OpenAI has officially opened the doors of ChatGPT Health to every adult user in the United States, ending a six-month pilot phase and turning one of the most-used features of its chatbot into a nationwide product. The company confirmed on Thursday that the health experience is now rolling out to all logged-in US users aged 18 and above, across the Free, Go, Plus, and Pro plans, on both the web and iOS.
The expansion marks one of the boldest moves yet by an AI company into personal healthcare, arriving at a moment when Americans are already asking ChatGPT more than 300 million health-related questions every single week up sharply from the 230 million weekly queries OpenAI reported when the pilot began in January.
From Limited Pilot to Nationwide Rollout
ChatGPT Health was first unveiled in early January 2026 as an invite-style pilot, housed inside a dedicated hub within the chatbot. The idea was simple but ambitious: give users a private, secure space where ChatGPT could ground its answers in their actual health data rather than generic medical information pulled from the internet.
During the pilot, users could connect popular wellness services such as Apple Health, MyFitnessPal, and Function Health, allowing the assistant to see real data points like sleep patterns, activity levels, medications, and nutrition logs. The full rollout keeps all of those integrations and goes further, letting users pull in official medical records from major hospital systems, including Epic and Oracle Health, as well as care platforms like One Medical. Users whose providers are not yet supported can upload their records manually.
Once connected, ChatGPT can help people compare lab results over time, prepare questions for an upcoming doctor's appointment, understand confusing visit summaries, and explore how daily habits connect to their overall wellbeing.
Health Answers Now Work in Everyday Conversations
Perhaps the most significant change in this release is where health conversations can happen. During the pilot, users were required to start health-related chats inside the dedicated Health section. That restriction is now gone.
OpenAI said its testing revealed a telling pattern: roughly 70 percent of health-related questions were being asked outside the dedicated hub, woven naturally into everyday conversations. Someone planning a dinner menu might casually ask about a food allergy; someone booking a trip might ask about medication timing across time zones. Forcing those moments into a separate tab simply did not match how people actually use the assistant.
With the update, users who grant permission can now let ChatGPT draw on their connected health context in any conversation. A general chat about meal planning, for instance, can automatically account for a user's documented allergies or dietary restrictions with no tab-switching required.
Smarter Models Behind the Scenes
The rollout rides on the back of OpenAI's newest model family, GPT-5.6. The company says every model in the new lineup outperforms GPT-5.5 on HealthBench, its open-source benchmark built to evaluate how well large language models handle health queries. Even the smallest model in the release, GPT-5.6 Luna, beats its predecessor on the evaluation, while GPT-5.6 Sol delivers the strongest performance on complex health-related tasks for paying subscribers.
The medical grounding did not happen in a vacuum. OpenAI built the feature in close collaboration with a network of more than 260 practicing physicians across some 60 countries and dozens of specialties. Those doctors reviewed and scored model outputs over 600,000 times across 30 health domains, feedback the company says shaped the model to prioritize safety, clarity, and knowing when to push a user toward professional care. The same physician network helped design HealthBench itself, which grades responses against physician-written rubrics rather than exam-style multiple-choice questions.
Fidji Simo, who led OpenAI's applications division through the product's development, framed the mission around a broken system where clinicians have only minutes per patient. In her launch essay, she argued that under such conditions doctors struggle to "see the full picture of someone's health," and described the product as "another step toward turning ChatGPT into a personal super-assistant."
Privacy Built as the Foundation
Given the sensitivity of medical data, OpenAI has wrapped the feature in layered safeguards. The company states firmly that connected health records, Apple Health data, and the conversations that use them are never fed back into training its foundation models. Health conversations are encrypted both in transit and at rest, with additional access controls and audit logging, and all health files and chats live in a dedicated space kept separate from a user's regular ChatGPT history.
Access is also permission-based at a granular level. When ChatGPT wants to reference health records to tailor a response, it asks first and users can enable or disable that access at any time.
A Rollout Shadowed by a Lawsuit
The timing of the expansion has raised eyebrows. Just one day before the announcement, a Florida-based former pastor, Scott Winters, filed a lawsuit against OpenAI and CEO Sam Altman in San Francisco County Superior Court, alleging that ChatGPT discouraged him from seeking medical care in July 2025. According to the complaint, the chatbot advised him to stay home and limit movement despite worrying symptoms, advice he says preceded a massive pulmonary embolism that brought him to the brink of death.
OpenAI has pointed to its terms of service, which state its products are "not intended for use in the diagnosis or treatment of any health condition." Company spokesperson Drew Pusateri was blunt in a statement, saying "ChatGPT is not a doctor," and stressing it should never substitute for medical care. The company also told reporters it is actively working to make health and medicine-related answers safer, and with this rollout it is urging users to verify important information and base medical decisions on professional advice.
Independent research adds weight to that caution. Multiple studies, including work from Oxford's primary care researchers, have warned that AI chatbots frequently deliver problematic or unreliable health advice, a sobering counterpoint to the benchmark gains.
The Race for AI Healthcare Is On
None of these concerns has slowed the industry's charge into healthcare. Anthropic announced Claude for Healthcare in January, shortly after OpenAI's initial reveal, while Google has rolled out an AI-powered personal health and fitness coach for Fitbit users. OpenAI itself has expanded on multiple fronts, launching ChatGPT for Clinicians in April, a free documentation and research tool for verified physicians alongside an enterprise-facing OpenAI for Healthcare initiative.
For now, ChatGPT Health remains a US-only offering, and the rollout will reach all eligible accounts over the coming week. Whether it becomes the trusted health companion OpenAI envisions, or a cautionary tale about AI overreach in medicine, may depend less on benchmarks and more on how carefully millions of Americans and the company itself handle the line between information and medical advice.
