Hypertension affects over one billion people globally, with more than half of affected adults unaware of their condition, according to data noted by Aiunderstanding. Against that backdrop, Google Research published a study in Nature on November 5, 2026, introducing an artificial intelligence system that detects signs of hypertension using standard consumer smartwatch sensors without requiring any arm cuff calibration. Developed in collaboration with Beth Israel Deaconess Medical Center and Harvard Medical School, the system bypasses traditional calibration requirements by analyzing passive physiological signatures over multiple days.
How Sensors and Machine Learning Process Wrist Data
Dubbed AI-PPG-ACC-HTN, the multimodal system uses data from two sensors already active on modern consumer smartwatches: photoplethysmography (PPG) to track changes in blood volume at the wrist by measuring reflected light, and accelerometers (ACC) to capture motion, posture, and artifact data. Rather than trying to calculate precise systolic and diastolic blood pressure numbers at a single moment—which are easily thrown off by temporary environmental disturbances—the model zeroes in on recognizing a person’s constant, ongoing hypertensive condition by compiling passive biological indicators across several days.
To build the system, researchers used a two-stage machine learning approach. In the first stage, a 1-D deep convolutional neural network was trained using self-supervised contrastive learning on a large, unlabeled dataset of 492,347 hours of sensor data from 2,109 consenting participants. This allowed the model to learn complex physiological features of vascular and cardiac function without manual feature engineering or medical labels. In the second stage, the model combines these extracted features with user height as a proxy for arterial path length, feeding them into a logistic regression classifier.
Evaluating Clinical Validation Across Five United States Sites
The system was validated in a prospective, multicenter clinical study involving 196 participants across five United States sites who had no known cardiovascular disease and took no antihypertensive medications.
For confirmatory testing among a subgroup of 48 participants with elevated initial office blood pressure, the smartwatch system achieved a sensitivity of 78.4 percent, a specificity of 90.9 percent, and a PPV of 96.7 percent, outperforming return office blood pressure measurements, which yielded a sensitivity of 67.6 percent, specificity of 63.6 percent, and a PPV of 86.2 percent. During initial screening, the smartwatch-based AI achieved a sensitivity of 65.8 percent, a specificity of 90.0 percent, and a positive predictive value (PPV) of 80.6 percent. By comparison, office-based screening showed a sensitivity of 55.3 percent and a PPV of 77.8 percent at the same 90.0 percent specificity.
Home Monitoring Yields Higher Sensitivity than Office Screening
Further evaluation in the study included multi-day home monitoring, which yielded a sensitivity of 89.2 percent, a specificity of 81.8 percent, and a positive predictive value of 94.3 percent. The initial office-based screening strategy specifically achieved 55.3 percent sensitivity with a 95 percent confidence interval ranging between 43.4 percent and 66.7 percent, alongside a specificity of 90.0 percent with a 95 percent confidence interval spanning 83.2 percent to 94.7 percent.
The study’s authors acknowledged numerous contributors to the research effort. The writing team and co-authors of the paper involved Lawrence Cai, Jiang Wu, Mingwu Gao, Bhavna Daryani, Jonathan Wang, Utkarsh Khanna, Ho Ko, Anupam Pathak, Mark Malhotra, Shwetak Patel, Jacqueline B. Shreibati, Stephen P. Juraschek, Pramod Rudrapatna, and Matthew Thompson. Key contributors to the work also included Jennifer Block, Anu Gabbita, Sushalini Sunku, Keerthana Natarajan, Alexandros Pantelopoulos, Tuan Phan, Tracy Giest, and Florence Thng.
Bringing Passive Screening to the Google Health App
While the findings helped inform the blood pressure trends feature in the Google Health app, researchers emphasize that the technology is currently a research project rather than a standalone diagnostic tool or medical device. The passive screening tool aims to gently prompt at-risk individuals, particularly younger adults aged 18 to 44 where missing diagnoses can delay early interventions, to visit a clinic for formal assessment and confirmatory diagnosis using established cuff-based methodologies.