People who regularly turn to artificial intelligence chatbots for health advice are significantly more likely to believe false claims about vaccines, a major US survey has found – prompting fresh concerns among UK health experts about the rise of an “unregulated AI healthcare system” alongside the NHS.
The poll, conducted in May by the health research firm KFF and published on Tuesday, surveyed a representative sample of 2,480 US adults. It found that 35% of those who use AI tools to find health information at least once a week believe it is “definitely or probably true” that the measles, mumps and rubella (MMR) vaccine causes autism. Among US adults who never use AI for health advice, only 20% hold the same belief. The pattern held even when the researchers controlled for age, race, education and political partisanship.
The correlation extended to other widespread anti-vaccine myths. The KFF poll found that 29% of frequent AI users for health believe mRNA vaccines can alter a person’s DNA – a claim that is not true – compared with 20% of non-users. Similarly, 22% of regular AI health seekers think the measles vaccine is more dangerous than the measles virus itself, while only 15% of those who never use AI for health share that view.
Social media and demographics
Turning to social media platforms for health advice was also linked to a higher likelihood of believing misinformation, the survey found. KFF reported that adults who use social media for health information at least weekly are more than twice as likely as those who do not to say the MMR–autism myth is probably or definitely true – 37% compared with 16%.
The poll revealed a demographic split in how people seek health information online. Lower-income groups and adults without a college education are more likely to consult social media, while people in households earning more than $90,000 a year or who hold a university degree tend to turn to AI tools. Additional research cited in the briefing notes that younger adults, Black and Hispanic adults, and those without a college degree are also more likely to use social media for health information, whereas adults aged 30–49 are more likely to use AI. The KFF poll also found that individuals who do not have a trusted healthcare provider are more likely to believe vaccine myths.
Why AI might spread vaccine misinformation
The KFF survey did not ask respondents which specific AI models they used, but the quality and accuracy of chatbot health advice varies widely. Each chatbot carries biases from its training data and the decisions its developer makes about how to handle contentious questions. A study published in Nature Medicine found that large language models present risks because they often provide inaccurate and inconsistent information, leading to what researchers described as a “communication breakdown” with users. An audit of five popular chatbots concluded that nearly half of their health-related responses were problematic, with the chatbot Grok generating a significantly higher number of problematic responses than its competitors.
AI companies have acknowledged the volume of health queries they handle. In a January blog post announcing the creation of a specialised ChatGPT Health tool, OpenAI stated: “Health is already one of the most common ways people use ChatGPT, with hundreds of millions of people asking health and wellness questions each week.” The reliance on chatbots for medical advice continues a longer-standing pattern in how people use search engines: about 5% of all Google searches concern health, and roughly 77% of people use search engines to look up information about a new diagnosis, according to a 2025 research paper by a Georgetown University researcher.
Concerns about AI amplifying health misinformation are not limited to the United States. In the UK, a study found that one in seven people use AI chatbots for health advice instead of seeing a GP, with a quarter of those users citing long NHS waiting lists as the reason. Experts from King’s College London and the London School of Hygiene & Tropical Medicine have warned about the accuracy and safety of AI in healthcare, describing a growing “unregulated AI healthcare system” operating alongside the NHS. The UK government is aware of the threat, particularly from AI-generated deepfake videos that impersonate medical professionals, and is leveraging the Online Safety Act to require social media platforms to remove harmful content. Meanwhile, the Medicines and Healthcare products Regulatory Agency (MHRA) has launched reports and an AI regulatory sandbox to inform future rules on the safe use of AI in healthcare.
The falsehood linking MMR vaccines to autism – a central pillar of the anti-vaccine movement – gained prominence after a study published in The Lancet in the 1990s was later fully retracted and refuted by multiple subsequent studies. The movement has gained additional influence since the Covid-19 pandemic and the appointment of Robert F. Kennedy Jr. as US health secretary. The KFF poll’s findings show that those who regularly consult AI for health advice are more likely to encounter and believe such debunked claims, a pattern that persists regardless of age, education or political affiliation. The poll did not determine whether the chatbots themselves are generating misinformation or simply reflecting false claims users already hold, but separate audits have found significant variation in chatbot performance – with Grok producing the highest proportion of problematic health responses.