OpenAI’s announcement of ChatGPT Health landed with predictable shock across health care. A consumer platform with hundreds of millions of users can now securely ingest medical records, wearable data, and personal health history to deliver context-aware guidance in real time. For many health executives, the instinctive response was alarm, but that reaction misses the point.
This announcement was not a disruption; it was a revelation. When OpenAI reports that 230 million people already ask health or wellness questions on ChatGPT every week, it is not declaring a new ambition. It is acknowledging a reality that health systems have spent years ignoring. Health care did not lose consumers to artificial intelligence overnight. It lost them gradually, through friction, delays, fragmented experiences, and a persistent inability to meet people where they are.
The failure of episodic care
The most uncomfortable question raised by ChatGPT Health is not whether consumer AI should exist in health care. It is this: If a consumer application can synthesize my health data, remember my history, and respond instantly with personalized guidance, why cannot my health system do the same?
That question should be unsettling because it exposes a long-standing failure of imagination. For decades, health care organizations have optimized around billing rules, compliance requirements, and episodic encounters. We digitized records but not relationships. We introduced portals without continuity. We measured productivity in visits rather than longitudinal impact. The result is a system that knows a great deal about patients but behaves as if it remembers very little. Consumer AI did not lower expectations. It revealed how low they already were.
Redefining primary care as a continuous service
The debate now unfolding tends to frame this moment as a contest between machines and clinicians. That framing is wrong and dangerous. Artificial intelligence does not deliver care; care teams do. AI changes how care is delivered, and the distinction matters.
Consumer AI operates in the domain of advice. Primary care, at its best, operates as an ongoing service. It is not a transaction, and it is not a visit. It is a continuously available capability that adapts as patients’ lives, risks, and needs change over time. This is where the current system breaks down.
Primary care today is still largely organized around episodic access and individual memory. It asks clinicians to reconstruct context at every visit and patients to reintroduce themselves to the system again and again. That model cannot scale, and it cannot compete with tools that feel persistent, responsive, and personal.
AI as service infrastructure
Artificial intelligence, properly applied, is not a chatbot layered onto an outdated workflow. It is service infrastructure. It enables continuous monitoring rather than intermittent check-ins. It synthesizes signals across encounters, settings, and time. It prompts proactive outreach instead of reactive documentation. Critically, it allows primary care teams to operate as a coordinated service rather than a series of disconnected interactions.
Some health systems are already beginning to redesign primary care around continuity rather than encounters. They are using AI to embed intelligence directly into clinical operations, so the system itself maintains patient-specific context over time. The goal is not to replace clinicians or to mimic consumer AI, but to ensure that the care experience itself becomes as continuous and context-aware as patients now expect.
Responding to the structural challenge
When primary care functions as a service, the comparison to consumer platforms becomes less threatening. Patients are not choosing between a chatbot and a doctor. They are choosing between fragmented care and care that feels present in their lives, even when they are not actively seeking it.
ChatGPT Health puts pressure on health systems not because it is irresponsible, but because it is responsive. It highlights the gap between what technology can deliver and what health care organizations have chosen to prioritize. The real risk is not that consumer AI exists; the real risk is treating it as the problem.
Issuing restrictive policies, warning clinicians, or dismissing patient enthusiasm misses the point. Those responses defend the status quo rather than addressing why the status quo no longer meets expectations. The harder response is structural. It requires redesigning primary care around continuity, anticipation, and service delivery. It requires embedding AI deeply into care teams and workflows so that responsiveness is not outsourced to consumer platforms but built into the care model itself.
The systems that succeed in this next era will not be the ones with the flashiest technology announcements. They will be the ones that quietly rebuild primary care as a service patients can rely on and more easily access. ChatGPT Health did not change health care overnight. It simply made clear how much change has already occurred without us.
David Carmouche is a physician executive.
