Nearly half of adults surveyed in 13 countries, including the United States, believe that someone without medical training who knows how to use artificial intelligence can perform at least one health task as well as, or better than, a trained health care professional. The finding comes from a new special analysis by the Edelman Trust Institute and the Yale School of Public Health.
The result measures belief, not performance. No AI tool or AI user was tested for accuracy. The same survey found that 7 in 10 respondents believe at least one of six divisive health claims, including the claim that taking acetaminophen during pregnancy can lead to autism. For patients who consult chatbots before calling a doctor, that combination matters.
The report comes as U.S. clinicians face falling confidence in routine prevention. A Reuters/Ipsos poll found that 76% of U.S. adults say the measles, mumps, and rubella (MMR) vaccine is safe for children, down from 84% in 2020, CIDRAP reported. Meanwhile, the CDC has confirmed 3,294 measles cases this year as of Sept. 10.
Confidence in AI Health Skills Runs Highest Among Younger Adults
The analysis surveyed 12,998 people online between Feb. 28 and March 11, 2026, with about 1,000 respondents per country, including 995 in the United States. The AI question was asked of half the sample. Participants chose which of four tasks, if any, a person without medical training who knows how to use AI could do as well as, or better than, a trained professional.
About 26% chose deciding whether someone needs to see a health care provider, 19% chose performing basic medical procedures, 19% chose determining the proper treatment or medication, and 16% chose diagnosing an illness. Overall, 49% selected at least one task.
Belief was strongest among adults ages 18 to 34, at 59%, compared with 52% of those ages 35 to 54 and 35% of adults 55 and older. It was also higher among people with university-level science education, at 56%, than among those without it, at 42%. The report pages MedicalDaily reviewed did not break out this result for U.S. respondents alone.
Conflicting Advice Is Reshaping Everyday Health Decisions
Among respondents who had received conflicting recommendations from a doctor and another source, 51% said they always followed the doctor. Another 34% said it depended on the situation, and 15% said they always followed the other source.
Divisive beliefs track closely with reliance on other voices. Among respondents who believed three or more divisive claims, 46% said they had let an AI platform influence their health decisions, compared with 13% of those who believed none. In the same group, 52% said they had followed medical advice from social media over their clinician's guidance in the past year, compared with 15% of those who believed none.
Respondents who interacted more frequently with the health care system were more likely to hold many divisive beliefs, a pattern the report describes but does not explain. More than half said they want clinicians who listen first to understand their situation (58%), welcome doubts without judgment (57%), and help them work through their concerns (55%). The authors urged health care providers and communicators to "move beyond persuasion and invest in the experiences, relationships, and community presence that build trust and ultimately shape health decisions."
A Survey of Beliefs, Not a Test of Chatbots
MedicalDaily Evidence Check: This is a cross-sectional online survey of 12,998 adults, published as an Edelman Trust Barometer special analysis rather than a peer-reviewed journal article. It captured what people believe about AI and health; it did not measure whether AI tools give correct answers. The report notes that online samples in countries with lower internet access tend to skew younger, more urban, and more affluent. Its associations do not show that AI use causes mistaken beliefs. Edelman is a communications firm, and the report pages MedicalDaily reviewed did not include a funding statement.
Separate research points to real limits. MedicalDaily recently reported on a study in which free AI chatbots dropped correct sleep apnea referral advice in about a third of conversations when simulated patients pushed back. That study used scripted cases rather than real patients, but it shows how a chatbot can bend toward what a user wants to hear.
The people most likely to be affected include younger adults, members of online health communities, and patients who manage chronic conditions between appointments. Parents weighing childhood vaccines face particular risk from bad information, and the CDC has counted 38 new measles outbreaks in the U.S. this year.
Practical Ways to Use AI Without Replacing Your Clinician
AI tools can help people prepare questions, understand medical terms, or organize notes before a visit. They cannot replace an exam, lab tests, or a clinician who knows a patient's history. Readers can check AI answers against trusted sources such as the CDC, the FDA, or their own provider, and they should be cautious when a chatbot agrees with everything they suggest.
Do not start, stop, or change a prescription based on chatbot advice. Chest pain, trouble breathing, sudden weakness or confusion, and severe allergic reactions call for 911. Anyone having thoughts of self-harm can call or text 988 for immediate support.
Cost and convenience are part of the appeal. In the Edelman report, easy access without paying or waiting for an appointment ranked among the top five things people said they get from health voices without formal credentials. People without insurance can look for community health centers that charge on a sliding scale, and many insurers and hospital systems offer nurse advice lines or telehealth visits.
The survey shows how much faith many people place in AI, but it does not show that the faith is earned. AI works best as a starting point for a conversation with a qualified clinician.
Key Questions Answered
What did the survey find about AI? About 49% of respondents in 13 countries said a person skilled with AI but without medical training could do at least one of four health tasks as well as, or better than, a trained professional. Deciding whether someone needs care was the most common choice, at 26%.
Does this mean AI is as accurate as a doctor? No. The survey measured beliefs only, and no AI tool was tested for accuracy.
Who was surveyed? The survey reached 12,998 adults online from Feb. 28 to March 11, 2026, including 995 in the United States.
What divisive health claims were included? The claims involved acetaminophen use in pregnancy and autism, raw milk, childhood vaccine risks, animal versus plant protein, fluoride in drinking water, and vaccines as population control. About 70% believed at least one.
How can people use AI tools more safely? Use AI to prepare questions, verify answers with trusted health sources, and never change medications without a clinician. Seek emergency care for severe symptoms.
.png?w=600)
