Research institute · Pew Research Center
34% of US Adults Use AI Chatbots for Health Information
A third of US adults now put health questions to an AI chatbot, and nearly all of them find the answers useful. The trust gap sits somewhere else: only 29% are comfortable giving those tools their personal health data.
Source published Summarised
Key findings at a glance
- 34% of US adults use AI chatbots for at least one health or medical reason
- 28% use chatbots because they get health information quickly
- 25% use chatbots to work out what is causing their symptoms
- 22% use chatbots because the information costs little or nothing
- 20% use chatbots to help make sense of their lab results
- 18% use chatbots for health topics they are uncomfortable raising with a person
- 47% of chatbot health users find the information extremely or very helpful
- 5% of chatbot health users find the information not too or not at all helpful
- 29% of chatbot health users are comfortable sharing personal health data with a chatbot
- 26% of chatbot health users are not comfortable sharing personal health data
- 44% of US adults aged 18 to 29 are chatbot health users
- 17% of US adults aged 65 and over are chatbot health users
What the report covers
Pew Research Center surveyed 3,488 US adults in late June 2026 on how they use AI chatbots for health and medical questions. The study covers eight specific reasons people turn to a chatbot, ranging from working out what is causing a symptom to understanding a diagnosis a doctor has already given them, then asks how helpful the answers are and how comfortable users feel sharing personal health details with the tool.
The headline number is 34%. That is the net share of US adults who use chatbots for at least one of the eight reasons Pew asked about, and Pew notes that nearly all of those users cite more than one. Speed leads at 28%, followed by self-diagnosis of symptoms (25%), low or no cost (22%), understanding a doctor's diagnosis (22%) and interpreting lab results (20%). A second chapter asks a harder question: whether chatbots help or hurt people who use them for loneliness, depression or stress. On balance Americans say they hurt, by 39% to 19% on loneliness and 36% to 17% on depression.
It builds on Pew's wider 2026 AI programme, which we covered in our report on chatbot adoption reaching half of US adults. This one is narrower and more revealing, because it shows what people actually ask a chatbot to do once the novelty has worn off, in a category where a wrong answer carries real cost.
What this means for UK marketers
The health framing makes this easy to file under "not my sector". That would be a mistake. What Pew has measured is a behaviour rather than a vertical: people now route high-consequence questions through a chatbot before they speak to a professional, because it is fast (28%), cheap or free (22%) and does not judge them (18% use chatbots for things they are uncomfortable raising with a person). Every one of those motivations transfers cleanly to a UK buyer researching a mortgage, a legal claim, a B2B software purchase or a builder.
Three things follow. First, the confidence gap is the opportunity. Only 47% of users call the information extremely or very helpful, yet just 5% call it unhelpful, so most people sit in the lukewarm middle. That middle is exactly where clear, well-structured, verifiable content gets cited and clicked, which is the practical case for treating AI visibility as a channel rather than a curiosity. It is increasingly what our SEO work for UK businesses is built around.
Second, the audience skew is real but it is not mainly about age. Chatbot health use runs at 44% among 18 to 29 year olds against 17% of the over-65s, a wide gap. The attitudinal split is wider still: 60% among adults who feel more excited than concerned about AI, against 21% among those who are more concerned. Enthusiasm predicts channel use better than demographics do, which means the people least likely to see your AI-surfaced content are the ones already sceptical of it. We saw the same fault line in Pew's data on under-30s growing warier of AI.
Third, treat the privacy figure as a hard ceiling. Only 29% of these users are comfortable sharing personal health details with a chatbot and 26% are actively uncomfortable, so any brand designing chatbot intake, triage or advice tooling is building against the grain of its own audience. If the tool edges into diagnosis or treatment it may also fall under the MHRA's rules on software and AI as a medical device. Answer the regulatory question before the marketing one.
How they got the data
Pew surveyed 3,488 US adults between 22 and 28 June 2026 as Wave 195 of its American Trends Panel, a nationally representative panel recruited by address-based sampling from the US Postal Service delivery file. SSRS ran the fieldwork, with 3,362 interviews completed online and 126 by live telephone, in English and Spanish. The margin of sampling error for the full sample is plus or minus 1.8 percentage points.
The survey-level response rate was 86% of the 4,069 panellists sampled, though the cumulative rate accounting for recruitment non-response and attrition is 3%. Non-Hispanic Asian adults were deliberately oversampled to give more precise subgroup estimates, then weighted back to their true population share. Two respondents were removed before weighting after data-quality checks for satisficing. Full detail sits in Pew's American Trends Panel methodology.
Research institute
Read the full report
From Pew Research Center · published 25/08/2026