Ask Fitbit's AI coach how to improve your cholesterol today and it will not give you the generic answer it would have given a year ago. It will pull up your actual lipid panel, compare it against your last three visits, cross-reference the statin in your medication list, and then tell you what your resting heart rate and sleep data suggest you do about it. That capability — a Gemini model reading a consumer's full clinical history — is now live for U.S. users, and it is the single most consequential expansion of what a fitness tracker is allowed to know about you.
Google announced the medical records integration at The Check Up, its annual health event, on March 17, 2026, opened it to U.S. Public Preview users in early April, and folded it into general availability on May 19 when the Fitbit app was rebranded as Google Health and the Personal Health Coach became the Google Health Coach. In the months since, the feature has moved from a preview curiosity to the default assumption behind Google's consumer health strategy — and the reason the company is now competing directly with OpenAI, Anthropic, and Microsoft for the right to hold your chart.
How the records actually get in
The mechanics matter more than the marketing. Users can search for a doctor or health system by name or address, enter a few details, and log into the provider's portal directly. "To get started, we make it really easy for patients to find their doctor or health system by searching for a name or address," said Rishi Chandra, VP/GM for Health and Home at Google, at the Check Up event. "Once you have found your provider, enter a few quick details and then simply log in to your provider's portal. After we verify your identity, your historical and future records are synced."
The more aggressive path skips the portal hunt entirely. Google partnered with CLEAR, b.well, and Athenahealth, and is coordinating with the Centers for Medicare and Medicaid Services, to run identity-based record retrieval against IAL2-certified standards. "You first verify your identity with Clear by sharing a selfie and then taking a photo of a valid form of ID," Chandra said. "Once verified, we can use that identity to try to locate your medical records automatically so the records come to you." In practice, a selfie and a driver's license can pull records from providers a user may not even remember visiting — availability varies by provider.
Once synced, lab results, medications, and visit history sit inside the app, and the user can hand them to the coach. Google is also rolling out Smart Health Link sharing, which packages a health summary into a URL or QR code that can be added to Google Wallet and handed to a family member or a new physician. The whole bundle sits behind Google Health Premium at $9.99 a month or $99.99 a year — up from Fitbit Premium's old $79.99 annual price — and is included at no extra cost for Google AI Pro and Ultra subscribers in more than 30 countries, provided they own a paired Fitbit or Pixel Watch.
The clinical framing is deliberately modest. "When your coach understands your medical history, its guidance becomes safer, more relevant and more personalized," wrote Florence Thng, Health Intelligence Product Management Director at Google, in the launch blog. Google pairs that with an unambiguous disclaimer: Fitbit "is not intended to use your medical records to diagnose, treat, cure, prevent, or monitor any disease or condition."
Why It Matters
That disclaimer is doing enormous regulatory work. By declining to make any diagnostic or monitoring claim, the coach stays outside FDA medical-device classification — the same general-wellness carve-out that has let a decade of consumer health features ship without premarket review. Meanwhile, the underlying models are getting real clinical scaffolding: Google shipped a 15% improvement in sleep staging accuracy validated against polysomnography, and published work in Nature on predicting insulin resistance from wearable data. The upside is genuine. Roughly 75% of U.S. adults say they want digital health tools that deliver personalized insights and about 70% want wearable data integrated with their medical records, per a July 2025 Verily survey. Most patients cannot get a clinician to sit down and explain a lipid panel; a model that has read the panel and the last two years of sleep data can at least try.
The risk is structural, not hypothetical. Fitbit is not a HIPAA covered entity. The records are protected by Google's contractual promises — that medical data is never used for ads, that users can delete it — not by federal health privacy law. "Health data collected by Fitbit, for example, is not governed by the HIPAA Privacy Rule," University of Illinois Chicago health informatics professor Andrew Boyd has noted of consumer wearables. "It is governed by contract law and the licensing agreement with your software provider." A policy change, an acquisition, or a subpoena operates on very different terms than a hospital's chart would. And an LLM summarizing a real oncology history or a psychiatric medication list has a hallucination surface that no wellness disclaimer neutralizes — the failure mode is not a wrong step count, it is a confidently wrong reading of a lab trend that a user acts on before their next appointment.
What to watch
Three things. First, the CMS coordination Chandra promised more detail on "later this year" — if federal record-locator infrastructure becomes the pipe feeding consumer AI apps, the policy question stops being about Google and starts being about every app on the other end. Second, whether the FDA's relaxed January 2026 clinical decision support guidance holds as these tools get more assertive. Third, competitive pressure: OpenAI connected ChatGPT to Epic EHRs on September 4, and Microsoft's Copilot Health already pulls from more than 50,000 U.S. providers. The race to hold your chart is now a four-way sprint, and none of the runners are covered entities.
“When your coach understands your medical history, its guidance becomes safer, more relevant and more personalized.”— Florence Thng, Health Intelligence Product Management Director, Google