ChatGPT Can Now Read Epic Charts - But It Can't Write Back

OpenAI has connected Epic’s electronic health record system to ChatGPT, so an authorized clinician can pull a patient’s chart into a chat window and ask what changed since the last visit. The connection is read-only: ChatGPT can retrieve and summarize the record, but it does not write anything back into it. Announced on September 1, 2026, the Epic integration arrived alongside a second feature - a Healthcare Public Data plugin that wires ChatGPT into nine official datasets including PubMed, DailyMed, and CMS Coverage.

What OpenAI actually connected

Two capabilities shipped the same day, and they solve different problems. The first is the Epic integration: OpenAI said healthcare organizations can now connect their Epic environments to ChatGPT for Healthcare, its governed workspace for clinical, research, and administrative teams. Instead of hunting across appointment notes, laboratory results, medications, and specialist documentation, a clinician can ask “What has changed since this patient’s last visit?” - one of the sample prompts in OpenAI’s announcement. ChatGPT pulls the relevant pieces from the authorized record, summarizes the important developments, and points back to the supporting chart information.

The integration comes in two shapes. In the first, authorized patient information from a supported EHR is brought into ChatGPT. In the second, available only in what OpenAI calls supported deployments, ChatGPT is embedded directly into an EHR layout so a pre-visit review happens without leaving the patient chart. Fierce Healthcare reported it can pull together clinical notes, lab results, medications, and specialist documentation.

Read-only is the most important phrase in the announcement

OpenAI was explicit that the Epic connection allows read-only access and that the AI does not write anything back to the chart - a point TechCrunch, Fierce Healthcare, and PYMNTS each repeated. That sets the ceiling: this is a retrieval and summarization layer beside the record, not a system that changes it.

Scale is why the announcement drew attention. Epic’s system holds data for more than 325 million patients, according to TechCrunch, a figure PYMNTS also cited. Wiring a general-purpose assistant into a record system of that reach is not a departmental pilot.

TechCrunch noted that despite these deeper integrations, OpenAI has maintained that AI is not suitable for diagnosis or treatment. The use cases the company named - pre-visit review, clinical timelines, medication review, handoff summaries - are preparation work OpenAI says clinicians and staff can review.

Nine official datasets, addressable by record and version

The Healthcare Public Data plugin is the half that got the shorter treatment in the coverage, and arguably the more usable one. It bundles dedicated connectors to nine official public healthcare sources; OpenAI named five: ClinicalTrials.gov, CMS Coverage, RxNorm, DailyMed, and PubMed.

What separates it from ordinary web search is granularity: OpenAI said teams can work with specific records, fields, identifiers, and versions while focusing a task on those authoritative sources, making it easier to compare and verify precise information - trial eligibility criteria, medication identifiers, coverage policy versions, provider records - without searching each source separately.

OpenAI’s examples are concrete: a research team using ClinicalTrials.gov to find actively recruiting trials and compare eligibility criteria, a pharmacy team using DailyMed to confirm the latest label and warnings for a medication, and a population health team pulling research, active trials, and Medicare coverage into one source-backed view for a diabetes-prevention program.

Who can actually switch this on

This is where the news stops applying to most readers. According to Unite.AI, existing ChatGPT for Healthcare customers should ask their workspace administrator to enable both features, while ChatGPT Enterprise customers were directed to their OpenAI account team to confirm eligibility for a Regulated Workspace. Eligible U.S. ChatGPT for Clinicians users can install the public data plugin - but the EHR integration is not available for individual accounts.

So if you are comparing the consumer ChatGPT tiers, none of them include this. It is an organizational purchase with a contract in the loop.

Unite.AI listed AdventHealth, Baylor Scott & White Health, Boston Children’s Hospital, Cedars-Sinai, HCA Healthcare, Memorial Sloan Kettering Cancer Center, and UCSF as launch partners, with UCSF Health named as pilot partner for the EHR feature. “As a pilot partner, we’re exploring how the new EHR integration with ChatGPT for Healthcare can help clinical teams understand what has changed and what matters most across a complex patient record,” Suresh Gunasekaran, president and CEO of UCSF Health, said in the announcement, adding that it “has the potential to reduce time spent synthesizing data and give clinicians more time with patients.”

