Disclosure: This post contains affiliate links. If you click through and make a purchase, we may earn a small commission at no extra cost to you. Thank you for supporting this site!
You’ve probably seen the marketing: “Monitor your blood oxygen saturation anytime, anywhere.” But when I strapped six smartwatches to my arm and compared their SpO₂ readings against a Masimo Rad-7 medical pulse oximeter, the results ranged from surprisingly accurate to dangerously misleading. The average absolute error across all devices was 2.1%, but individual errors hit 6% during exercise and sleep. That’s the difference between a “normal” 96% and a concerning 90%. In this test, I’ll show you exactly which sensors you can trust, which hardware actually drives accuracy, and where wearable SpO₂ still belongs in the “nice to know” bucket rather than the “clinical decision” one.
Every smartwatch SpO₂ sensor uses photoplethysmography (PPG) — shining red (660 nm) and infrared (940 nm) LEDs through your skin and measuring light absorption. Oxygenated hemoglobin absorbs more infrared, deoxygenated more red. The ratio gives SpO₂. But here’s the catch: medical-grade pulse oximeters use transmission PPG (through a fingertip), while watches use reflectance PPG (from the wrist), which is far more susceptible to motion artifacts, skin pigmentation, and poor contact. The sensor hardware matters enormously. The Apple Watch Ultra 2 uses a custom Apple-designed optical sensor with four photodiodes and a TI AFE4900 analog front-end, while the Garmin Fenix 7X relies on the Elevate V4 sensor with a Bosch BHI260AP accelerometer for motion cancellation. The Withings ScanWatch uses a Philips-developed sensor, and the Oura Ring 4 uses a Samsung BioSensor. In my test, the Masimo Rad-7 (FDA-cleared, ±2% accuracy) served as the gold standard. Key difference: medical oximeters average over 10–15 seconds; watches often average over 30–60 seconds, smoothing out real dips. That’s not a bug — it’s a design choice to reduce noise, but it also masks transient desaturations that might be clinically relevant.
I tested six wearable devices over two weeks: Apple Watch Ultra 2 (v10.2 watchOS), Garmin Fenix 7X (v19.20), Fitbit Sense 2 (v1.6.2), Samsung Galaxy Watch 6 (One UI 5), Withings ScanWatch (v1.4.0), and Oura Ring 4 (v4.0.2). Each was worn simultaneously on the same arm (non-dominant) with the Masimo Rad-7 on the index finger. I collected 150 paired readings per device across three conditions: resting (sitting, 5 minutes), exercise (treadmill at 5 mph, 10 minutes), and sleep (overnight, 6 hours). The Masimo logged SpO₂ every 10 seconds; the watches logged their own automatic readings. I matched timestamps within ±30 seconds. To control for skin tone, I tested on fair and medium skin (Fitzpatrick II and IV). The results:
Key takeaway: resting accuracy is decent across the board, but exercise and sleep introduce significant error — especially for Fitbit and Oura.
The Apple Watch Ultra 2’s sensor array uses four photodiodes arranged in a cross pattern, combined with the TI AFE4900 — a dedicated analog front-end that separates the PPG signal from ambient light and motion. The AFE4900 includes programmable gain and a 22-bit ADC, allowing it to resolve small changes in blood volume. Garmin’s Elevate V4 uses a similar approach but with a single photodiode and the Bosch BHI260AP coprocessor for motion artifact reduction. In practice, the BHI260AP does a good job canceling walking motion but struggles with the erratic arm movements of running. That’s why Garmin’s exercise error was higher than Apple’s. Withings uses a Philips OEM sensor with two photodiodes and a dedicated ASIC — the same core hardware found in some hospital spot-check monitors. It’s no surprise it matched the Apple in resting accuracy. The Oura Ring 4 uses Samsung’s BioSensor, which is designed for finger-based PPG but suffers from ring fit variability — if the ring rotates, the optical path changes. Fitbit’s sensor is the oldest design here, using a single LED-photodiode pair and no dedicated motion coprocessor. In my sleep tests, Fitbit consistently reported SpO₂ values 3–5% lower than Masimo during REM sleep, likely because of poor contact during side sleeping.
Sleep SpO₂ is where most users want clinical insight — for sleep apnea screening. I compared overnight averages from each watch against the Masimo Rad-7 worn on the finger. The Apple Watch Ultra 2 averaged 96.2% vs Masimo’s 96.8% (difference 0.6%). Garmin averaged 95.3% (difference 1.5%). Fitbit averaged 93.8% (difference 3.0%). The Withings ScanWatch averaged 96.5% (difference 0.3%). But the real story is in the dips: the Masimo recorded 14 desaturation events (SpO₂ below 90%) during the night; the Apple Watch caught 8, Garmin caught 6, Withings caught 9, and Fitbit caught only 3. This is a critical limitation — watches miss transient desaturations because they sample less frequently. For altitude testing, I took the devices to 10,000 ft (simulated in a hypoxic chamber). At true SpO₂ of 88% (Masimo), the Apple Watch read 86%, Garmin 85%, Withings 87%, and Fitbit 82%. Skin tone effects: on medium skin (Fitzpatrick IV), the Apple Watch error increased by 0.5% on average, Garmin by 0.8%, and Fitbit by 1.9%. The Withings ScanWatch was least affected (0.3% increase). These results align with published studies — for example, a 2023 study in JAMA Internal Medicine found that smartwatch SpO₂ accuracy degrades on darker skin, with some devices exceeding 3% bias.
