328
Audio & Video Production320
Software Development244
Automation & Workflow210
AI Infrastructure & MLOps151
Marketing & Growth193
Writing & Content Creation200
Data & Analytics121
Design & Creative148
Customer Support123
Photography & Imaging141
Voice & Speech132
Sales & Outreach113
Operations & Admin86
Education & Learning121
More Americans are using AI chatbots for health questions. Hospitals are responding with their own tools, but experts say proof of benefit is still limited.
In short: As more Americans turn to AI chatbots for health information, hospitals are launching their own versions inside patient portals.
Many Americans now ask large language model chatbots (AI tools that predict the next words in a sentence, like a very advanced autocomplete) for health advice. A recent KFF poll found that 1 in 3 US adults have used an AI chatbot for health information. Among those users, 41 percent said they uploaded personal medical details, such as test results.
Hospitals and health systems see this behavior and want to guide it. Hartford HealthCare in Connecticut, working with K Health, is expanding a chatbot called PatientGPT to tens of thousands of existing patients. PatientGPT can answer general questions, and it can also collect symptoms and suggest next steps, including booking an appointment or telling someone to seek urgent or emergency care.
Other systems are rolling out Epic’s Emmie inside MyChart, which many people already use to view test results and message their doctor. Sutter Health says Emmie can answer general questions and summarize information already in a patient’s chart, but it does not give personalized medical advice or make care decisions.
Experts warn that the benefits are still not proven. Adam Rodman, a physician and researcher, told Stat News there is not yet strong evidence that adding chatbots to health systems improves patient outcomes. Accuracy is also a concern, especially because real people may not know what details to include, like telling a nurse only half your symptoms (it changes the whole picture). A Nature Medicine study found performance dropped sharply when participants wrote their own prompts, compared with scripted scenarios.
Source: Arstechnica