Blood Oxygen

Blood oxygen, written SpO2, is the percentage of hemoglobin in your arterial blood that is carrying oxygen, estimated by a pulse oximeter from how your blood absorbs red and infrared light. Most healthy people at sea level read 95 to 100 percent, and readings run about a point lower during sleep.

For a healthy athlete at sea level the number barely moves. The useful information sits in the exceptions: a drop at altitude, a run of low nights, or a reading that disagrees with how you feel. Whether a 93 on your wrist deserves attention depends on what the sensor can resolve and where it goes wrong. The wearable data quality guide compares blood oxygen accuracy with other wearable metrics.

01What SpO2 measures

Arterial oxygen saturation, SaO2, is the share of hemoglobin binding sites carrying oxygen, measured directly from an arterial blood draw in a co-oximeter. SpO2 is the same quantity estimated through the skin by a pulse oximeter. The "p" stands for pulse. Clinicians treat SaO2 as the reference, and every accuracy figure below compares a device's SpO2 with it.

Saturation and oxygen pressure are different quantities. The relationship between them, the oxyhemoglobin dissociation curve, is S-shaped. Near the top it is almost flat, so large changes in arterial oxygen pressure move saturation very little. Jubran's 2015 review in Critical Care notes that because of this shape, oximetry may not detect a falling oxygen pressure when it starts from a high level. Lower down the curve steepens. Luks and Swenson's 2011 review of oximetry at high altitude warns that saturation there "changes rapidly in response to small changes in oxygen tensions."

Saturation also says nothing about how much hemoglobin you have. On the Caudwell Xtreme Everest expedition, Grocott and colleagues drew arterial blood from 10 climbers. Oxygen pressure fell with altitude, yet hemoglobin concentration rose enough to keep arterial oxygen content at or above sea-level values up to 7,100 m. At 8,400 m the mean arterial oxygen pressure was 24.6 mmHg (NEJM, 2009). SpO2 is one link in the oxygen cascade, and it tells you nothing about cardiac output, capillary supply or mitochondrial use, the links that set your VO2 max.

02How pulse oximetry works

Oxygenated and deoxygenated hemoglobin absorb light differently. Blood fully loaded with oxygen looks bright red and darkens toward brown as saturation falls, because deoxygenated hemoglobin absorbs more red light while oxygenated hemoglobin absorbs relatively more infrared. A pulse oximeter shines both wavelengths into tissue. Clinical finger probes typically use 660 nm red and 940 nm infrared (Jubran, 2015).

Most of the light the detector sees has passed through skin, bone, muscle and venous blood, none of which change with each heartbeat. Arterial blood swells slightly with every pulse. The oximeter isolates that pulsing part of the signal, called the photoplethysmogram or PPG, at each wavelength and computes the ratio of red to infrared modulation. A calibration curve, built from studies in which healthy volunteers breathe low-oxygen gas while their arterial blood is sampled, converts that ratio to a saturation percentage. Any bias in how the calibration group's skin and tissue handle light carries into every reading, which is where the skin tone problem described below enters.

A finger clip sends light through the finger to a detector on the other side. A watch cannot do that, so it uses reflectance, with emitters and detectors on the same surface reading light that scatters back out of the wrist. Apple's 2022 white paper states that pulse signals are smaller at the wrist than at the fingertip, which makes stillness and fit matter more.

Conventional oximeters can tell apart only two substances, oxyhemoglobin and reduced hemoglobin. They assume that dyshemoglobins such as carboxyhemoglobin from carbon monoxide are absent (Jubran, 2015), so a smoker's or a fire victim's SpO2 can look normal while oxygen delivery is impaired. Motion, cold hands, low perfusion and a loose sensor are the everyday sources of error.

