Some US doctors are using ChatGPT to communicate more empathetically with patients, including when breaking bad news and explaining medical recommendations
Gina Kolata / New York Times : Tweets: @davidgratzer , @_eric_reinhart , @josourcing , and @sarahdigregorio Tweets: David Gratzer / @davidgratzer : “Most surprising to Dr. Lee, though, was a use he had not anticipated - doctors were asking ChatGPT to help them communicate with patients in a more compassionate way.” This is a must-read article on AI. https://www.nytimes.com/... Eric Reinhart / @_eric_reinhart : Machines are helping doctors become more like humans. https://www.nytimes.com/... Nicole Miller / @josourcing : Thanks to Greg's stupidity, every doctor named in this article will be reported to the FTC and other relevant health law monitoring agencies. https://twitter.com/... Sarah DiGregorio / @sarahdigregorio : Good lord, if you need AI to tell you to say something like “I know this is a lot of information to process and that you may feel disappointed...” when delivering news about incurable cancer, then we need med school to deal w empathy, not outsource to AI. https://www.nytimes.com/...
Context & Ripple Effects
The story lands mid-arc in a fast-moving sequence of clinical AI adoption. Earlier in 2023, US hospitals were already testing GPT-3 to draft replies to patient messages, with a study claiming the chatbot's first-version answers beat doctors', and weeks later doctors named documentation relief as generative AI's best use given the hours it consumes daily and its role in burnout. The new wrinkle is direction: rather than replacing clerical work, doctors like Dr. Lee are asking ChatGPT to make them sound more human when breaking bad news — a use commentators did not anticipate.
It matters because it moves generative AI from the back office into the most sensitive channel in medicine, and the backlash arrived immediately: critics including Eric Reinhart argue empathy should be taught in medical school, not outsourced, and some observers are flagging such use to regulators.
First-order effects
- US doctors adopting ChatGPT for compassionate communication gain a drafting tool for bad-news conversations and treatment explanations, while critics respond by arguing the skill belongs in medical curricula and reporting the practice to regulators.
Second-order effects
- Hospitals that had only justified AI on efficiency grounds — cutting reply times on patient queries and easing the documentation burden behind clinician burnout — now face governance decisions over AI-drafted emotional content they never explicitly approved.
Third-order effects
- If the pattern holds, the doctor-patient message stream becomes substantially machine-composed from both ends — clinicians drafting with ChatGPT while patients independently feed years of medical records into chatbots despite privacy risks and unreliable diagnoses — forcing disclosure norms for a channel where, as with earlier AI-powered decision support, most patients remain unaware AI is involved at all.
The trend: Generative AI in medicine is migrating from administrative relief into the empathetic front line of doctor-patient communication, outpacing both medical training norms and regulatory clarity.