Microsoft Copilot Health entered public preview on May 29, 2026, marking the company’s most ambitious move yet into AI-powered personal health management. The new web experience is available immediately to U.S. users aged 18 and older who hold a Microsoft 365 Personal or Family subscription. Built with a sharp focus on security, the preview brings four core capabilities: sleep analysis, lab result interpretation, health record summarization, and care search. It is the first Copilot branded product dedicated entirely to health, following years of Microsoft’s incremental investments in cloud, AI, and wellness.

The Launch and Availability

The preview release landed on a Friday afternoon, a quiet window that suggests Microsoft intends to gather measured feedback before a broader rollout. Access is restricted to the United States and requires a valid Microsoft 365 Personal or Family subscription. Users must be at least 18 years old, reflecting both the sensitivity of health data and the company’s compliance with age-specific privacy regulations. There is no native mobile app at this stage—Copilot Health lives exclusively on the web, accessible through a modern browser. Microsoft has not announced plans for integration into Windows 11, Microsoft Teams, or the main Copilot sidebar, though such expansions would align with its historical pattern.

The gateway is simple: Microsoft 365 subscribers navigate to a dedicated Copilot Health portal, authenticate with their Microsoft account, and grant consent to process health information. First-time users are walked through a privacy setup that clearly enumerates data sources, retention policies, and user controls. The experience is offered at no additional cost during preview, but Microsoft has hinted that certain premium features might eventually be bundled into higher-tier subscriptions.

What Is Copilot Health?

Copilot Health is not a medical device or a diagnostic tool. Microsoft positions it as a “secure AI companion” that helps individuals make sense of their own wellness data. It synthesizes information from multiple sources—a user’s manually entered health metrics, uploaded lab reports, connected wearable device data, and self-reported symptoms—and generates plain-language insights, suggestions, and organized summaries. The system runs on a specialized instance of Microsoft’s large language models, isolated from the main Copilot consumer stack and tuned exclusively for health content.

At its core, the service attempts to solve a universal pain point: fragmented health information. Most people track sleep with one app, receive lab results from a separate patient portal, and store medical history across half a dozen clinics. Copilot Health aggregates these threads into a single, searchable conversation. The AI can cross-reference data, such as noticing when a new lab result falls outside a personally established range or when recent sleep disturbances coincide with medication changes logged in the health record.

Feature Deep Dive: Sleep Analysis

Sleep tracking has become a commodity feature in wearables, but Copilot Health advances the concept by combining raw data with contextual interpretation. Users can connect devices like the Microsoft Band successor (or any compatible smartwatch via Health Connect) to import sleep stages, heart rate variability, and oxygen saturation. Alternatively, manual entry works for those without wearables.

Once data flows in, Copilot generates a nightly sleep report that goes beyond a simple score. It compares sleep architecture against the user’s historical baseline, highlights anomalies, and correlates them with lifestyle factors the user has shared. For example, if a user mentions they started a new evening medication or changed their workout time, the AI will flag corresponding shifts in deep sleep or REM cycles. During preview, the feature avoids making clinical recommendations; it sticks to observational statements such as “Your average REM sleep has decreased by 18% over the last week, which coincides with your reported increase in late-night screen time.”

Privacy is reinforced by on-device processing where possible—sleep data is encrypted in transit and at rest, and the AI does not retain raw biometric signals after analysis. Users can delete any night’s data with a single click, and a timeline view enables easy auditing of what the system has stored.

Feature Deep Dive: Lab Results Interpretation

Lab results arrive as PDFs, portal screenshots, or CSV exports—often filled with abbreviations and reference ranges that leave patients confused. Copilot Health accepts these documents through a secure upload tool. Using optical character recognition and a health-specific natural language parsing layer, it extracts every biomarker and maps it to plain English.

The output mimics a conversation: “Your A1C is 5.4%, which is within the normal range. This has improved from 5.7% six months ago, suggesting your dietary changes are working. Keep an eye on your fasting glucose, however—it’s trending slightly upward.” The AI draws on evidence-based guidelines but never oversteps into prescribing treatment. Each summary is accompanied by a disclaimer urging users to consult their physician before making health decisions.

Users can ask follow-up questions in natural language. “What lifestyle changes could lower my LDL?” prompts Copilot to retrieve public-domain educational content from the National Library of Medicine, CDC, and similar sources—all clearly cited. It will not answer direct queries about dosage, diagnosis, or predicted disease risk, a boundary that Microsoft has implemented through a dedicated safety classifier.

Feature Deep Dive: Health Records Summarization

For people managing chronic conditions, a sprawling timeline of doctor visits, procedures, and prescriptions is hard to navigate. Copilot Health ingests Continuity of Care Documents (CCDs) and FHIR data that many health systems now provide to patients. It parses these structured records to build a cohesive narrative: a chronological condition summary, an active medication list, upcoming screening reminders, and a care team directory.

The AI can generate a one-page “visit prep” sheet to bring to appointments. This sheet pulls forward recent changes, open questions the user has typed in, and a list of current medications with dosages—all formatted clearly. During the preview, the system warns users to verify accuracy and offers an editable version for corrections.

Importantly, the summarization is not reliant on a cloud-based personal health record (PHR) like Google Health’s failed attempt. Copilot Health stores records encrypted in the user’s Microsoft account envelope, with access governed by per-document permissions. Microsoft claims it cannot read the files; they are decrypted only in the user’s browser session.

