Patients Turn to AI for Medical Advice as Hospitals Deploy Their Own Chatbots

Americans are increasingly using AI chatbots for health information, with one in three adults having consulted one for medical guidance, according to a KFF poll from March 2026. Health systems across the country are now rolling out their own branded chatbots—including PatientGPT through Hartford HealthCare in Connecticut—to capture this demand and offer what they frame as safer alternatives to commercial versions.

Hartford HealthCare partnered with clinical AI company K Health to deploy PatientGPT to tens of thousands of existing patients, positioning the move as a way to meet patients where they are and provide digital equity. Allon Bloch, K Health's CEO, stated that the chatbot would operate "inside a health system that connects to your medical records and your care team," distinguishing it from unaffiliated tools patients currently use.

However, a February 2026 study published in Nature Medicine involving nearly 1,300 participants revealed significant accuracy gaps. When researchers provided LLMs with structured medical scenarios, the models correctly identified diagnoses 95 percent of the time and recommended appropriate next steps 56 percent of the time. But when participants used their own language to describe the same scenarios, diagnosis accuracy dropped to roughly 33 percent and appropriate next-step guidance fell to 43 percent. Andrew Bean, lead author and AI researcher at Oxford University, told NPR that the findings showed "people don't know what they are supposed to be telling the model."

Among Americans who used AI for health information, 41 percent uploaded personal medical data like test results. When asked their primary reasons for turning to AI, 19 percent cited inability to afford care, 18 percent said they lacked a regular provider or couldn't get appointments, and 65 percent simply wanted quick answers. A significant portion did not follow up with doctors: 58 percent of those who asked about mental health and 42 percent who asked about physical health never consulted a physician afterward.

Adam Rodman, an internist and clinical reasoning researcher at Beth Israel Deaconess Medical Center in Boston, expressed caution about hospital chatbot rollouts. "It's a tempting idea," Rodman told Stat News, but "there isn't yet evidence to show that integrating chatbots into health systems improves patient outcomes."

Context

The US health care system has long underperformed compared to other high-income nations. Americans have lower life expectancy, higher rates of avoidable deaths, and worse maternal and infant mortality outcomes than peers in comparable countries. A 2023 report found that nearly a third of Americans—more than 100 million people—lack a primary care provider, creating a gap that AI tools have begun to fill.

This gap explains why Americans are turning to chatbots at scale. The combination of cost barriers, appointment scarcity, and convenience has driven adoption faster than evidence of effectiveness has accumulated. Health systems view branded chatbots as a way to retain patients and address demand that already exists, rather than as a solution to underlying access problems.

What's Next

The critical question ahead is whether hospitals will implement sufficient monitoring and liability frameworks as these tools expand. Rodman's observation that "we're not there yet" in terms of outcome evidence suggests that health systems are deploying chatbots based on demand and competitive positioning rather than clinical validation. Regulators have not yet established clear standards for hospital-deployed AI health tools, leaving questions about accountability and patient safety unresolved as rollouts accelerate across the country.