ChatGPT Health Plugs Into Epic's Patient Records

OpenAI just gave clinicians a read-only bridge between ChatGPT and the electronic health record system that holds data on 325 million patients.

Welcome Back,

Every clinician knows the feeling of opening a patient's chart minutes before an appointment and scrolling through years of notes, labs, and referrals trying to reconstruct a story that should already be obvious. OpenAI thinks an AI assistant sitting right inside that chart can fix a lot of that wasted time.

On September 1, 2026, OpenAI announced it is integrating ChatGPT Health with Epic's electronic health record system, which holds data for over 325 million patients, letting clinicians import patient data and ask AI questions about it directly, according to TechCrunch.

Today we look at what the Epic integration actually lets clinicians do, how it fits into the seven-month journey ChatGPT Health has taken since launch, what guardrails OpenAI has built in, and why a read-only connection to hospital records is still a meaningfully big deal.

🩺 What the Epic Integration Actually Does

With the integration live, clinicians can access information like appointment notes, laboratory results, medications, and specialist documentation, and have ChatGPT summarize it in seconds instead of scrolling through a full chart, according to TechCrunch.

It supports two distinct workflows. Clinicians can pull authorized patient data directly into ChatGPT to review medical histories, spot changes over time, and prepare for upcoming visits, or use ChatGPT within supported Epic workflows without ever leaving the patient chart, per Digital Health News.

In practice, that means a doctor can ask ChatGPT to build a clinical timeline, flag what's changed since the last visit, or generate a quick pre-visit review, all pulled from a patient's real record rather than a generic knowledge base, according to Yahoo Tech.

Becker's Hospital Review reports that UCSF Health is serving as the pilot partner for the rollout, giving OpenAI a real hospital environment to test the integration before wider release, according to Becker's Hospital Review.

The integration is designed to fit into a clinician's existing workflow, whether that's inside ChatGPT or embedded directly in the Epic chart.

🕒 How ChatGPT Health Got Here

This launch is the latest step in a fast build-out. OpenAI first introduced ChatGPT Health on January 7, 2026, as a dedicated tab for answering health questions and letting users upload medical records or connect apps like Apple Health and MyFitnessPal, according to Reuters.

That early version relied on a partnership with b.well, a data connectivity company that aggregates records from roughly 2.2 million U.S. health care providers, and was initially limited to a small group of early users outside the EU, Switzerland, and the UK, per Medical Economics.

By July 23, 2026, OpenAI expanded Health to all logged-in U.S. users 18 and older across Free, Go, Plus, and Pro plans on web and iOS, with support for connecting records from systems like Epic and Oracle Health, and platforms like One Medical and Function Health, according to TechCrunch.

Alongside the Epic integration, OpenAI also added a Healthcare Public Data plugin connecting ChatGPT to nine official public sources, including ClinicalTrials.gov, CMS Coverage, RxNorm, DailyMed, and PubMed, giving clinicians a way to cross-reference patient data against current medical literature and guidelines, according to Becker's Hospital Review.

🔒 Why "Read-Only" Is the Key Detail

Handing an AI model access to hospital records raises an immediate question: can it change anything? OpenAI's answer is a firm boundary. The company specified that the integration allows only read-only access to health records, and the AI doesn't write anything back, according to TechCrunch.

OpenAI's own help documentation is specific about what that means in practice: the connection cannot update medical records, place orders, message patients, or override existing patient-chart permissions, and clinicians remain responsible for reviewing the underlying record and making care decisions, per the OpenAI Help Center.

Access is also tightly gated. The Epic plugin is only available for approved ChatGPT for Healthcare and HIPAA-enabled Enterprise workspaces, not individual ChatGPT for Clinicians accounts, and each clinician has to connect their own Epic sign-in, meaning they can only see patient records their existing Epic permissions already allow, according to the same documentation.

UCSF Health is serving as OpenAI's pilot partner as the Epic integration rolls out to approved healthcare organizations.

⚙️ Getting It Set Up Isn't Simple

This isn't a toggle a clinician flips on their own. The organization first needs to be approved for the Epic integration under an applicable Business Associate Agreement covering protected health information, and its Epic administrator has to register or approve an OAuth application for the correct Epic environment, according to the OpenAI Help Center.

From there, a workspace admin decides whether the plugin is simply available for eligible users to install themselves, or installed automatically for specific roles. Either way, every individual clinician still has to sign in with their own Epic credentials before they can pull up a single patient record, per the same documentation.

That layered approval process, organization sign-off, admin configuration, individual sign-in, mirrors how hospitals typically roll out any new EHR-connected tool, treating an AI assistant with the same access controls as any other clinical software, rather than a shortcut around them.

⚖️ Consumer Health AI vs. Clinical Health AI

It's worth separating this from the consumer-facing ChatGPT Health that launched in January. That version lets everyday users upload their own records and ask questions about their own health, with OpenAI explicit that it's designed to "support, not replace, medical care" and isn't intended for diagnosis or treatment, according to the BBC.

That said, privacy advocates flagged concerns even with the consumer version at launch, warning that letting an AI company aggregate sensitive medical records, even under strict data controls, is a meaningful shift in how health information moves through the system, according to the BBC.

The bigger question is whether AI-assisted chart review actually gives doctors more time with patients, or just shifts where the time goes.

🧠 AI Spotlight Analysis

The real significance here isn't the chatbot, it's the plumbing. Epic sits underneath a huge share of U.S. hospital systems, so connecting to it directly, rather than asking every hospital to build its own custom integration, is what makes this scale quickly instead of staying a boutique pilot.

OpenAI's own framing of the strategy is telling. One of its team members described the goal as bringing ChatGPT "closer to the systems, information" clinicians already rely on, rather than asking doctors to adopt an entirely new tool, according to a LinkedIn post from an OpenAI team member.

The read-only design is a smart first move for trust, but it also means the immediate value is narrow, faster chart review and summarization, not diagnosis or decision-making. Whether OpenAI eventually pushes further into clinical recommendations will be the real test of how far health systems are willing to let this go.

🔗 Sources and Further Reading

Thanks for reading AI Spotlight.

Our mission is simple: deliver clear, trustworthy, and actionable AI insights that help professionals stay ahead without the hype.