Poll: Most Who Use Artificial Intelligence Doubt AI Chatbots Provide Accurate Health Information

Amid rising interest in and use of artificial intelligence (AI) by individuals and businesses, most of the public (63%), including most AI users (56%), are not confident that AI chatbots provide accurate health information, a new KFF Health Misinformation Tracking Poll finds.

The poll comes as AI chatbots such as ChatGPT, Google Gemini, and Microsoft CoPilot have become widely available and public use has risen. About one in six (17%) adults now say they use such chatbots at least once a month to find health information and advice. That includes a quarter (25%) of those under age 30.

"While most of the attention around AI in health is focused on how it can transform medical practice and create new business opportunities, consumers are also using it, and the jury is still out on whether it will empower or confuse them," KFF President and CEO Drew Altman said. "At KFF, our focus will be on how AI and other information technologies affect people."

Other findings include:

• AI users' trust in chatbot responses varies based on the type of information provided. For example, most users say they trust chatbots' responses related to practical tasks (66%) and technology (61%) at least a fair amount. Far fewer say so about responses related to health (36%) and politics (24%).
• When asked about AI chatbots generally, most of the public (56%) say they are not confident that they can tell what information is true and what is false in their responses. Even among those who use AI, half say they aren't sure they can tell fact from fiction.
• For most of the public, the verdict is still out on whether they believe AI mostly helps or hurts people trying to find accurate health information. Similar shares say it mostly helps (21%) or mostly hurts (23%), with a majority (55%) saying they are unsure.

Designed and analyzed by public opinion researchers at KFF. The survey was conducted June 3-June 24, 2024, online and by telephone among a nationally representative sample of 2,428 U.S. adults in English and in Spanish. The margin of sampling error is plus or minus 3 percentage points for the full sample. For results based on other subgroups, the margin of sampling error may be higher.

Highlights from the poll will be featured in an upcoming edition of the KFF Health Misinformation Monitor, a twice-a-month briefing that tracks the evolution and spread of health misinformation. Both the poll and the monitor are part of KFF's new Health Misinformation and Trust Initiative.