03Normal ranges awake and asleep

Apple's support documentation states that "the majority of people have a blood oxygen level of 95 - 100%." A large dataset of wrist readings from daily life comes from the Apple Heart and Movement Study. Shapiro and colleagues analyzed 72.2 million readings from 33,080 participants wearing Apple Watch Series 6, taken on demand and in the background at roughly 30-minute intervals when the wearer was still (npj Digital Medicine, 2023).

The overnight mean was 0.8 percentage points lower than daytime in every demographic group. Differences between racial and ethnic groups were small, at most 0.13 points lower daytime SpO2 in Hispanic compared with White participants, with no significant differences at night. These are watch readings with no blood reference, so they cannot show whether the watch reads some groups less accurately than others.

For your own data, a nightly average of 94 or 95 is ordinary. About 2 percent of all readings in the study fell between 85 and 90, so a single background reading in the high 80s means little on its own. A pattern across nights means more.

04Altitude

Air at altitude still contains about 21 percent oxygen, but lower barometric pressure means each breath delivers fewer oxygen molecules. The effect on saturation is small at moderate elevation and large above about 2,500 m. Rojas-Camayo and colleagues measured 6,289 healthy, acclimatized residents aged 1 to 80 at 15 locations in Peru, from sea level to the highest permanent settlement, and found that SpO2 fell with altitude, "especially at altitudes above 2500 m" (Thorax, 2018).

Visitors fare worse than residents. Tannheimer and colleagues recorded continuous overnight SpO2 in 10 male mountaineers over a three-week Peruvian expedition (Journal of Travel Medicine, 2017). On the first night at 3,050 m, the average was 83.4 percent in the first half of sleep and 86.3 percent in the second. At the same camp on night 21 the figures were 88.3 and 90.1 percent. Nighttime values were lower than resting morning values, and the rise across each night shrank as acclimatization improved. The authors suggest that the size of that overnight rise could serve as a marker of individual acclimatization.

Whether falling SpO2 predicts acute mountain sickness is less settled. A 2024 systematic review by Goves and colleagues in Experimental Physiology found seven studies of 1,406 lowlanders ascending to 3,952 to 6,300 m. Falling saturation during ascent showed a positive predictive relationship with mountain sickness, but ascent profiles and measurement methods varied too much to pool the data. Luks and Swenson recommend treating the normal saturation at any elevation as a range, and not basing decisions on small differences over time or between people.

Take a ski trip as an example. The Heart and Movement coefficients predict about 2.8 points lower nighttime SpO2 for someone living at 2,400 m than for someone at sea level. That coefficient comes from residents, most of whom live near sea level, so it is a rough guide. A visitor on the first nights will usually sit below it. Apple labels readings taken at barometric pressures typical of elevations above roughly 5,000 feet as "high elevation environment" in the Health app, which helps you separate those nights from home nights.

05How Apple Watch measures blood oxygen

Blood oxygen is available on Apple Watch Series 6 and later and on every Ultra model. SE models do not measure it. The sensor uses four clusters of green, red and infrared LEDs and four photodiodes on the back crystal. According to Apple's 2022 white paper, the red and infrared LEDs run at about 660 nm and 850 nm, with green at 525 nm. The Blood Oxygen app reports values from 70 to 100 percent.

The watch takes readings two ways. An on-demand reading takes 15 seconds with your arm held still. Background readings happen opportunistically when the watch detects that you are still with your wrist in position, including during sleep if Track Sleep with Apple Watch is on. Apple's white paper says that these intermittent readings "are not the same as the continuous second-by-second measurement capability commonly available with bedside pulse oximeters." The app withholds a reading when the PPG signal is inadequate or your arm position risks venous pulsation, which can falsely lower the value.

Fit matters more than most people expect. Apple recommends a snug band with the sensor centered on the wrist and clear of the wrist bone. In Apple's own validation data, three sessions in which the watch was worn looser than recommended produced much of the outlier scatter, and removing them improved accuracy from 1.97 to 1.67 percentage points. Apple describes the measurements as designed for general fitness and wellness purposes and not intended for medical use.

