If you've typed your symptoms into ChatGPT before calling a doctor — or after, trying to figure out what the diagnosis actually meant — you're in a 34% majority now.

Pew Research surveyed 3,488 US adults in late June 2026 and found that a third of Americans now use AI chatbots for at least one health-related task. The most common: fetching quick health information (28%), understanding what's causing symptoms (25%), getting access to health information at low or no cost (22%). A quarter use AI specifically to try to figure out a diagnosis. Similar shares use it after a doctor's visit — to understand test results, make sense of what was prescribed, decode the explanation they got in a 12-minute appointment.

This is not a niche behavior anymore. It is the way a third of the country is navigating the health system.

Source spread

• Pew Research — "From Diagnoses to Treatments: Why Americans Use AI Chatbots for Health" [academic] — primary source; 3,488 adults surveyed June 22–28, 2026; all the usage and helpfulness data.
• Pew Research — "Do Americans Think Chatbots Help or Hurt People Using Them for Loneliness, Depression or Stress?" [academic] — companion report; the mental health harm/help percentages.
• KFF — "Poll: 1 in 3 Adults Are Turning to AI Chatbots for Health Information" [academic] — separately confirms the Pew figure and compares to social-media health usage.
• Healthcare Dive — "Most health AI users don't rate chatbots as highly accurate" [skeptic] — the accuracy-versus-helpfulness divergence.

Pros & cons

What this data confirms:

• AI has moved past "early adopters" for health information. 34% in a nationally representative survey is not a niche. It is a mainstream behavior that policy and providers are not fully ready for.
• The cost-access angle matters. 22% cited "low or no cost" as a reason. Not everyone has a primary care doctor they can call with a quick question. AI filling that gap is not imaginary.
• The under-30 adoption rate — 44% of adults under 30 use AI for at least one health task — tells you this behavior is going to grow, not shrink, as that cohort ages.

What deserves careful attention:

• Only 29% of AI health users are very comfortable sharing personal health data with AI tools. 47% call the answers very helpful. Those two numbers coexist. People are getting value from AI health advice while being uncomfortable with the privacy implications of that same tool knowing their health information. That's not a stable equilibrium.
• The mental health data is the one I'd flag loudest. For loneliness: 39% of Americans say AI chatbots do more to hurt the people using them, versus 19% who say they help. For depression: 36% harm, 17% help. That's a two-to-one ratio saying AI makes things worse in exactly the use case where people are most vulnerable.

Samwise's take

What builders need to know

• Health AI is now mainstream, which means the liability question is real. Any product that intersects with health queries is getting questions it wasn't designed to handle. Have a clear answer for when your AI routes to "please consult a healthcare professional."
• The comfort-vs-use gap is a design signal. Users are using health AI despite privacy discomfort. That creates both a trust debt and an opportunity: the tools that build genuine transparency about data handling will differentiate. The ones that ignore it will lose trust when something goes wrong.
• Mental health use is growing fastest among the highest-risk demographic. Under-30 adoption at 44% includes a population with high rates of depression and anxiety. If you're building in this space, the 39% "AI hurts people using it for loneliness" finding is not background noise.
• Asian Americans at 56% adoption — the highest of any group surveyed — suggests health AI is doing something specific for communities navigating language, cultural, or access barriers. That use case deserves direct product attention.

Further reading

• Pew Research — "From Diagnoses to Treatments: Why Americans Use AI Chatbots for Health" — the primary survey and all usage percentages
• Pew Research — "Do Americans Think Chatbots Help or Hurt People Using Them for Loneliness, Depression or Stress?" — the mental health companion report
• KFF — "Poll: 1 in 3 Adults Are Turning to AI Chatbots for Health Information" — corroborating data from KFF
• Healthcare Dive — "Most health AI users don't rate chatbots as highly accurate" — the accuracy versus convenience distinction