Google Research: an AI system detects signs of hypertension from a consumer smartwatch

Google Research unveils an AI system capable of detecting signs of hypertension using a consumer smartwatch without the need for a cuff device. The research, published on November 5, 2026, in a Nature paper titled "Opportunistically Detecting Signs of Hypertension on a Consumer Smartwatch," introduces a system called AI-PPG-ACC-HTN. This system leverages two sensors commonly found in smartwatches: photoplethysmography (PPG) and accelerometry (ACC).

The AI-PPG-ACC-HTN system operates in two stages. Initially, a one-dimensional deep convolutional network is trained using self-supervised contrastive learning on 492,347 hours of unlabeled sensor data collected from 2,109 participants. This training employs overlapping 15-second blocks of PPG and accelerometer waveforms. Subsequently, features are averaged across wear days, and a regularized logistic regression combines the aggregated representation with participant height to calculate the probability of hypertension.

In a prospective multicenter study involving 196 participants at five centers across the United States, the model was tested. The results showed a sensitivity of 65.8 percent, specificity of 90 percent, and a positive predictive value of 80.6 percent for primary screening. When compared to real-world wear for seven days, the primary screening achieved sensitivity of 55.3 percent and positive predictive value of 77.8 percent.

Notably, a subset of 48 participants with elevated baseline office pressure was evaluated using a confirmatory test. This subset demonstrated a sensitivity of 78.4 percent, specificity of 90.9 percent, and a positive predictive value of 96.7 percent for the AI-PPG-ACC-HTN system. The model does not provide an estimation of systolic and diastolic pressure at any given moment. Instead, it focuses on detecting sustained hypertensive states by aggregating features across multiple days.

The implications of this research for engineers are significant. Google emphasizes that this is early-stage research and not a replacement for traditional cuff measurements or a definitive diagnosis. The study's findings contribute to the development of the Health Guardian blood pressure tracking feature within the Google Health app.

For engineers, passive screening on mass-market wearable sensors is emerging as a tangible engineering challenge. The transition from signal processing to thresholds, validation, and error handling marks the hard part of this emerging class of tasks in wearable sensor technology.