Across the country, Americans are quietly adding a new name to their list of go-to health resources: the chatbot. A new national poll finds that roughly one in three adults has turned to artificial intelligence for health information in the past year, often because seeing an actual doctor feels too expensive, too slow, or both.
According to KFF, its Tracking Poll, fielded Feb. 24 to March 2, 2026, among 1,343 adults, found that about 32% of U.S. adults used AI chatbots for health guidance in the past year. The survey breaks that into about 29% who asked about physical health and 16% who asked about mental health, and it suggests many people were looking for quick, on-demand answers rather than formal medical care.
The trend hit home in Northeast Ohio, where Cleveland.com reporter Julie Washington highlighted how rising health care costs and long waits for appointments are nudging people toward AI tools. Her coverage framed chatbot use as a symptom of broader strain in the health care system, especially for younger adults and residents without stable insurance.
KFF’s data also show that about 41% of AI users have uploaded test results or doctors’ notes into a chatbot to get personalized explanations, even as roughly three-quarters of the public say they worry about the privacy of medical data shared with AI tools. The poll found that 58% of people who asked chatbots about mental health did not later see a doctor, and 42% who asked about physical health likewise did not follow up with a clinician, according to KFF.
Health systems urge caution
Local health systems and frontline clinicians are not thrilled about patients swapping in a chatbot for a checkup. “No chatbot or large language model (LLM) is a substitute for expert medical care,” the Cleveland Clinic warns in a primer on clinical AI. At the same time, the Clinic and other major systems are piloting EHR-linked chatbots that limit answers to vetted clinical data, a strategy experts say is key to cutting down on misinformation and risky self-triage. Cleveland Clinic describes these internal tools as a safer way to give patients quick, context-aware information without letting the AI “freelance” outside trusted sources.
How to use chatbots more safely
Physicians and consumer guides advise treating chatbot answers like a starting point, not a final diagnosis. That means stripping out personal identifiers before pasting in test results, avoiding requests for medication decisions or emergency-care calls, and bringing any worrying response to a real-world clinician. As Forbes explains, patients should share only the minimum necessary information and cross-check chatbot output against trusted health sites or a provider’s guidance. Many health systems are also experimenting with integrated chatbots constrained to verified content and the patient’s own record, in an effort to keep convenience without inviting chaos.
Policy questions are already here
Lawmakers have noticed that “Dr. Algorithm” is suddenly doing a lot of unofficial office visits. Policy proposals in Ohio and other states would require chatbots to display safety warnings, limit access for minors, and in some cases penalize platforms whose models appear to encourage self-harm. Cleveland.com reported on an Ohio bill that would impose escalating penalties on developers whose models prompt harmful behavior, underscoring how quickly regulation is racing to catch up with consumer habits.
The KFF numbers highlight a basic tension: AI can lower the barrier to information, but it also opens up follow-up gaps and privacy risks for the very people who lean on it most. As the technology and the rules around it evolve, local health systems say the safest path is to keep AI embedded inside clinical workflows and to keep human clinicians firmly in the loop.