The US feature after the Masimo dispute

Masimo, a maker of hospital pulse oximeters, won an International Trade Commission exclusion order against Apple Watch in 2023 over its patents. From January 2024, Apple sold Series 9 and Ultra 2 watches in the US with blood oxygen disabled. On August 14, 2025, Apple shipped a redesigned version in iOS 18.6.1 and watchOS 11.6.1 for US Series 9, Series 10 and Ultra 2 owners, following a US Customs ruling. In the redesign, the watch collects the sensor data and the paired iPhone does the calculation. Results appear in the Respiratory section of the Health app. Watches that already had the original feature, and watches bought outside the US, were unaffected.

The redesign has held up. The ITC opened a new proceeding over it in November 2025, and on March 19, 2026 an administrative law judge found that it does not infringe Masimo's patents. The Federal Circuit separately affirmed the original 2023 exclusion order, so the original on-watch version stays barred. On April 17, 2026 the full Commission declined to review the judge's finding and closed the case, though Masimo can still appeal (9to5Mac, 2026). A separate $634 million jury verdict for Masimo, over heart rate monitoring and notification features, survived Apple's challenge on July 21, 2026. Apple plans to appeal, and that case does not affect blood oxygen.

As of Apple's support page published September 14, 2026, watches bought in the US on or after January 18, 2024 with part numbers ending in LW/A or LS/A analyze blood oxygen on the iPhone, and the results cannot be viewed on the watch. The analysis is unavailable if you use a second iPhone that is not paired to the watch. The readings still land in Apple Health, which is where Titan reads them.

06How accurate wrist SpO2 is

Oximeter accuracy is reported as Arms, the root mean square of the differences between SpO2 and SaO2 across a population. About two thirds of readings fall within plus or minus Arms of the true value. The ISO standard allows up to 4 percentage points. FDA guidance for medical oximeters allows 3.0 or 3.5 points depending on sensor type.

Apple's validation

Apple's 2022 white paper reports a study in 50 healthy adults aged 19 to 40, each fitted with a radial artery catheter while inspired oxygen was lowered in steps to push saturation from about 100 percent down to about 70 percent. Watch signals were processed offline with the Blood Oxygen algorithm. The overall Arms was 1.97 points, with a mean bias of 0.21 points and 95 percent limits of agreement of -3.61 to +4.02. The eight participants with Fitzpatrick skin types V and VI had an Arms of 1.73, compared with 2.02 in the 42 lighter-skinned participants. The study ran under ideal laboratory conditions, and Apple conducted the analysis itself.

Independent studies

Independent results put the average error in a similar place and show more spread in individual readings.

Jiang and colleagues tested four watches against the same reference. Apple Watch Series 7 had the lowest error, and the Garmin Venu 2s the highest at 5.8 points. The Withings ScanWatch failed on 31 percent of attempts. In the COVID-19 study, the watch's specificity was 97.5 percent. When it reported low oxygen it was usually right, but it missed about two of every three low readings the ward oximeter caught.

The average bias is close to zero, and single readings are loose. A limit of agreement near ±4 points means a true saturation of 96 percent can appear anywhere from about 92 to 100. That spread is why a daily or nightly average says more than any one reading.

Skin tone bias

In 2020, Sjoding and colleagues compared paired pulse oximeter and arterial blood gas readings in the New England Journal of Medicine. In 10,789 pairs from 1,333 White and 276 Black patients at the University of Michigan, readings of 92 to 96 percent hid an arterial saturation below 88 percent in 11.7 percent of Black patients' measurements and 3.6 percent of White patients' measurements. In 37,308 pairs from intensive care units at 178 hospitals, the rates were 17.0 and 6.2 percent. The authors concluded that "Black patients had nearly three times the frequency of occult hypoxemia that was not detected by pulse oximetry as White patients." Those devices were hospital finger oximeters.