Feature Deep Dive: Care Search and Recommendations

The fourth pillar of the preview helps users find appropriate medical services. A simple search bar lets someone type “cardiologist near me that takes my insurance” or “urgent care open late.” Copilot Health uses a local provider database augmented by user reviews and insurance network feeds to return ranked, annotated results. It can also factor in the user’s recorded health history: if a person has a history of migraines, searching for “neurologist” will prioritize practitioners with that sub-specialty when such information is available.

Care search is integrated with Microsoft’s Bing Maps, so users see locations, hours, and estimated wait times on a map interface directly inside the Copilot conversation. A notable privacy promise: the system does not send any personal health data to Bing or third-party APIs during the search. Provider matching happens entirely within the Copilot Health environment, using anonymized criteria.

Security and Privacy Design

Security is the headline the product team wants to own. Copilot Health operates within a dedicated “health boundary” inside Microsoft’s cloud. According to the preview documentation, user prompts, uploaded documents, and AI-generated responses are never used to train foundation models or to target advertising. Data is encrypted with customer-managed keys for M365 subscribers who opt in, and advanced threat protection monitors the ingest pipeline.

The consent flow is granular: each capability—sleep, labs, records, care search—must be individually enabled. A dashboard shows exactly what data the AI has accessed, and users can revoke permissions at any time. The AI won’t prompt for data it doesn’t already have permission to read, and any attempt to ask about another person triggers an automatic block with a reminder that the tool is for individual use only.

Microsoft has also contracted an independent health data auditor to verify that the system complies with HIPAA where applicable. Though Copilot Health is a consumer product and Microsoft is not acting as a covered entity, the company volunteered to meet HIPAA Security Rule standards as a design benchmark. A shareable privacy report will be available for users who want to review compliance artifacts.

The Competitive Landscape

Copilot Health enters a field littered with cautious optimism and high-profile failures. Apple Health has built a vast data repository but offers minimal interpretation. Google’s various health efforts, including the discontinued Google Health app and the early-launch demise of its PHR, highlight the difficulty of consumer health engagement. Startups like Ada and Babylon (now eMed) tried AI-powered symptom checking with mixed regulatory outcomes.

What differentiates Copilot Health is its deep integration with the Microsoft 365 ecosystem. Users can already export Copilot Health summaries to Word for sharing with a doctor, or add care plan reminders to Outlook Calendar. The underlying language model benefits from Microsoft’s partnership with OpenAI, but the health instance runs on a separate fine-tune that has been heavily constrained for safety. This blend of productivity and health is new; no competitor owns the desktop-to-cloud stack that Microsoft commands.

Still, the preview faces skepticism. Many consumers trust their healthcare provider’s portal, not a generic AI. Microsoft will need to prove that its privacy promises hold and that the AI reduces—rather than adds to—anxiety. The exclusion of users under 18 also limits its reach for family health management, though Microsoft says pediatric features are under evaluation.

Early Impressions and Limitations

During the preview, Microsoft has imposed deliberate constraints. The AI refuses to answer questions about mental health crises, suicidal ideation, or weight loss planning without immediate escalation to crisis resources. Every session begins with a reminder that Copilot Health is not a substitute for professional medical advice. The system has also been tuned to be conservative with drug interaction queries, deferring entirely to a pharmacist unless the interaction is well-documented and non-critical.

Users testing the preview have noted that lab result parsing struggles with handwritten or poorly scanned documents—a known issue that is common across OCR tools. Sleep analysis, while detailed, currently lacks integration with non-wearable sleep trackers like bedside smart hubs. The care search database, while robust, has gaps in rural areas. Microsoft openly acknowledges these limitations and plans iterative updates roughly every two weeks during preview.

Performance is snappy on Edge and Chrome, but Firefox users have reported occasional formatting errors—likely a preview-era bug. The web-only nature frustrates those who want offline access to their health summaries. There is no API or export standard for developers yet, though Microsoft promises FHIR export by the time Copilot Health reaches general availability.

What’s Next for Copilot Health

Microsoft has publicly shared a six-month preview roadmap. Immediate priorities include expanding the provider directory to all U.S. ZIP codes, adding support for Spanish-language lab results, and launching an iOS and Android web app with offline caching. By late 2026, the team expects to introduce an opt-in research mode where de-identified data can contribute to academic studies—with compensation for participants in the form of Microsoft Rewards points.

Longer term, Microsoft engineers have hinted at integration with Windows 11’s health dashboard, which currently tracks basic device usage and posture. A unified well-being panel that brings together digital health and physical health metrics is being explored. There is also talk of a Copilot Health API that telehealth startups could embed into their own platforms, creating a B2B2C channel.

The preview’s success will be measured by engagement and trust metrics rather than raw user counts. Microsoft plans quarterly transparency reports detailing data access by law enforcement and any security incidents. If Copilot Health can maintain user confidence while delivering genuinely useful insights, it might finally crack the consumer health code that has eluded big tech for a generation.

Copilot Health is a calculated bet that AI’s interpretive power can turn scattered health data into a coherent story. The preview answers the “what if” with a carefully sandboxed experience that respects the sensitivity of its domain. For Microsoft 365 subscribers curious about what their own numbers mean, the new tool offers a compelling first look—but it lands with the appropriate weight of responsibility that health AI demands.