avoided.
What is AI in medicine?Artificial intelligence (AI) is rapidly becoming an indispensable tool in every realm of medicine, shaping how doctors and other health professionals work, from basic research to the diagnosis and treatment of patients. AI systems learn patterns from large amounts of data and then use these patterns to perform tasks that are usually associated with human thinking. Large language models (LLMs) are a type of AI that can recognize and generate human-like language. AI systems can also perform tasks such as identifying signs of cancer on mammograms that might be hard for humans to see. An increasing number of clinicians use AI tools, including LLMs, to assist with clinical decision making, writing notes in medical charts, and keeping up with medical knowledge. |
Green-light questions
“Green-light questions are those you can feel comfortable asking without discussing it with a health care provider,” says Dr. Rodman. For example, you can ask a chatbot to help you understand general medical information: “Explain diabetes.” “What is hypertension?”
You might ask a chatbot for advice about diet: “I have lactose intolerance. Can you help me develop a menu plan?”
Chatbots can also explain the meaning of doctors’ notes, lab reports, and imaging test findings. If you don’t understand what’s written in your medical record, ask a chatbot to “explain what the doctor meant in this note.” (Remove personal identifiers first.)
An AI chatbot can also help you prepare for a doctor’s appointment. Enter information from your last visit and ask, “How should I prepare for this visit? List three ways.”
Yellow-light questions
You can ask yellow-light questions, but be aware that the answers won’t replace a doctor’s clinical judgment.
People often ask chatbots about new symptoms. “There are studies suggesting that AI large-language models are very good, if not better, than humans, at exploring new symptoms,” says Dr. Rodman. However, the accuracy depends on the context.
Answers can reflect the way you interact with the chatbot. “For example, they often tell you what they detect you want to hear, which can be dangerous,” says Dr. Rodman. That may lead to misinformation and anxiety.
Instead of typing in a list of symptoms, Dr. Rodman advises using a prompt something like this: “I have a new headache. Can you interview me like you’re a doctor to help explore my symptoms?” The chatbot will ask questions that put the symptoms into context. In addition to the symptoms you’re experiencing, it will ask about symptoms you’re not having.
“While you can explore new symptoms and potential diagnoses with a chatbot, this should be done only as a first step before talking to a health care professional,” says Dr. Rodman.
Red-light questions
Based on the current limitations of AI, you can’t rely on the accuracy of answers about how to treat and manage health conditions.
Avoid questions such as the following: Is this the right treatment plan for me? Should I take this medication? Is this the correct chemotherapy regimen? These are red-light questions.
“I tell my patients, just don’t do it,” says Dr. Rodman. There’s too much involved in treatment decisions, including physical examinations, medical history, and clinical judgment, to trust a chatbot.
Can AI replace doctors?
By the time you read this article, AI chatbots will have evolved. That’s how fast technology changes. But it doesn’t alter the bottom line: you shouldn’t rely on an AI medical diagnosis, and don’t think of a chatbot as your personal AI doctor. “AI is not a replacement for health care professionals,” says Dr. Rodman.
Image: © PonyWang/Getty Images
About the Author
Lynne Christensen, Staff Writer
About the Reviewer
Adam Rodman, MD, MPH, FACP, Contributor
Disclaimer:
As a service to our readers, Harvard Health Publishing provides access to our library of archived content. Please note the date of last review or update on all articles.
No content on this site, regardless of date, should ever be used as a substitute for direct medical advice from your doctor or other qualified clinician.