The follow-up work widened the finding and tied it to outcomes. Wong and colleagues analyzed 87,971 patients across 215 hospitals and found hidden hypoxemia in every group, at 6.8 percent of Black, 6.0 percent of Hispanic, 4.8 percent of Asian and 4.9 percent of White patients. Patients with hidden hypoxemia had more organ dysfunction a day later and higher in-hospital mortality (JAMA Network Open, 2021). In 1,216 COVID-19 patients at Johns Hopkins, Fawzy and colleagues found that SpO2 overestimated arterial saturation by 1.7 points in Asian, 1.2 points in Black and 1.1 points in Hispanic patients compared with White patients. Black patients had a 29 percent lower hazard of having their eligibility for COVID-19 treatment recognized (JAMA Internal Medicine, 2022).

A 2024 meta-analysis by Singh and colleagues in the Journal of Medical Internet Research pooled 23 studies with 197,353 paired readings. Oximeters overestimated saturation for both light and dark skin, with a mean bias of 0.70 points for light, 0.27 for medium and 1.27 for dark. None of the SpO2 studies used wrist devices. In January 2025 the FDA issued draft guidance asking makers of medical oximeters to test at least 150 participants spread across skin-tone groups, assessed with the 10-shade Monk Skin Tone scale and an objective measurement. The guidance does not cover wellness devices, which is how Apple classifies the watch feature. When Hughes and colleagues applied the proposed criteria to 34 oximeters in a 2025 preprint, 22 met the 2013 FDA standard and only one met the new criteria in cohorts of 24 adults. Most, 18 of 34, read higher in dark skin than in light skin, and some showed the opposite.

The mechanism is optical. Melanin absorbs strongly in the green and red parts of the spectrum. Apple says its watch compensates by raising LED current, photodiode gain and sampling rate when less light returns. Apple's own dark-skin sample was eight people, and Jiang's independent study found no significant skin-tone difference in average error but a possible relationship between Fitzpatrick type and the direction of error. The wrist evidence is thin in both directions.

The hospital data point to overestimation, so the failure that matters is a falsely reassuring reading when true saturation is low. Comparing your nights with your own median removes a steady offset, which helps with trends. It does nothing for the low readings, where the hospital studies found the problem. If you have darker skin, treat a wrist reading of 94 as weaker evidence that you are fine, and trust symptoms such as breathlessness over the number.

07What blood oxygen can and can't tell an athlete

At sea level, blood oxygen works as a check on the rest of your data. A night well below your usual range alongside higher resting heart rate, faster respiratory rate and raised wrist temperature is a reason to look closer. At altitude it becomes a direct acclimatization readout, provided you compare readings taken at the same time and in the same position, since Tannheimer's climbers gained up to 3 points across a single night.

It cannot tell you whether you recovered from yesterday's session. I found no study linking day-to-day wrist SpO2 at sea level to training readiness in healthy athletes, and the physiology gives little reason to expect one. Saturation sits near the flat top of the curve regardless of fatigue. HRV and resting heart rate carry that information.

It also cannot capture what happens during hard exercise. Apple's white paper describes maintained or rising SpO2 as the normal response to exercise in healthy people. Dempsey and Wagner's review in the Journal of Applied Physiology showed that a significant number of fit, healthy people desaturate at or near sea level during heavy work. They defined mild exercise-induced arterial hypoxemia as 93 to 95 percent, or 3 to 4 points below rest, moderate as 88 to 93 percent and severe as below 88 percent. They report that it lowers VO2 max. The watch withholds readings during movement, so it cannot see this.

It cannot screen for sleep apnea by itself. Background readings are intermittent, and repeated dips during apnea events can fall between samples. Apple's sleep apnea notifications use the accelerometer to detect breathing disturbances. The support documentation does not mention blood oxygen for that feature. A persistent run of low nights at sea level, or breathlessness at rest, calls for a clinician and a proper sleep study or medical oximeter.

