How Patients Are Using ChatGPT to Navigate Their Health
For years, patients have turned to search engines late at night, typing symptoms into a browser and hoping for clarity. Now, many are having an actual conversation instead — with artificial intelligence.
Across social media and patient communities, a new kind of anecdote is surfacing with increasing frequency: stories of people who say ChatGPT helped them recognize a serious medical issue and seek care sooner than they otherwise would have. Some describe it as reassuring. Others call it lifesaving.
One such story comes from New York-based consultant Bethany Crystal. After waking from a nap to find red spots spreading across her legs, she turned to ChatGPT for guidance. According to Crystal, the AI urged her to seek immediate medical evaluation, warning of a possible bleeding risk.
What followed was a frightening three-day stretch that ended with a diagnosis of immune thrombocytopenic purpura — a rare autoimmune disorder that can cause dangerously low platelet levels. Crystal has said she might not have gone to the emergency room as quickly without that prompt.
Her experience sits at the center of a growing conversation: What role should AI play in modern health care — and where are the limits?
Why Patients Are Turning to AI
Hundreds of millions of people now consult ChatGPT weekly for wellness-related questions, according to OpenAI. In early January, the company announced the launch of ChatGPT Health, a platform designed to allow users to share medical information with added security.
It joins a growing ecosystem of AI tools promising to help patients understand symptoms, interpret test results, and prepare for doctor visits.
Advocates say one advantage is simple but powerful: time.
Doctors often operate under intense time pressure, relying on experience and probability to guide diagnosis. There’s a saying in medicine — “If you hear hoofbeats, think horses, not zebras” — a reminder that common explanations are usually correct.
AI doesn’t work under the same constraints.
Unlike clinicians with packed schedules, large language models can ask follow-up questions indefinitely, exploring edge cases and rare conditions that might not immediately rise to the top of a differential diagnosis. For some patients — especially those with uncommon illnesses — that exhaustive questioning can feel like being heard for the first time.
Cancer survivor and patient advocate Dave deBronkart says he frequently hears from people who believe AI helped flag symptoms that didn’t fit a standard pattern. “Doctors are really good at horses,” he says. “They just don’t know all the special stuff.”
Living With AI as a Care Companion
For others, AI isn’t about diagnosis — it’s about daily management.
Burt Rosen, a 60-year-old marketing professional in Oregon, has been diagnosed with two rare cancers. He uses AI tools to track symptoms, translate lab results into plain language, and explore treatment options ahead of appointments.
In one instance, after experiencing migraines and nausea, Rosen described his symptoms to an AI system. It asked how he slept and suggested adjusting his pillow height. The headaches stopped.
Rosen says AI has fundamentally changed how he interacts with his oncologist. “I’m no longer going in to have him explain my scans,” he says. “My appointments are about action planning.”
For many patients, that shift — from passive recipient to informed participant — is the appeal.
The Risks No One Can Ignore
Still, even proponents stress that AI is not a doctor — and treating it like one can be dangerous.
As a consumer product, AI health platforms are not regulated like medical systems. Privacy protections differ, oversight is limited, and errors can have serious consequences.
Physicians have already seen examples of AI giving advice that, while plausible on the surface, could delay effective treatment. Dr. Robert Wachter, chair of medicine at UCSF, recalls a case where an AI system suggested ivermectin as a treatment for testicular cancer.
The drug itself might not cause harm, Wachter says — but relying on it instead of proven therapies absolutely could.
Other documented cases include users ingesting inappropriate substances or misinterpreting guidance without clinical supervision. Mental health remains a particularly sensitive area, with ongoing lawsuits alleging psychological harm tied to AI interactions.
“The capacity for badness here is pretty high,” Wachter says.
A Tool That Doctors Are Already Using
Despite those risks, AI is rapidly embedding itself into clinical practice — often behind the scenes.
Wachter and many of his colleagues now use AI-powered documentation tools that transcribe patient visits in real time, freeing doctors to maintain eye contact instead of typing notes. Others rely on AI systems that instantly surface medical literature or help evaluate complex cases.
Studies published in major medical journals suggest that large language models can perform competitively with human clinicians in simulated diagnostic reasoning — sometimes falling just short, sometimes matching expert performance.
“The performance was remarkable,” Wachter says.
What the Future May Look Like
For patients like Rosen, AI is already part of the care team. And while he acknowledges that not everyone is equally comfortable with technology, he believes the learning curve will fade quickly.
Two generations from now, he says, this may all feel routine.
Wachter agrees — cautiously. He worries about AI’s impact on politics, employment, and misinformation. But in health care, he sees a system under enormous strain.
“It’s falling apart,” he says. “And it can’t meet people’s needs without this kind of help.”
For now, the message from both doctors and patients is clear: AI can inform, support, and empower — but it cannot replace clinical judgment. Used carefully, it may help bridge gaps in a system stretched thin. Used carelessly, it may create new ones.
