Apple Watch AI May Detect Pregnancy with 92% Accuracy, New Study Reveals

Apple Watch is already famous for tracking vast amounts of health information, but a new research study indicates its capabilities could soon increase exponentially — possibly even to the detection of pregnancy.

A team of researchers backed by Apple has created an AI model that can predict several medical conditions with accuracy based on behavioural patterns tracked by the Apple Watch and iPhone, without using raw biometric data.

The research, "Beyond Sensor Data: Foundation Models of Behavioural Data from Wearables Improve Health Predictions", was recently spotted by 9to5Mac. It presents the Wearable Behaviour Model (WBM) — a new machine learning framework that transforms attention away from classic sensor data (such as heart rate or SpO₂) towards prolonged behavioural patterns. These are user mobility, sleep, heart rate variability, and activity levels — all features Apple's devices already monitor and study.

Built as part of the ongoing Apple Heart and Movement Study (AHMS), WBM was trained on a record 2.5 billion hours of data from more than 160,000 participants who agreed to share their wearable data. The AI model was tested across 57 diverse health prediction scenarios, and in the researchers' words, it surpassed traditional models based on short-term sensor data every time.

AI-Powered Pregnancy Detection
Among the most striking advances was the model's capacity for pregnancy detection. By merging Apple Watch's photoplethysmography (PPG) signals with behavioral observations, WBM gained a maximum of 92% accuracy in detecting early-stage pregnancy. The scientists discovered that slight mobility changes in gait and sleep behavior were among the strongest predictors — even more precise than the changes in heart rate alone.

Instead of trying to supplant conventional sensor data, the researchers suggest a hybrid model: applying behavioral analysis to spot long-term patterns, while still taking advantage of biometric spikes for short-term or acute health incidents. This two-input system proved to work elsewhere, such as to identify respiratory infections, monitor recovery from injury, or even whether a user may be taking medications such as beta blockers.

The WBM model is based on a time-series AI framework known as Mamba-2, which is effective at detecting changes in human habits over time. This allows the AI to capture slight but regular variations in a user's lifestyle that can indicate underlying changes in health.

While Apple hasn't officially indicated pregnancy detection as a future feature, this study heavily suggests what could be on the horizon for future Apple Watch versions. The study is an important advancement in AI-assisted health diagnostics, and further erases the distinction between wearables and proactive medicine.

As Apple continues to put money into healthcare innovation, the discoveries may open doors for increasingly anticipatory and personalized features — enabling devices such as the Apple Watch not only to monitor health, but to enable people to comprehend and act upon it better.