SpO₂ monitoring is power-hungry — the LEDs draw up to 50 mA during readings. I measured battery drain in two scenarios: GPS-on continuous tracking (1 hour outdoor run with SpO₂ logging every 5 minutes) and daily use (24 hours with automatic sleep SpO₂ and occasional manual checks). Results:
Garmin’s dominance in battery life is due to its larger battery (500 mAh vs Apple’s 542 mAh but with lower-power GPS chipset) and less frequent SpO₂ sampling. But that lower sampling rate is exactly why Garmin misses more desaturations. The Withings ScanWatch uses a tiny battery (45 mAh) but lasts days because it only takes SpO₂ readings on demand — no automatic night tracking. That’s a conscious trade-off: accuracy at rest, but no continuous data.
Every brand claims “medical-grade accuracy” — but only the Withings ScanWatch has FDA clearance for SpO₂ (as a spot-check device, not continuous monitoring). Apple’s SpO₂ feature is not cleared; it’s classified as a “wellness” feature. Garmin, Fitbit, and Samsung similarly avoid FDA clearance. What does that mean in practice? If your watch says 92% while resting, you should not assume you have hypoxemia. Instead, check with a medical pulse oximeter. The clinically useful data from these watches is trends, not absolute numbers. For example, a consistent drop in overnight SpO₂ over weeks could indicate worsening sleep apnea — but only if the watch is accurate enough to detect the trend. In my data, the Apple Watch and Withings showed a correlation of r=0.89 with Masimo trends over 7 nights; Fitbit showed r=0.71. That’s still useful for spotting changes, but not for diagnosis. Another marketing claim: “SpO₂ during exercise helps optimize training.” In reality, during high-intensity exercise, SpO₂ stays near 95-98% in healthy individuals; only elite athletes at VO2max see significant drops. For most users, the data is noise. The one genuinely useful application is altitude acclimatization — tracking SpO₂ over days at high altitude can indicate when to descend. But even then, rely on the more accurate devices (Apple, Withings).
Based on accuracy, battery, and real-world performance, here’s my advice:
No smartwatch replaces a medical pulse oximeter for diagnosis. But if you need trend data for personal health awareness, prioritize devices with dedicated PPG front-ends (TI AFE4900 or similar) and frequent sampling rates.
First, don’t trust a single SpO₂ reading from any smartwatch — always take multiple readings at rest and compare over time. Second, Apple Watch Ultra 2 and Withings ScanWatch are the most accurate in this test, with MAE under 2% across all conditions. Third, use SpO₂ trends, not absolute values, for health insights — a downward trend over weeks is more meaningful than a one-time 92% reading. If you need to track sleep apnea or altitude adaptation, buy the Apple Watch Ultra 2 for continuous monitoring or the Withings ScanWatch if you prefer a watch that looks like a traditional timepiece. For everything else, save your money and stick with a $30 finger pulse oximeter.
Not reliably enough for diagnosis. While some watches (Apple, Withings) can spot overnight desaturation patterns, they miss 30–50% of events compared to polysomnography. The Apple Watch Ultra 2 caught 8 of 14 desaturations in my test; that’s a 43% miss rate. If you suspect sleep apnea, you need a home sleep test or in-lab polysomnography. However, consistent low overnight SpO₂ trends (e.g., average below 94%) warrant a conversation with your doctor.
You can’t calibrate consumer smartwatches — the sensor firmware is fixed. But you can validate accuracy: take a simultaneous reading with a medical pulse oximeter (like the Masimo Rad-7 or a cheap CMS-50D) at rest. If the difference exceeds 3%, the watch’s SpO₂ feature is likely compromised by poor fit, skin tone, or motion. For best results, clean the sensor window regularly and ensure the watch is snug but not tight.
Two reasons: motion artifacts and peripheral vasoconstriction. During exercise, blood flow diverts from the skin to muscles, reducing the PPG signal. The watch’s algorithm may misinterpret this as lower SpO₂. In my test, exercise errors were 1.5–3% higher than resting errors across all devices. This is why many watches disable continuous SpO₂ during GPS workouts — the data is too noisy to be useful. If you see a reading below 90% during exercise, stop and check with a finger oximeter.