Blood Oxygen (SpO2)
How accurate your wearable's blood oxygen reading actually is, especially the overnight number most devices report, and a real, documented accuracy gap by skin tone the research doesn't let us ignore.
Blood oxygen, or SpO2, shows up on nearly every modern watch and ring as a clean percentage, usually somewhere in the high 90s. It feels like a lab reading. It isn't one. The sensor shines red and infrared light through your skin and infers oxygen saturation from how much of each color bounces back, a method with real, well-documented limits, some of them specific to the device sitting on your wrist or finger, not just to blood oxygen itself. Here's what the research actually says about how well that estimate holds up, and where it holds up less evenly than the single number on your screen suggests.
| Device | What they call it | What it actually measures |
|---|---|---|
| Apple Watch | "Blood Oxygen app" | An on-demand, roughly 15-second spot-check reading, available on Series 6 and later plus Ultra models. Not continuous or automatic overnight in the US market as of this writing. |
| Garmin | "Pulse Ox" | Can run as an on-demand spot check or as continuous all-day/overnight tracking, depending on device settings, unlike Apple's on-demand-only model. |
| Oura Ring | "Blood Oxygen Sensing (SpO2)" | Overnight-only by design. Measures during sleep periods longer than three hours and reports Average Blood Oxygen plus a separate Breathing Regularity score. No daytime spot-check exists on Oura. |
| Fitbit | "Estimated Oxygen Variation" (EOV) | Deliberately not a raw SpO2 percentage. It's a relative overnight graph showing how much blood oxygen fluctuates, with a red/yellow/green pattern flagging possible breathing disturbances, a different approach from every other device listed here. |
Two different, real findings, not one: general accuracy against a reference oximeter holds up reasonably well across the studies reviewed here. But a separate, well-documented body of research finds pulse oximetry accuracy, including in consumer wearables, is reduced by skin pigmentation, and the direction of that bias is more nuanced than a simple light-skin/dark-skin split, one study found devices missed a dangerously low reading in 21% of darkly pigmented subjects under low-perfusion conditions, versus about 1% in lightly pigmented subjects.
General accuracy: Established · see the file Accuracy by skin tone: Uncertain · see the fileSources: Apple's own support documentation (support.apple.com) for the Blood Oxygen app. Garmin's own Pulse Ox support pages (garmin.com) for on-demand vs continuous modes. Oura's own Blood Oxygen Sensing blog post and support article (ouraring.com) for overnight-only measurement and the two reported metrics. Fitbit's own help documentation for Estimated Oxygen Variation's relative-graph design. Full study-by-study evidence lives on the Research Desk linked above.