Google Just Unified the Android Health Stack: What Google Health, the Gemini Coach, and FHIR Medical Records Mean for Developers

In May 2026, Google did something it had been circling for years: it consolidated its fragmented health ecosystem into a single app. Google Health — the rebranded Fitbit app, which began rolling out on May 19, 2026 — pulls together data from Fitbit and Pixel Watch devices, Health Connect, and connected sources like Apple Health and US medical records, then layers a Gemini-powered assistant, Google Health Coach, on top as a paid Google Health Premium feature ($9.99/month) [1][2][7].

For consumers, it’s a cleaner front door. For developers, it’s a signal worth reading carefully: the platform owners are racing to own the intelligence layer of health data, not just the collection layer. This article breaks down what actually shipped, what it changes for teams building on Android health data, and — importantly — what it still doesn’t solve.

What actually launched

Google’s announcement bundles three distinct things that are easy to conflate. Keeping them separate is the key to understanding the developer impact.

1. Google Health (the consumer app)

The unified app aggregates previously scattered sources — wearables, Health Connect, Apple Health, and clinical records — into one timeline. Alongside the launch, Google published a roadmap of 39+ features and fixes rolling out through summer 2026 [2], including:

• A 24-hour total sleep view and easier nap discovery
• A fix for missing Sleep Scores — a long-standing complaint
• Apple Health integration, so iOS users aren’t locked out
• Smart Health Links for sharing medical records
• Health Coach message tuning and an “Ask Coach” that can log data like core body temperature

The sleep-score fix is a small line item with a large lesson, which we’ll return to below.

2. Google Health Coach (the Gemini layer)

The headline feature is an AI coach built on Gemini that reasons over the unified data set [1]. It exits public preview as a paid Google Health Premium feature ($9.99/month, or $99/year) [7] — Google’s bet that AI coaching is what finally gets consumers to pay a recurring fee for their own health data. It’s also Google’s answer to the “what do I do about it?” problem: the gap between a number on a screen and an action a user will take.

3. Health Connect Medical Records (the developer layer)

Underneath the consumer app, Health Connect continues to be the API and on-device data store that Android apps read from and write to. Its Medical Records feature extends the platform to handle clinical data in FHIR format, with read/write APIs and a permissions UI for users to grant or revoke access [3]. This builds on capabilities Google added in prior cycles — background reads and history reads — that let apps access Health Connect data without forcing the user to open the app first [4].

Why this is a strategic move, not just a UI refresh

Health data has always had a value-chain problem. The raw data lives in silos. Aggregating it is hard. But the real value — turning data into something a person will act on — sits at the top of the stack. For years, that top layer was wide open. Google Health is Google’s move to claim it.

The pattern is worth naming explicitly, because it’s the same pattern playing out across the industry:

When the platform owner starts shipping the top two layers, app teams face a strategic question: where does your product actually differentiate? If your value proposition was “we show users their Fitbit data in a nicer chart,” Google Health just ate your lunch. If your value is a specific, opinionated experience built on health signals, the platform’s move is a tailwind — it normalizes the behavior of acting on health data and grows the market.

The contrast with Apple is the real story

Google shipping a working Gemini coach in May 2026 is more striking when you set it against Apple’s quarter.

According to multiple reports, Apple delayed its own AI health coach — internally codenamed Project Mulberry — and the associated Apple Health+ subscription, pushing them past the WWDC 2026 launch window. The reported reason is instructive: Apple chose to first overhaul the reliability of its underlying biometric telemetry, including a rebuild of the Apple Watch’s heart-rate tracking engine, before layering AI coaching on top [5][6].

That decision encodes a principle every team building on health data should internalize:

An AI coach is only as trustworthy as the data underneath it. A confident recommendation built on a noisy sleep estimate or a misattributed heart-rate reading doesn’t just fail to help — it actively erodes user trust.

Google and Apple have made opposite bets on sequencing. Google shipped the coach and is fixing data quality in parallel (note the “missing Sleep Scores” fix in the same roadmap as the coach launch). Apple is hardening the data first and shipping the coach second. Neither is obviously wrong — but the tension between them is the defining product question in consumer health right now.

What Health Connect still doesn’t solve for developers

Health Connect is genuinely good infrastructure, and the medical-records and FHIR additions are meaningful. But it’s important to be precise about what it is and isn’t, because the gaps are exactly where teams underestimate the work.

It’s Android-only. Health Connect collects Android data. Your iOS users live behind HealthKit, a completely separate integration with a different schema, permission model, and data semantics. Supporting both is two integrations, not one.

It doesn’t normalize across brands. Health Connect is a data store, not a referee. When a user has an Apple Watch, an Oura Ring, and a phone all writing to it, you get overlapping, duplicated, and sometimes contradictory records. Health Connect hands you the pile; deduplicating and reconciling it is your problem. (We’ve written about why those sources disagree — the disagreement is real and unavoidable.)

It doesn’t derive intelligence. Health Connect gives you steps, heart rate, and sleep stages. It does not give you a sleep score, a readiness signal, a behavioral archetype, or a trend. The “missing Sleep Scores” fix in Google Health is a reminder that even Google finds scoring non-trivial — and that scoring is a product layer, not a free byproduct of collection.

FHIR is for clinical data, not fitness metrics. The Medical Records feature is excellent for structured clinical records. But mapping consumer fitness signals (HRV windows, step cadence, sleep regularity) into FHIR resources adds complexity without clear benefit for most consumer use cases. FHIR solves a clinical interoperability problem, not a wearable-fragmentation one.

What this means for builders

The practical takeaways for product teams:

1. Treat Health Connect as one input, not the whole stack. It’s the best way to collect Android data. It is not a cross-platform, normalized, intelligence-bearing layer. Budget for the iOS side and the normalization layer separately.

2. Don’t compete with the platform on collection — compete on experience. Google and Apple will always win the collection layer. Your differentiation lives in what you do with the data: the specific decisions, timing, and experiences your product creates.

3. Invest in data quality before AI features. Apple’s sequencing is the lesson. If you’re building an AI coach or personalization engine, the trustworthiness of the underlying scores and biomarkers determines whether users believe — and keep using — the feature.

4. Plan for derived signals as a build-or-buy decision. Scores, biomarkers, and behavioral segmentation are their own engineering discipline — be honest about whether rebuilding them in-house is the best use of your team’s time, or whether it’s table stakes you should buy.

This is where infrastructure that sits above Health Connect and HealthKit earns its place. A normalization layer that collects across both platforms, deduplicates overlapping sources into a single record, and derives the scores and behavioral signals that platform stores leave to you turns the collection layer into a foundation rather than a recurring project — it’s the layer we work on at Sahha. As Google and Apple turn health data into a coaching battleground, the teams that win will be the ones spending their engineering on a differentiated experience, not rebuilding sleep scoring for the third time.

Google just raised the bar on what a health experience should feel like. The opportunity for builders is to clear it with something the platform can’t ship: a product that actually knows its users.