Google Research, in partnership with Beth Israel Deaconess Medical Center and Harvard Medical School, released results showing that a smartwatch‑based artificial‑intelligence model can screen for high blood pressure without a cuff. The study, published in Nature on November 5, 2026, monitored 196 volunteers for seven days using ordinary pulse‑wave and accelerometer data from a consumer‑grade watch. When compared with 24‑hour ambulatory blood‑pressure monitoring, the AI system—named AI‑PPG‑ACC‑HTN—identified 65.8% of participants who met clinical criteria for hypertension, while correctly labeling 90% of normotensive subjects as negative.

ETH / USD
Source: CoinGecko. Prices are indicative.

How the algorithm works

The model combines optical photoplethysmography (PPG) signals with motion data from the watch’s accelerometer, aggregating measurements across multiple days. Height is also incorporated to generate a probability score for hypertension. Unlike traditional cuff‑based devices, the algorithm does not output a specific systolic or diastolic value; instead, it provides a binary screening outcome that can trigger a recommendation for formal clinical assessment. The underlying waveform model was pretrained on 492,347 hours of unlabeled sensor data collected from 2,109 consenting participants, giving it a broad base of physiological patterns before being fine‑tuned on the study cohort.

Study design and performance metrics

Researchers enrolled participants with no known cardiovascular disease and who were not on blood‑pressure medication, creating a risk‑enriched sample that included a higher proportion of hypertensive individuals than the general population. Over the monitoring period, 80.6% of positive smartwatch screens corresponded to hypertension on the reference test, yielding a positive predictive value of roughly four‑fifths. Sensitivity confidence intervals ranged from 54% to 76.3%, while specificity confidence intervals spanned 83.2% to 94.7%. The AI missed about one‑third of true hypertension cases, underscoring the need for confirmatory cuff‑based measurements before any diagnostic decision.

Implications and next steps

According to Google Research, the technology could serve as an early‑warning system, prompting users to seek professional evaluation when a watch‑based screen flags potential hypertension. The cuff‑free approach may lower barriers to regular blood‑pressure monitoring, especially for individuals who rarely visit a clinic. However, the study’s limited demographic—participants without existing cardiovascular treatment—means results may not translate directly to patients already on antihypertensive therapy. Further trials with broader, more diverse populations are required to validate the algorithm’s utility across age groups, ethnicities, and comorbid conditions.