The Journal of mHealth posted a detailed article on how wearables and smartphones attempt to measure blood pressure without the traditional method of measuring the external pressure sufficient of occlude the blood flow.
Bottom line: Most of these devices use Pulse Wave Velocity to estimate how much the blood pressure has changed from previous readings. They require frequent calibration using a conventional sphygmomanometer. Some of the more sophisticated techniques combine sensor data with machine learning and can give good estimates of blood pressure, but may not be useful in some important circumstances:
The collateral data alone can make a good estimate of your blood pressure. What it finds is the normal blood pressure for a person of your age, height, sex and weight. Since (by definition) most of us are normal, this gives a good estimate in most cases. Of course, it does not find the people who are abnormal, for whom accurate and early detection is vital. The international standard for blood pressure meters measure the average accuracy, so wearables or devices that tell you what your blood pressure should be can pass because the few anomalies are lost in the large number of normal people.
It’s worth reading the entire article: thejournalofmhealth.com/wearables-smartphones-and-apps-do-they-really-measure-blood-pressure