Pastor says ChatGPT’s medical advice nearly killed him—now he’s suing OpenAI

As generative AI becomes a routine part of patients’ lives, physicians are encountering a new kind of “Dr. Google” consultation—one that doesn’t just retrieve information but offers personalized recommendations with remarkable confidence. 

A new lawsuit filed by a Florida pastor against OpenAI illustrates why that distinction matters.^[] The complaint alleges that the patient relied on ChatGPT for ongoing medical guidance, delayed seeking care despite worsening symptoms, and ultimately suffered a life-threatening pulmonary embolism after following ChatGPT’s advice to remain largely immobile.

OpenAI disputes the broader allegations and maintains that ChatGPT is not a substitute for medical care. The case is likely to become another closely watched test of where legal responsibility begins and ends when patients rely on AI-generated medical advice.

For physicians, however, the legal outcome may be less important than the clinical lesson: Many patients are already asking AI, rather than their doctor, what to do next.

What happened?

He later developed a pulmonary embolism that nearly proved fatal. The lawsuit accuses OpenAI of negligence and the unauthorized practice of medicine and seeks both damages and stronger safeguards for medical conversations.^[]

“Scott Winters nearly died because ChatGPT acted like a medical authority while having none of the responsibility. If ChatGPT were a physician dispensing medical advice, it would be guilty of medical malpractice,” said attorney Matthew P. Bergman in a statement.^[]

OpenAI spokesperson Drew Pusateri said that “ChatGPT is not a doctor and should never be used as a substitute for medical care, diagnosis, or treatment.”^[]

What doctors can learn

As LLM use becomes more common, physicians should expect AI use to become part of every history.

“With millions of people consulting LLMs for medical advice regularly, healthcare practitioners now need to know what to expect from patients with LLM-based opinions about their care,” wrote the authors of a research article in Nature Medicine.^[]

Just as clinicians now routinely ask about supplements or internet research, it may become worthwhile to ask: “Have you asked an AI chatbot about this?”

Simply telling patients “don’t use ChatGPT” is unlikely to succeed.

But emphasize that it should not diagnose urgent symptoms, determine whether emergency care is necessary, or replace clinical evaluation.

It's also important to identify high-risk AI behaviors.

These conversations may reveal delayed presentations that otherwise seem difficult to explain.

Finally, reinforce "red flag" symptoms. Patients often understand that chest pain deserves attention, but they may be less aware that leg swelling, syncope, severe dyspnea, neurological deficits, or persist abdominal pain also warrant prompt evaluation regardless of what an LLM suggests.

Explicitly discussing these “AI-proof” red flags may become an increasingly important part of patient education.

The communication challenge

Patients frequently interpret fluent, detailed explanations as evidence of accuracy—even when the information is incomplete or wrong. In the 2026 study, researchers found that AI-generated advice can increase users’ confidence while making them less likely to acknowledge uncertainty, particularly when the AI provides incorrect answers.^[]

That makes physician communication even more important. Explaining why a diagnosis remains uncertain, why additional testing is necessary, and why emergency symptoms require evaluation may help patients distinguish evidence-based clinical reasoning from AI-generated certainty.