
The AI Health Answer You Want Could Be Wrong
Can you trust ChatGPT with medical advice? Dr. Michael Turken and I explain how to check AI health answers, verify their sources, spot missing context, and decide what to do when AI and your doctor disagree. Dr. Michael Turken, MD, MPH, is a practicing internist at UCSF and the founder and CEO of My Doctor Friend, an AI health assistant for patients and caregivers navigating complex care. He previously worked in health technology at Google, Forward Health, and Omada Health. Michael and I trained together in internal medicine at Stanford and have been friends ever since. Chapters: 00:00 Can You Trust AI With Your Health? 01:49 From Dr. Google to AI 03:32 Why Patients End Up Managing Their Own Care 05:16 Three Ways AI Can Get It Wrong 06:31 Confabulation and Hallucinations: When AI Makes Things Up 10:01 AI Memory and Context Limits 13:08 Sycophancy: When AI Tells You What You Want to Hear 15:52 AI, Medical Triage, and Real-World Risks 19:35 Building a Harness Around AI 22:15 Using AI to Navigate Healthcare 27:10 Practical Ways to Get Better AI Answers MICHAEL TURKEN, MD, MPH + MY DOCTOR FRIEND My Doctor Friend: https://mydoctorfriend.ai/r/lin Offer: Use the referral link or code “lin” for one month of premium access, with no redemption deadline. After that, continue with the free plan (10 messages per month) or choose a paid plan. Learn more about My Doctor Friend: https://about.mydoctorfriend.ai Disclosure: This channel may receive compensation if you make a paid purchase through the My Doctor Friend link or code, at no added cost to you. YouTube: https://www.youtube.com/@DrMichaelTurken LinkedIn: https://www.linkedin.com/in/michaelturken/ Sovereign Care: https://michaelturken.substack.com/ WE DISCUSS AI hallucinations and real citations attached to wrong answers What AI remembers about your medical history and what it misses Sycophancy: when a chatbot agrees with you too readily What to check when AI and your doctor disagree Why patients managing complex care need better tools THREE PROMPTS FROM THE CONVERSATION “Why could you be wrong?” “What do you know about me?” “Don't guess. Look it up.” Compare an AI-generated health summary with your current records and open the sources yourself. Agreement between models doesn't prove an answer is right. Subscribe for more conversations about health, AI and longevity. More from Dr. Hillary Lin: https://hillarylinmd.com/
- Transcript