Governance is the pitch around it. ChatGPT for Healthcare pairs healthcare-specific capabilities with enterprise controls such as role-based access, single sign-on, and audit logs, and with an applicable business associate agreement, customers can use ChatGPT Work, Codex, apps, and plugins in the same workspace to support HIPAA-compliant workflows. Plugins for Microsoft SharePoint, Google Drive, Salesforce, and Slack widen that context while preserving existing permissions.

The safety numbers come from OpenAI

There is real evaluation work behind the launch, and OpenAI is upfront that it ran that work itself. The company said it partners with hundreds of physicians across 60 countries, 49 languages, and 26 medical specialties to define and improve health responses in ChatGPT, and that they have reviewed more than 700,000 model responses to date.

For connected EHR context specifically, physicians evaluated responses across 27 clinical use cases including pre-visit review, clinical timelines, medication review, and handoff summaries. Across 4,363 ratings, they rated 99.1% of responses safe. In a separate two-round evaluation of nuanced clinical questions based on large U.S. healthcare datasets, more than 93% of responses were rated “good” or better on accuracy for each of the five connected data sources tested.

Unite.AI attached the caveat that matters: “These figures are OpenAI’s own reported evaluation results.” None of the coverage describes an independent auditor. TechCrunch was blunter about the remainder, writing that even a few unsafe answers can lead to harmful results for humans. So run the exercise any buyer should run against a vendor benchmark - decide what you measure and who owns the measurement before the tool is switched on, not after.

What happens next

First, expect naming confusion. TechCrunch headlined the story as ChatGPT Health adding an Epic integration, while OpenAI’s post and Fierce Healthcare call the product ChatGPT for Healthcare, which Fierce dates to January and describes as running on GPT-5 models that went through doctor-led testing. PYMNTS traces a January launch of ChatGPT Health for consumers and an April launch of ChatGPT for Clinicians. The coverage does not use the three names consistently.

Second, the consumer pressure behind this is large. TechCrunch said OpenAI rolled out ChatGPT for health to all U.S. consumers last month, and that the company said at the time people ask ChatGPT 300 million health-related queries every week.

Third, the legal backdrop is live. TechCrunch reported that days before this rollout a Florida-based pastor sued the company alleging ChatGPT gave him a near-fatal recommendation, and that family members of a user sued in May over wrongful advice related to dosage. The coverage does not tie either case to the Epic workflow, but they will shape how cautiously hospital policies get written.

What to watch: whether other EHR vendors follow Epic, whether an accuracy evaluation appears from someone other than OpenAI, and how far the in-chart experience spreads beyond the supported deployments OpenAI has not enumerated. Regulated buyers keep converging on one governed front door for general-purpose models - the Pentagon’s secure portal for ChatGPT and Grok is the closest analogue outside healthcare.

FAQ

Can ChatGPT write notes back into Epic? No. OpenAI said the integration allows read-only access and does not write anything back to the chart, and TechCrunch, Fierce Healthcare, and PYMNTS all reported the same limit.

Can I use this with a personal ChatGPT subscription? No. Unite.AI reported the EHR integration is not available for individual accounts. Eligible U.S. ChatGPT for Clinicians users can install the public data plugin, but the EHR connection needs a workspace administrator.

Which public health datasets does the plugin cover? It connects nine official public healthcare sources. OpenAI named five: ClinicalTrials.gov, CMS Coverage, RxNorm, DailyMed, and PubMed.

Is ChatGPT now approved to diagnose or treat patients? No. TechCrunch noted that despite these integrations, OpenAI has maintained that AI is not suitable for diagnosis or treatment. The described use cases are preparation work a clinician reviews.