AI in Personal Health Tracking: The Benefits, the Risks, and How Accurate Wearables Actually Are

How accurate are they, really?

This is where the marketing and the evidence diverge most. Physicians reviewing the current evidence generally describe a consistent pattern: wearables tend to be quite reliable for resting heart rate and step counts, but show meaningfully larger error margins for blood pressure, arrhythmia detection outside of AFib specifically, and blood oxygen (SpO2) readings during everyday use rather than controlled testing.

Several factors drive this variability:

• Sensor and hardware quality, which differs substantially between a $25 smart ring and a clinical pulse oximeter.
•  How well the underlying algorithm was validated, and against what population.
• User behavior — fit, skin contact, and movement all affect optical sensor readings.
• Demographic bias in training and validation data. This is a significant, underappreciated problem: a 2026 bench study from Johns Hopkins Biomedical Engineering quantified how skin tone affects the accuracy of the light-based (photoplethysmography) sensors most wearables rely on, and multiple device studies have found meaningfully reduced accuracy in people with darker skin tones (Fitzpatrick V–VI) and low BMI — populations that were often underrepresented in the datasets used to train and validate these algorithms in the first place.

Brand-level differences are real, too. A widely cited 2025 meta-analysis of AFib-detection studies found real variation in sensitivity and specificity across major brands, though the strength of evidence itself varies by brand — some manufacturers simply have far more independent research behind their claims than others.

The regulatory reckoning

The line between "wellness product" and "medical device" has become one of the most contentious issues in the industry, and it came to a head in July 2025, when the FDA sent a warning letter to WHOOP over its Blood Pressure Insights feature. The FDA's position was that estimating blood pressure is inherently a medical function — tied directly to diagnosing hypo- and hypertension — regardless of how a company markets it, and it specifically flagged WHOOP's own description of the feature as delivering "medical-grade" insights as evidence that the company intended more than general wellness use. WHOOP publicly disagreed and kept the feature on the market through most of its dispute with the agency; a class-action lawsuit citing the warning letter followed in November 2025. Roughly a year later, in June 2026, the FDA told WHOOP it would not pursue further enforcement after the company adjusted the feature — but the underlying regulatory question the case raised hasn't gone away, and the FDA has since issued broader safety communications warning consumers generally about unauthorized blood-pressure-measuring smartwatches and rings.

The bigger picture: as of 2026, the FDA is applying more scrutiny to wearable features that touch on genuinely medical functions (blood pressure, arrhythmia detection, glucose trends) even when they're marketed under the more lightly regulated "general wellness" category — a legal exemption created for products intended to support a healthy lifestyle rather than diagnose or treat disease.

'