One reporter’s notebook: Like many journalists this week, I demoed ChatGPT Health, which just debuted. The new tool is extremely useful, and made recommendations on chronic health issues that, quite frankly, I was never going to visit a doctor about. But I wonder if maybe it is going a little too far into what is basically the practice of medicine.

Chrissy and I debate the good and bad of AI all day long, so be sure, the below are my thoughts only on ChatGPT’s Health. She has different views, which is part of the fun!

This week OpenAI launched Health, which allows users to connect their medical records and Apple Health data to the platform. I could only connect my fitness data from Apple Health and a hospital record from when I gave birth. Still, the models it uses are trained for the task of making health recommendations. OpenAI works with “more than 260 physicians across 60 countries, 49 languages, and 26 medical specialties” to train models for Health.

The good: I tested Chat on a few things. Serendipitously, I was sick. I was convinced I had West Nile virus. Chat was not so sure. The app told me it was more likely a viral infection, strep throat, COVID-19, influenza, or another infection. It also suggested I might have meningitis at one point and urged me to go to a doctor. I reassured Chat that my symptoms were improving (I’m fine). 

“If you're not mostly back to normal within about a week, or if the swollen lymph nodes remain enlarged after the rest of the illness has resolved, it would be worth making an appointment with your primary care clinician to be examined,” it told me. 

I also queried it about my fitness routine to see if it’s good enough and what other steps I could take to ensure that I actually get around to doing it (my most persistent problem). It suggested that rather than trying to work out at a specific time of day, I think about placing the work between work tasks. A worthwhile suggestion.

Finally, we chatted about my sleep issues. I’ve had insomnia, though I’ve never been officially diagnosed with it, since I was a child. I have never thought to mention this to my primary care doctor, because it comes in clusters and isn’t usually happening during my annual exam. I had a recent spell of it, so I told Chat about when it started for me, the circumstances under which it started, and how it was manifesting in adulthood. 

Chat said I have chronic sleep-maintenance insomnia. The app gave me several suggestions for what to do when I have these wake-ups, but it repeatedly suggested I look into Cognitive Behavioral Therapy for Insomnia. Then it proceeded to demo what a session would look like, replete with therapist-patient banter. 

The questions: So what do I think of this experience? I honestly don’t know how to evaluate the information I was given. Was it useful? Absolutely. Is it accurate? I don’t know.

My sense is that though ChatGPT can be really helpful, especially for people with longstanding complex health issues that doctors have not been able to figure out, it overreaches when it comes to health queries. It offered answers to questions about my health problems and though it did encourage me in more than one instance to seek out a doctor, it also gave me enough information to stay home. It repeatedly told me I probably just have a regular virus. Even with the suggestion that I seek out CBT-I, it gave me a good indication of how I could do it myself if I was motivated. 

This overreach might be fine, if there were guardrails in place, particularly around accuracy. But the federal government hasn’t regulated this technology yet. And allegedly, some people are potentially experiencing harms. This week a man sued OpenAI for repeatedly misdiagnosing him and assuring him he didn’t have a serious illness.

Final thoughts:  As many doctors and even health tech founders know, there are a lot of barriers to seeing a doctor. When ChatGPT makes it this easy to believe you’ve found the answer, it could represent one more thing that will keep people from seeing a professional. 

These are just my initial thoughts; I welcome yours. Now onto the news. 

—Ruth

Co-hosted by Christina Farr and Stephanie Davis, this episode dives into the biggest debates shaping digital health right now. Keaton Bedell, co-founder and CEO of Bridge, joins to unpack an exclusive data report on where virtual care is headed next, and Joe DeVivo, President, CEO, and Chairman of Butterfly Network, sits down to talk about the Midjourney partnership that's redefining the future of full-body imaging.