When a patient asks a health question online, most first turn to a chatbot like Claude, Gemini, or ChatGPT instead of Google. Unlike a search engine that hands back links, these AI assistants spit out a concise, clinician‑like summary that often feels authoritative—hence the nickname “Dr. Google on steroids.” The same power that makes the tools useful can also mislead, with physicians reporting frustration when patients arrive armed with contradictory or incomplete information.

Digital‑health specialist Christopher Khoury, vice president of strategic insights at the American Medical Association (AMA), stresses that chatbots can be a helpful pre‑visit companion. “Don’t underestimate stage fright,” Khoury says. “Patients typically have limited time with their physician in the exam room, and it’s easy to forget what to ask in the moment.” An AI assistant can help patients list the questions they need to bring to the appointment.

The article offers four sample prompts that patients can tailor with personal details such as age, gender and medical history:

1. I have an upcoming annual physical, what are five questions I should ask my doctor to get the most out of my appointment? 2. I am experiencing these symptoms—or I am going in for my annual physical. Can you tell me what questions my physician may ask me so that I can prepare now and give the most complete answers possible? 3. Can you create a symptom tracker for me that I can then share with my physician? 4. Here is data from my wearable device. Is there anything my physician should know based on the data?

The third prompt is especially useful for patients with chronic or episodic symptoms. A chatbot can generate a spreadsheet that tracks pain location, severity, time of day, bowel movements and stool appearance, giving patients a richer record to bring to the doctor.

Robert M. Wachter, MD, an AMA member and author of A Giant Leap: How AI Is Transforming Healthcare and What That Means for Our Future, cautions that chatbots are not yet ready to dispense definitive medical opinions. “As long as you are not using it to do anything more than say ‘Is there anything here that I should be concerned about?’ or ‘Anything my doctor might want to know?’ or ‘Any questions that I might want to ask my doctor based on this?’ I think that is reasonable,” Wachter said. He added that asking the chatbot to “play doctor” based on wearable data has been a little sobering.

Dr. Wachter, who chairs the medicine department at the University of California, San Francisco, emphasizes the importance of context. “If you put into a tool ‘I’m short of breath’ and it knows nothing about you, it probably gives you a different answer than if you say ‘I have a history of blood clots,’ or if you say ‘I have diabetes,’ or if you say ‘I’m taking so‑and‑so medicine,’” he explained. The more the chatbot knows about a patient’s background, the better its responses.

The AMA has issued guidance on how patients can use AI chatbots safely and efficiently. The organization cautions that users should not share personal or identifying details because the chatbots are not governed by the same privacy safeguards that protect physician offices. The AMA’s Center for Digital Health and AI, which was recently launched, aims to give clinicians a voice in shaping how AI and other digital tools are used to improve patient and clinician experience.

In short, AI chatbots can help patients prepare for visits, create symptom trackers and interpret wearable data, but they are not a substitute for a medical professional. Patients should use the technology as a supplement, not a replacement, and should follow the AMA’s privacy recommendations. As the technology matures, clinicians and patients alike will need to balance the convenience of instant answers with the need for accurate, personalized care.