On October 30, 2025, Simar Bajaj (The New York Times) reported that 1 in 6 adults in the U.S. regularly asks AI chatbots about their health, and this rate is increasing. ChatGPT can even pass medical licensing exams, but experts warn that AI is still prone to “making things up,” sometimes causing real harm to patients.

Dr. Ainsley MacLean advises that AI is useful for drafting questions for your doctor, explaining medical terms, or summarizing a diagnosis, but it should not be used for self-diagnosis or to replace a doctor.

5 tips suggested by experts for using AI safely and effectively in healthcare:

1. Practice with low stakes: Don’t wait until you’re seriously ill to ask an AI. Try asking small questions you’ve had in past appointments to understand the chatbot’s strengths and limitations. Avoid leading questions like, “Should I get an MRI?” as AI tends to “please” the user.
2. Share context — but with limits: Providing your age, medical history, and current medications helps the AI give more accurate answers. However, avoid sending personally identifiable information like your address, Social Security number, or full medical records, as AI is not HIPAA-compliant. You can use incognito mode or HIPAA-compliant chatbots like My Doctor Friend or Counsel Health.
3. Check in periodically during long conversations: AI can easily “forget” or confuse facts. It’s better to use a paid version with a longer memory and frequently ask it to summarize what it knows to detect inaccuracies.
4. Encourage the chatbot to ask questions: AI rarely asks follow-up questions like a real doctor. Prompt it with:”Ask me any additional questions needed to make a safe diagnosis.” Providing feedback helps the AI create a list of differential diagnoses, but remember that AI lacks the clinical experience to assess their true probability.
5. “Cross-examine” the AI itself: Ask for its sources, request explanations for its reasoning, then tell the AI to critique itself and compare its previous answers. You can try asking it to “act as a general practitioner” or a “specialist” to see the difference. However, always verify information with a real doctor or a reputable medical source.

📌 5 tips suggested by experts for using AI safely and effectively in healthcare: 1. Avoid leading questions like,”Should I get an MRI?” as AI tends to “please” the user. 2. Share context — but with limits: Providing your age, medical history, and current medications helps AI answer more accurately. However, avoid sending personally identifiable information. 3. Use a paid version with a longer memory and frequently ask for summaries to detect inaccuracies. 4.Encourage the chatbot to ask questions, such as, “Ask me any additional questions needed for a safe diagnosis.” 5. Ask for sources, “Ask me any additional questions needed for a safe diagnosis.” 5. Ask for sources, request explanations, and then have the AI critique itself and compare previous answers.