A new study has revealed the concerning side effect of relying on artificial intelligence (AI) chatbots for health advice: a heightened likelihood of believing anti-vaccine myths. A recent poll in the US found that those who regularly consult AI tools for health information are more likely to endorse false claims, such as the measles, mumps, and rubella (MMR) vaccine causing autism in children.
The survey, conducted by health research firm KFF in May, involved a representative sample of 2,480 US adults. It discovered that individuals who use AI tools for health information at least once a week are more likely to hold false beliefs. Of these frequent users, 35% believed it was "definitely or probably true" that MMR vaccines cause autism – a higher percentage than those who do not use AI for health advice (20%) and even occasional AI users (29%).
The poll also highlighted other misconceptions prevalent among those who frequently use AI chatbots. Around 29% of this group believed that mRNA vaccines can alter human DNA, a claim with no scientific basis. This compares to 20% of non-AI users holding the same belief. Furthermore, 22% of frequent AI users thought the measles vaccine posed more danger than the measles virus itself, compared to 15% of those who do not use AI for health information.
The trend of Americans turning to AI chatbots for medical guidance is clear: a March survey by KFF found that about one-third of US adults seek health advice from AI. AI developers acknowledge this shift too; OpenAI noted in a January blog post that "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," when announcing a specialised ChatGPT Health tool.
This emerging pattern follows a historical trend seen with search engines, where health-related queries constitute a significant portion of overall searches. The KFF poll didn't specify which particular AI models respondents used, but it's understood that different chatbots can produce varying levels of misinformation, influenced by their training data and company policies on handling sensitive or divisive questions.