Gemini for Healthcare Enters US Preview

Google Cloud on Nov. 11 launched Gemini for healthcare in preview in the United States, an agent package that reads hospital records, checks insurance policies and assembles prior-authorization paperwork for hospitals and health insurers. Google Cloud CEO Thomas Kurian announced the product in a company blog post, arguing that "a general AI model is not enough" for an industry whose data sits in disconnected systems.

Key takeaways

• Gemini for healthcare is in U.S. preview and is built on the general-purpose Gemini agent Google introduced in October at its Gemini at Work event.
• Google lists more than two dozen connectors, including read-only links to Epic and Oracle Health records, Optum eligibility tools and Medicare coverage rules.
• Google says patient data, prompts and outputs stay in the customer's environment and are not used to train its foundation models.

What Gemini for healthcare does

The product is built from four parts: prewritten "skills" for clinical and administrative tasks, connectors to existing systems, a partner ecosystem and a governance layer that lets organizations choose among models. Google says it is built to support HIPAA compliance, with audit logs and citations so reviewers can see where each answer came from.

For hospitals, the skills build a timeline of a patient's history from scattered records, prepare charts before visits and pull chart evidence into prior-authorization packets. If a claim is denied and the record supports it, the agent can draft an appeal. For insurers, it checks incoming requests against medical policy, organizes findings for clinical reviewers and estimates members' out-of-pocket costs from their plan documents.

In the announcement, Google cites a 2024 report on administrative burden estimating that clinicians spend nearly 28 hours a week manually reconciling information. AI Newsway could not independently verify that figure.

Epic, Oracle Health and Optum connectors

The connector list spans records, claims and coding. It includes Epic and Oracle Health EHRs, an InterSystems FHIR server, TriZetto and GoodRx, ICD-10 and RxNorm lookups, the NPPES provider registry and a CMS connector for national and local Medicare coverage determinations. Two Optum tools, for coordination of benefits and network eligibility, are still in limited release and testing.

Google's Epic connector documentation shows the guardrails. The connector is read-only, calls Epic's FHIR R4 API as the signed-in user, and does not index or copy Epic data. Each search covers one patient and returns at most 50 records per page. It cannot read PDFs, truncates notes longer than about 20,000 characters, and does not read imaging, care plans or family history.

Who is involved

Google named 16 systems integrators, including Accenture, Deloitte, KPMG and PwC. Executives from HCA Healthcare, Highmark Health, Northwell Health, Blue Cross Blue Shield of Massachusetts and the American Society of Clinical Oncology were quoted in the announcement. Several described their work as exploratory.

"The real promise of AI in healthcare isn't simply doing today's work faster. It's creating a better way for the health system to work," said Sandeep Dadlani, CEO of Optum Insight.

Why it matters

Prior authorization and claims review sit between hospitals and insurers, and Google is pitching one agent that serves both sides of that exchange. Gemini for healthcare adds to Google's industry packages for legal, financial services and government. It also comes as Google pushes its newest models, such as the Gemini 4 family, into enterprise use.

What's next

Google has not published pricing or a general-availability date. Key questions are how accurate the agent's coverage findings are in practice and how insurers disclose AI use in reviews. Google says all outputs are meant for staff review.

FAQ

Can Gemini for healthcare change patient records? No, not through the Epic and Oracle Health connectors. Google describes both as read-only with no write-back to the chart.

Is it available outside the U.S.? Not yet. Google says the preview is available in the U.S.

Does Google train its models on patient data? Google says clinical data, prompts and outputs are never used to train or fine-tune its foundation models.

References

• Introducing Gemini for healthcare — Google Cloud Blog (Nov. 11, 2026)
• Epic EHR connector overview | Gemini Enterprise — Google Cloud Documentation