How Google AI Detects High Blood Pressure on Smartwatches

High blood pressure affects more than one billion people worldwide, yet over half of adults with the condition do not know they have it. Google Research published a Nature paper on November 5, 2026, describing an AI system that uses ordinary smartwatch sensors to flag signs of hypertension without any cuff calibration.

The system, called AI-PPG-ACC-HTN, analyzes photoplethysmography (PPG) from the optical pulse sensor and accelerometer data that captures both motion and tiny wrist vibrations from each heartbeat. It combines these signals over several days with the wearer’s height to output a probability of persistent high blood pressure rather than a single systolic/diastolic number.

How the Model Works

Researchers pretrained the waveform encoder with self-supervised contrastive learning on 492,347 hours of unlabeled sensor data from 2,109 participants. The model learns which short segments come from the same person by comparing pulse and motion patterns, without needing diagnostic labels.

In everyday use the watch records passively. After enough data (typically several days with about one hour of usable still periods per day), the system averages patterns and produces a hypertension probability score. This approach differs from most current cuffless features that still require periodic calibration against a traditional arm cuff and only track relative changes.

Clinical Validation Results

Google validated the system in a prospective multicenter study of 196 adults across five U.S. sites. Participants had no known cardiovascular disease and were not taking blood pressure medication. They wore the smartwatch for seven days while also completing 24-hour ambulatory blood pressure monitoring (ABPM), the clinical gold standard.

For initial screening the smartwatch achieved 65.8% sensitivity and 90.0% specificity, with a positive predictive value of 80.6%. Initial office-based cuff measurements in the same group reached only 55.3% sensitivity at the same specificity and a PPV of 77.8%. In a confirmatory subset of people who already had high clinic readings, the watch confirmed sustained hypertension with 78.4% sensitivity and 96.7% PPV, outperforming a repeat office visit.

These numbers are comparable to traditional screening workflows, not superior in a statistically significant way for every metric. The practical difference is that screening happens continuously during ordinary life instead of only at a clinic visit.

What Users Can Do Today

The findings informed the blood pressure trends feature already present in the Google Health app (also called Health Guardian on some Fitbit and Pixel Watch devices). The feature delivers a monthly summary when it detects a persistently high trend and prompts users to confirm with a home cuff or talk to a clinician. It does not provide a diagnosis or replace clinical measurement.

To use existing Google Health blood pressure trends:

1. Pair a compatible Fitbit or Pixel Watch with the Google Health app.
2. Enable Health Guardian or blood pressure trends in the app settings if available on your device.
3. Wear the watch regularly, including during still periods such as sleep or seated work, so the model has enough clean data.
4. Review the monthly summary and follow the prompt to check with a validated cuff or healthcare professional if a high trend appears.

The research system itself remains early-feasibility work and is not yet a regulated diagnostic feature. Google emphasizes that any flag is only a prompt for formal clinical assessment.

Tips for Better Data Quality

Wear the watch snugly but comfortably on the wrist recommended by the manufacturer. Clean the sensor regularly and avoid heavy motion during key measurement windows. Consistency over multiple days matters more than any single reading. If you already track other health metrics, combine the trend summary with your existing routine rather than treating the watch as a standalone medical device.

People with known heart conditions, those on blood pressure medication, or anyone with symptoms should continue standard clinical care. The study excluded those groups, so real-world performance may differ.

Sources and Further Reading

Official details appear in the Google Research blog post “Detecting signs of hypertension using wearable sensors” (November 5, 2026) and the Nature paper “Opportunistically Detecting Signs of Hypertension on a Consumer Smartwatch.” Related wearable setup steps are covered in our guide to setting up Fitbit Edge with the Google Health app.

The research shows that passive sensing already present on millions of wrists can surface a major silent risk factor and encourage earlier clinical checks. It does not eliminate the need for proper diagnosis, but it expands the moments when screening can happen.