Ready or not, ‘ChatGPT Health’ has entered the exam room
Industry Buzz
It’s alarming because a lot of that information is misinformation, so that really undermines the trust between patients and their physicians… AI information is different because it’s hard to know who’s generated it, it’s hard to ensure where the validity and the science has come from.
—Margot Burnell, MD, CMA President, Montreal City News
The launch of ChatGPT Health has set off a ripple of concern and curiosity within the healthcare industry.[] This new tool allows patients to integrate personal medical records and wellness data into ChatGPT’s AI, enabling more tailored health-related advice.
But for healthcare providers, the real question is whether this innovation is a revolutionary patient resource or a potential clinical nightmare that further fuels misinformation, misinterpretation, and patient confusion.
Ahead, we unpack real-world risks of use, take an optimistic look at potential upsides, and break down what it could all mean for the clinic.
A simple solution? Maybe not
Health-related questions have long been among the most common uses of ChatGPT, with over 230 million people globally seeking wellness advice each week.[] With ChatGPT Health, OpenAI aims to offer users more than generic responses. The tool pulls from medical records and wellness apps, offering tailored insights that go beyond surface-level suggestions.
While the integration of electronic medical records and wellness data might seem like an improvement, clinicians may question whether better data truly leads to better understanding. With patients receiving more context-driven answers, the challenge becomes: Will they trust AI-generated advice, or will it create new complexities in patient care?
OpenAI has been clear: ChatGPT Health is not designed to diagnose or treat medical conditions. Rather, it aims to help users understand lab results, prepare for doctor visits, spot trends in health data, and navigate lifestyle choices like diet, exercise, and insurance. The product is intended as a care-adjacent tool (not a decision-maker), although patients may still rely on it for more than it was designed to offer.
The integration of personal medical data—via partnerships with platforms like b.well, Apple Health, MyFitnessPal, and others—has the potential to personalize interactions. But clinicians must carefully consider whether this level of data-sharing will lead to misunderstandings or reliance on AI instead of professional expertise.
OpenAI emphasizes that ChatGPT Health operates in a separate, encrypted space to ensure that health data is stored securely and is not used to train other models. Data is isolated from standard conversations, and users can review or delete their health-related memories at any time. These measures offer some reassurance, but healthcare providers will need to assess whether they can trust AI interpretations—or whether they should actively discourage patients from using the tool.
Related: Docs say this AI tool has finally allowed them to give patients their full attentionRisks in real-world use
Misinterpretation, not misinformation
Even when AI provides technically accurate responses, patients may misinterpret them: overestimating certainty, missing nuance, or drawing conclusions that feel authoritative because the system references their own lab values or chart data. Unlike a generic Google search, chart-informed AI can appear clinically validated, making it harder to recalibrate expectations in the exam room.
“It’s alarming because a lot of that information is misinformation, disinformation, and false information. And so that really undermines the trust between patients and their physicians,” Margot Burnell, MD, president of the Canadian Medical Association, told Montreal City News. “AI information is different because it’s hard to know who’s generated it, it’s hard to ensure where the validity and the science has come from,” she said.[]
Escalation confusion
Although the platform encourages follow-up with healthcare providers, patients may delay care if AI feedback seems reassuring. Conversely, ambiguous or alarming outputs could drive anxiety-fueled visits, additional portal messages, or requests for low-value testing, shifting how and when patients enter the system.
Workflow friction
In practice, this may mean longer visits spent unpacking AI-generated interpretations, correcting inaccuracies, or explaining why a suggested test or diagnosis is not indicated. What was once “I Googled this” could become “My AI analyzed my chart,” adding a new layer of perceived authority to navigate.
Liability issues
If a patient delays treatment—or pursues unnecessary interventions—based on AI guidance, where does responsibility fall? As AI becomes embedded in patient decision-making, documentation standards, informed consent conversations, and medicolegal exposure may need to evolve alongside it.
A potential upside?
Despite these challenges, ChatGPT Health offers tangible benefits when used thoughtfully.[]
Better-prepared patients: With clearer explanations of lab results and medical notes, patients can engage more effectively in their care.
Improved health literacy: AI-generated, plain-language explanations can help patients better understand complex medical terms and concepts, closing long-standing gaps in health literacy.
Pattern recognition: The ability to track trends in health data can uncover insights that might otherwise be overlooked by both patients and providers.
Reduced clinician burnout: By providing patients with baseline knowledge, ChatGPT Health could reduce time spent on repetitive educational tasks, freeing clinicians to focus on higher-priority care decisions.
Is ChatGPT Health a breakthrough, or a potential clinical nightmare? It’s likely both. The tool represents a shift from static health information to dynamic, personalized insights—but it also introduces the risk of amplifying misunderstandings or misdirecting patient care.
Whether clinicians embrace or resist it, ChatGPT Health is likely to enter the exam room—forcing providers to decide how to address AI-generated interpretations before patients act on them.