08How Titan uses blood oxygen

Titan reads blood oxygen from Apple Health. Any source that writes it there counts, including Apple Watch and Whoop. The Oura direct connection does not write blood oxygen.

On Today. The Health card shows your most recent blood oxygen sample as a whole-number percentage, next to HRV, resting heart rate and respiratory rate. Tapping it opens Vitals.

On the Vitals screen. The SpO₂ tile shows your latest reading and whether it falls above, within or below your personal range for the selected period, the 20th to 80th percentile of your daily values. The blood oxygen section lists the latest day, a baseline and the change between them. Each day's value is the average of that day's readings, and the baseline is the median of those daily values across the 7D, 30D, 60D or 6M period. The chart differs from the other vitals. It draws a fixed green band labeled Normal 95–100% in place of your personal range, plots each day as a dot, and turns any day below 95 percent red. The vertical scale is stretched so that 95 to 100 fills the top 70 percent of the chart, 90 to 95 the next 20 percent and 80 to 90 the bottom 10 percent, which makes a dip to 93 visible. Selecting a dot shows its real value.

Adding a reading. From Vitals you can add a blood oxygen reading by hand, for example from a finger oximeter. Titan accepts 70 to 100 percent and writes the reading to Apple Health.

In sleep scoring. Blood oxygen does not change the headline Sleep score. Titan includes it in a separate Sleep Quality calculation with a weight of 0.06, the same as breathing rate. The blood oxygen input scores zero at 93 percent and full marks at 97 percent. When Titan also has your median daily value over the past 30 or more days, three quarters of the input comes from that level and one quarter from stability, which falls to zero once the night is 5 percent away from your median. The Sleep screen does not display Sleep Quality.

Blood oxygen plays no part in Recovery. Recovery uses your overnight HRV and sleeping heart rate against your own baselines.

