Apple Watch may soon detect pregnancy with 92% accuracy, thanks to AI breakthrough

Apple Watch, known for its robust health-tracking capabilities, might soon gain a groundbreaking feature, the ability to detect pregnancy with remarkable precision. A new research study has demonstrated that an AI model developed with Apple’s support can predict health conditions like pregnancy using only behavioural data gathered from devices such as the Apple Watch and iPhone.

Also Read:Apple names Indian-origin executive Sabih Khan as new chief operating officer

The study, titled “Beyond Sensor Data”, introduces a next-gen AI system called the Wearable Behaviour Model (WBM). Unlike traditional methods that rely on direct biometric readings like heart rate or blood oxygen, WBM is trained on long-term patterns in user behavior, including sleep quality, activity trends, heart rate variability, and movement data.

This model is a result of the Apple Heart and Movement Study (AHMS), one of the largest health research efforts of its kind. Over 160,000 participants contributed more than 2.5 billion hours of wearable data, allowing researchers to train WBM across 57 different health prediction tasks.

In the case of pregnancy detection, the AI achieved up to 92% accuracy, identifying subtle weekly changes in user routines. Key indicators included alterations in gait, mobility, and sleep duration, signals often missed by traditional sensors.

WBM is powered by a new AI architecture called Mamba-2, optimized for time-based data, making it ideal for monitoring behavioural shifts over time. Researchers emphasize that this model doesn’t replace real-time biometric monitoring, but instead complements it. Combining raw sensor data with behavioural insights proved to be the most effective approach in predicting various conditions, from respiratory infections to medication effects.

Apple has not yet announced an official rollout, but this study opens the door for a major expansion of wearable health diagnostics in the near future.