It is three in the morning somewhere in America. A phone screen glows in a dark bedroom. Someone types a question they would never ask out loud at this hour: is it safe to take these two pills together?

The USA Trusts Chatbot

Photo by Pavel Danilyuk: https://www.pexels.com/photo/woman-in-red-long-sleeves-shirt-thinking-while-looking-afar-8637923/ 

The answer arrives in seconds. Calm, fluent, reassuring. They read it, half believe it, and then lie awake wondering whether they should have.

That small scene, repeated millions of times, captures something new about modern life. We have handed some of our most private questions to machines we are not entirely sure we trust.

The Confessional in Your Pocket

Health has always been one of the most intimate subjects there is. We hesitate before telling a doctor about a symptom. We rehearse the question in the waiting room.

A chatbot removes all of that. No appointment, no receptionist, no embarrassment. Just a text box that never judges.

It is no surprise that people have taken to it. A KFF poll found that 32 percent of US adults had used AI chatbots for health information or advice in the previous year, roughly the same share who turned to social media for health research.

The shift is not only happening on the patient’s side of the consulting room. According to the American Medical Association, more than 80 percent of physicians now use AI professionally, up from 38 percent in 2023.

So the machine is in the room from both directions. The patient consults it at night. The doctor consults it by day.

What the AI in Healthcare Statistics Actually Say

The most revealing numbers are not about adoption. They are about ambivalence.

SingleCare’s recent data roundup pulls together a wide range of AI in healthcare statistics, and its own survey found that 46 percent of US adults have asked an AI tool a question about their medication.

The contradiction sits just below that headline. Some 78 percent felt the information they received was at least somewhat accurate. Yet 83 percent said they were concerned AI could give them incorrect medication guidance.

Read those two figures side by side and they seem impossible. How can most people find the answers accurate while almost everyone fears they might be wrong?

The same survey adds one more detail that makes it matter. Nearly half of those users, 49 percent, said they had adjusted how they take a prescription based on what an AI tool told them.

This is not idle curiosity. People are acting on answers they do not fully believe.

The Transparency Problem

If patients feel uneasy about the chatbots they choose, they feel even more uneasy about the ones they do not see.

A Pew survey found that 72 percent of US adults think it is very or extremely important for healthcare providers to tell patients when they use AI. Over half said they have little or no say in whether it is used in their care, and only around one in five were comfortable with the say they currently have.

There is a quiet irony in that. Americans are comfortable using AI themselves, at their own pace, in their own bedrooms. What unsettles them is AI used on them, without being asked.

It is the difference between choosing a tool and discovering one has been chosen for you.

Confidence Without Accuracy

Part of the unease is justified, and the research explains why.

A 2025 evaluation published in the American Journal of Health-System Pharmacy found that AI responses to drug information questions were completely accurate only 19 percent of the time.

That does not mean the rest were dangerous. Many answers were partly right, or right in general but wrong in detail. But in medicine, detail is often the whole point.

The deeper problem is style. A chatbot does not hesitate, stammer or say “let me check”. It delivers every answer with the same polished confidence, whether it is correct or not. Humans are wired to trust fluency, and machines have become extraordinarily fluent.

So the doubt people report is not paranoia. It is an instinct that the tone and the truth have come apart.

Why We Keep Asking Anyway

If so many people doubt the answers, why do they keep asking?

Convenience explains some of it. Healthcare can be slow, expensive and difficult to navigate, particularly in the US, where cost shapes almost every decision. A free answer at midnight is more accessible than an appointment next week.

Privacy explains more. Some questions feel easier to ask a screen than a person.

But there is something cultural going on too. We have spent two decades learning to search for everything, from recipes to relationships. Asking a machine about our bodies is simply the next step in a habit already formed.

The chatbot is not replacing the doctor so much as filling the silence before the doctor, and sometimes the silence after.

A View From This Side of the Atlantic

It would be easy to treat this as an American story. It is not.

Anyone in Britain who has waited on hold to secure a GP appointment will recognise the appeal of an instant answer. The pressures are different, but the temptation is the same.

What the American figures offer is a preview. They show what happens when a population adopts a tool faster than it learns to trust it. They show people keeping one foot in the old system and one in the new, uncertain which is safer.

The UK is unlikely to be far behind.

The Uneasy Middle

For now, we live in an uneasy middle ground. We trust the chatbot enough to ask it, and sometimes enough to act on it. We doubt it enough to lie awake afterwards.

Perhaps that is the healthiest place to be. Blind trust in any source, human or machine, has never served patients well. A little doubt keeps people checking, asking follow-up questions and, ideally, bringing those questions to a pharmacist or doctor.

The chatbot is here to stay. The real question is whether our instincts can keep pace with its confidence.

written by Al Woods