Google Research has published a study in Nature describing an AI system that uses the sensors already in consumer smartwatches to flag signs of high blood pressure without requiring a traditional arm cuff for calibration. Hypertension affects more than one billion people worldwide, and more than half of adults with the condition do not know they have it. Because it rarely causes obvious symptoms, it is usually found by chance during routine checks, a gap that is most pronounced among adults aged 18 to 44. Existing cuffless wearables generally track changes from a cuff-measured baseline, which limits their use as standalone screening tools. The Google system instead estimates a person's underlying, persistent hypertension status by combining photoplethysmography, which measures blood volume changes at the wrist, with accelerometer readings that capture the faint vibrations of the wrist produced by blood leaving the heart. Those signals are averaged over several days of wear and combined with the person's height before a classifier produces a probability score. The model was first pre-trained on 492,347 hours of unlabelled sensor data from 2,109 people, then validated in a prospective study at five US sites involving 196 participants who had no known cardiovascular disease and were not taking blood pressure medication. Results were compared against 24-hour ambulatory blood pressure monitoring, the clinical reference standard. For initial screening, the smartwatch system had a sensitivity of 65.8 per cent, a specificity of 90.0 per cent and a positive predictive value of 80.6 per cent, compared with 55.3 per cent sensitivity and 77.8 per cent positive predictive value for office-based readings. In a smaller group of 48 people with elevated office readings, the system's confirmatory results were a sensitivity of 78.4 per cent, specificity of 90.9 per cent and positive predictive value of 96.7 per cent. The researchers say the findings helped inform a blood pressure trends feature in the Google Health app. The sensitivity figures mean the tool still misses a meaningful share of people who have hypertension, and the study population was relatively small and limited to the US. For Australian businesses and professionals, this changes little in the near term. The study does not say whether the feature is available in Australia or how it would be regulated here, and a screening estimate from a watch is not a diagnosis. Any employer, insurer or health provider considering wearable health data should note that under the Privacy Act 1988, health information is sensitive information that requires consent and strong safeguards. Individuals who already use a compatible smartwatch can treat the feature as a prompt to book a proper cuff reading with a GP, not as a substitute for one.