Nearly one in four Americans are now using AI chatbots to supplement their healthcare visits, according to a new Gallup poll. They are asking ChatGPT about symptoms. They are using Claude to interpret lab results. They are feeding their prescriptions into AI models and asking for second opinions.

Most of them never tell their doctor.

The Gallup data, published this week, reveals a healthcare system that is quietly being restructured from the bottom up. Not by policy. Not by technology companies. By patients who are too tired of waiting three weeks for a 12-minute appointment that costs $300.

The Numbers

24% of American adults say they have used AI to get health information or advice that supplements a visit to a healthcare provider. Among adults under 35, the number is significantly higher. The poll also found that AI use is highest among people with chronic conditions who interact with the healthcare system most frequently.

This is not hypothetical adoption. This is not a survey about interest or willingness. This is people actually doing it, right now, at scale.

The Trust Gap

The uncomfortable truth: these patients are not choosing AI because they trust it more than their doctor. They are choosing it because their doctor is not available. The American healthcare system has a 28-day average wait time for a new patient appointment. AI has a zero-second wait time. It is available at 3 AM when the anxiety hits. It does not charge a copay.

This creates a dangerous feedback loop. Patients get instant answers from AI. Those answers are sometimes wrong. But the patient has already made a decision based on them before ever seeing a doctor. When they do see the doctor, they either do not mention the AI consultation or they arrive with conclusions that are hard to undo.

States Are Already Fighting Over This

Maryland just passed legislation regulating AI use by health insurers. The state is specifically targeting insurers that use AI models to make coverage and denial decisions. Virginia, right next door, paused all AI healthcare legislation entirely. Missouri tried to pass AI liability and child safeguard bills but stalled because rural lawmakers warned it could cost the state $900 million in federal broadband funds.

The regulatory landscape is a mess. States cannot even agree on whether AI in healthcare is a problem worth solving, let alone how to solve it. Meanwhile, 24% of Americans have already decided for themselves.

The Real Risk

AI models are getting better at medical reasoning. Some studies show GPT-4 and Claude passing medical licensing exams at rates comparable to human physicians. But passing a test is not the same as treating a patient. AI does not know your family history. It cannot hear the thing you are not saying. It cannot read the body language that tells an experienced physician something is really wrong.

The biggest risk is not that AI gives wrong answers. It is that patients stop asking the right questions. When you have a confident, instantly available AI telling you what is probably happening with your health, the incentive to push through the broken appointment system drops. And sometimes "what is probably happening" is not what is actually happening.

A quarter of America is already running this experiment on themselves. Nobody asked permission. Nobody set up guardrails. The FDA has not weighed in. The AMA has not published guidelines. The patients just started doing it. The healthcare system will need to catch up, or it will find itself competing with a chatbot for relevance.

Source: Gallup poll published April 2026. State regulation data via WAMU, Columbia Missourian, and Daily Caller.