A University of Michigan review found many smartwatch health metrics are algorithmic estimates rather than direct body measurements, making single readings less dependable than longer-term trends.
Sensors such as optical heart-rate monitors, motion trackers and GPS collect raw signals, then proprietary software converts them into user-facing numbers using assumptions and personal characteristics.
Resting and steady-state heart rate, step count and outdoor pace were generally more reliable, while calories burned, sleep stages, body composition, hydration and recovery proved less dependable.
Accuracy also shifts with watch fit, movement, temperature, sweat, skin tone, tattoos and body composition, and results may not match across brands because devices use different sensors and definitions.
The findings, published in Sensors after a narrative review through June 2026, reinforce earlier concerns that complex wearable outputs can mislead users if treated as precise lab-grade measurements.