Most people can identify a specific type of late-night anxiety: the clinic has been closed for hours, something hurts, or something feels strange. That moment led people to Google ten years ago. It is now sending them to a chatbot more and more. Between the emergence of consumer AI and the enduring fact that visiting a doctor in America is costly, time-consuming, and occasionally just inconvenient, the change has occurred covertly and without notice. Finally, the numbers have caught up to the habit.
According to a West Health-Gallup Center on Healthcare in America survey published in April 2026, over 66 million Americans, or one in four adults, have used an AI tool or chatbot to get advice or information about their physical or mental health. The survey, which was conducted in late 2025 among a nationally representative panel of more than 5,500 adults, provides one of the most comprehensive depictions of how common people are incorporating artificial intelligence into decisions that were previously thought to be solely within the purview of physicians. Since then, the number may have increased even more. These things often move more quickly than the surveys that monitor them.
The dynamic is quite comforting for the majority of users. More than half claim to use AI to enhance their care, whether it’s to look up symptoms prior to a visit or attempt to interpret a diagnosis afterward. Seventy-one percent said they wanted answers right away, and another seventy-one percent said they simply wanted more details. That has an almost unremarkable quality. People fill in the gaps left by the system by using the tools at their disposal. According to 46% of AI health users, the experience increased their confidence when speaking with a provider. That result is not insignificant. Patients typically receive better care when they arrive at appointments with better questions.
However, there is a figure in the survey that is more difficult to ignore. Approximately 14 million adults, or 14% of recent AI health users, claim that the AI’s advice prevented them from seeing a medical professional they otherwise would have. That isn’t care supplementation. It is being replaced by that. Cost is a contributing factor; 32% of users with annual incomes under $24,000 used AI because they were unable to pay for medical care. In those situations, the chatbot is neither a luxury nor a curiosity. It is a stand-in for a pricing system.
Approximately 11% of current users claim that the AI provided them with advice they thought was risky. It is important to read that figure carefully because it is based on personal opinion rather than a clinical analysis of the recommendations. Something that was accurate could be reported as dangerous by a user. Alternatively, they may fail to identify advice that was actually detrimental. It is highly likely that the number itself is inaccurate. However, the direction it takes is fairly obvious, and it’s difficult to avoid feeling uneasy about what it suggests: millions of people using a tool that may be misguiding them on important issues in certain situations.
On the other hand, trust in technology is disjointed in a way that is distinctly American. About one-third of people who have used AI for health-related purposes in the last month say they trust it, one-third neither trust nor distrust it, and one-third distrust it. Just 4% of people firmly believe it to be accurate. This indicates that a significant portion of people are acting based on information they are not totally certain is accurate; this is not because they are careless, but rather because the alternatives seem more difficult or distant.
As you watch this unfold, you get the impression that medicine is facing something it has never had to deal with quite like this. Chatbots are capable of summarizing. They can provide comfort. They can assist someone who already knows what questions to ask when they arrive for an appointment. They are unable to examine you or take accountability for your future actions. The tool is truly helpful until it isn’t, and it is still a human problem to find that line in real time with actual symptoms.
