A smartwatch system turns daily data into a health age and a 12-week plan

Google Research describes an investigational system, meaning a research prototype, that turns smartwatch signals into a "coachable health age" and simulates 12 weeks of new habits. It has been tested only offline, on past research data.

Share
A smartwatch system turns daily data into a health age and a 12-week plan
Illustration: AI-generated

A smartwatch records resting heart rate, steps, sleep and active minutes. In a blog post dated 11 November 2026, Google Research describes an investigational system, meaning a research prototype, that turns those signals into a "coachable health age". It then simulates how a person's body might respond over 12 weeks to different habits.

The health age uses only daily behaviours and body measurements that people can change. Google Research compares it with the person's calendar age. A health age below calendar age means the body's profile looks younger; one above it signals faster ageing.

Google Research says its analysis of the All of Us research programme, which covered 22,748 people, found that healthy people whose health age was older than their calendar age were more likely to develop high blood pressure, heart disease or type 2 diabetes over the next 8 years.

From 2 weeks of one person's data, the simulation part, called Sensor World Model, forecasts how 20 body measurements may change over 12 weeks under different step, sleep and cardio routines. It was trained on 2.28 million person-weeks of data (one person's data for one week) from roughly 100,000 consenting Fitbit users, with identifying details removed.

A filter weighs projected gains against how flexible a person's habits have been and safety limits. In Google Research's example, the target was a 2.3-year drop in health age for a flexible person and 1.3 years for someone with a more fixed routine.

A language model, an AI system that writes text, turns the top options into a step-by-step 12-week plan.

In Google Research's own blinded study, reviewers were not told which plan came from which method. Five certified lifestyle coaches and five board-certified physicians reviewed 49 participant profiles. Each of the 490 paired reviews compared a plan written using the simulations with one written from wearable history alone.

Coaches chose the simulation plan as best overall in 71.4% of evaluations and rated it more likely to help people stick with it in 71.0%. Physicians rated 99.2% of the simulation plans as within safe exercise intensity limits. All of this used past research data.

A companion study of 819 WEAR-ME participants added three routine blood tests, to look at heart, kidney and metabolic strain separately. Google Research says metabolic and heart strain are often correlated, but early kidney decline frequently appears independently.

Google Research calls the models investigational research prototypes, not clinical diagnostic tools. It says they lay the foundation for future prospective trials, which would follow people forward in time.

Sources