09References

• Apple Inc. (2020). Apple Watch Series 6 delivers breakthrough wellness and fitness capabilities. Apple Newsroom. https://www.apple.com/newsroom/2020/09/apple-watch-series-6-delivers-breakthrough-wellness-and-fitness-capabilities/
• Apple Inc. (2022). Blood Oxygen app on Apple Watch. Apple white paper. https://www.apple.com/healthcare/docs/site/BloodOxygenapponAppleWatchOctober_2022.pdf
• Apple Inc. (2025). An update on Blood Oxygen for Apple Watch in the U.S. Apple Newsroom. https://www.apple.com/newsroom/2025/08/an-update-on-blood-oxygen-for-apple-watch-in-the-us/
• Apple Inc. (2026). How to use the Blood Oxygen app on Apple Watch. Apple Support. https://support.apple.com/en-us/120358
• Apple Inc. (2026). Measure blood oxygen levels with the Blood Oxygen app. Apple Watch User Guide. https://support.apple.com/guide/watch/blood-oxygen-apdaf17aa5ef/watchos
• Apple Inc. (2026). Sleep apnea notifications on your Apple Watch. Apple Support. https://support.apple.com/en-us/120031
• Dempsey JA, Wagner PD (1999). Exercise-induced arterial hypoxemia. Journal of Applied Physiology 87(6):1997-2006. https://doi.org/10.1152/jappl.1999.87.6.1997
• Fawzy A et al. (2022). Racial and ethnic discrepancy in pulse oximetry and delayed identification of treatment eligibility among patients with COVID-19. JAMA Internal Medicine 182(7):730-738. https://doi.org/10.1001/jamainternmed.2022.1906
• Goves JSL et al. (2024). Pulse oximetry for the prediction of acute mountain sickness, a systematic review. Experimental Physiology 109(12):2057-2072. https://doi.org/10.1113/EP091875
• Grocott MPW et al. (2009). Arterial blood gases and oxygen content in climbers on Mount Everest. New England Journal of Medicine 360(2):140-149. https://doi.org/10.1056/NEJMoa0801581
• Hughes C et al. (2025). Pulse oximeter performance and skin pigment, comparison of 34 oximeters using current and emerging regulatory frameworks. medRxiv preprint. https://doi.org/10.1101/2025.08.11.25332026
• Jiang Y et al. (2023). Investigating the accuracy of blood oxygen saturation measurements in common consumer smartwatches. PLOS Digital Health 2(7):e0000296. https://doi.org/10.1371/journal.pdig.0000296
• Jubran A (2015). Pulse oximetry. Critical Care 19:272. https://doi.org/10.1186/s13054-015-0984-8
• Lambe R et al. (2026). The accuracy of Apple Watch measurements, a living systematic review and meta-analysis. npj Digital Medicine 9:63. https://doi.org/10.1038/s41746-025-02238-1
• Luks AM, Swenson ER (2011). Pulse oximetry at high altitude. High Altitude Medicine and Biology 12(2):109-119. https://doi.org/10.1089/ham.2011.0013
• MedTech Dive (2025, January 6). FDA issues much-anticipated guidance on pulse oximeters. https://www.medtechdive.com/news/fda-draft-guidance-pulse-oximeter-accuracy/736555/
• Rajakariar K et al. (2024). Accuracy of smartwatch pulse oximetry measurements in hospitalized patients with coronavirus disease 2019. Mayo Clinic Proceedings: Digital Health 2(1):152-158. https://doi.org/10.1016/j.mcpdig.2024.02.001
• Rojas-Camayo J et al. (2018). Reference values for oxygen saturation from sea level to the highest human habitation in the Andes in acclimatised persons. Thorax 73(8):776-778. https://doi.org/10.1136/thoraxjnl-2017-210598
• Shapiro I et al. (2023). Pulse oximetry values from 33,080 participants in the Apple Heart & Movement Study. npj Digital Medicine 6:134. https://doi.org/10.1038/s41746-023-00851-6
• Singh S et al. (2024). Impact of skin pigmentation on pulse oximetry blood oxygenation and wearable pulse rate accuracy, systematic review and meta-analysis. Journal of Medical Internet Research 26:e62769. https://doi.org/10.2196/62769
• Sjoding MW et al. (2020). Racial bias in pulse oximetry measurement. New England Journal of Medicine 383(25):2477-2478. https://doi.org/10.1056/NEJMc2029240
• Tannheimer M et al. (2017). Oxygen saturation increases over the course of the night in mountaineers at high altitude (3050-6354 m). Journal of Travel Medicine 24(5). https://doi.org/10.1093/jtm/tax041
• 9to5Mac (2026, March 19). ITC judge says Apple Watch's redesigned blood oxygen feature doesn't infringe Masimo patents. https://9to5mac.com/2026/03/19/itc-judge-says-apple-watchs-redesigned-blood-oxygen-feature-doesnt-infringe-masimo-patents/
• 9to5Mac (2026, April 17). Apple wins latest round in Masimo fight as ITC closes Apple Watch import ban case. https://9to5mac.com/2026/04/17/apple-wins-latest-round-in-masimo-fight-as-itc-closes-apple-watch-import-ban-case/
• 9to5Mac (2026, July 21). Apple loses bid to overturn $634 million Masimo patent verdict. https://9to5mac.com/2026/07/21/apple-loses-bid-to-overturn-634-million-masimo-patent-verdict-new-trial-also-denied/
• Windisch P et al. (2023). Accuracy of the Apple Watch oxygen saturation measurement in adults, a systematic review. Cureus 15(2):e35355. https://doi.org/10.7759/cureus.35355
• Wong AI et al. (2021). Analysis of discrepancies between pulse oximetry and arterial oxygen saturation measurements by race and ethnicity and association with organ dysfunction and mortality. JAMA Network Open 4(11):e2131674. https://doi.org/10.1001/jamanetworkopen.2021.31674