How Accurate Is the Apple Watch's Blood Oxygen Reading?

Close at ordinary oxygen levels, unreliable once oxygen drops, and the gap between those two situations is large enough to matter. In 167 patients hospitalized with a COPD flare, Apple Watch 6 was off from arterial blood by a mean of 0.458 percent, a smaller mean error than the medical-grade pulse oximeter measuring alongside it. Under induced hypoxemia in healthy adults, Apple Watch 8 met the 4 percent accuracy standard used for medical oximeters and had the lowest error of three smartwatches tested. But in 9 healthy adults whose oxygen was lowered step by step, Apple Watch Series 7 fell within 2 percent of arterial blood only 32.14 percent of the time, against 49.03 percent for a clinical oximeter, and below 88 percent both devices exceeded the FDA accuracy threshold, the watch by more. And in 200 patients hospitalized with COVID-19, Apple Watch Series 7 flagged only 34.8 percent of the clinically significant low-oxygen readings a ward oximeter caught.

The question behind the question

Someone glances at their wrist after a workout, or in the middle of the night, and sees a blood oxygen number that seems slightly off from what they expected, or from what a fingertip clip reading showed the last time they were at a clinic. The number on the watch face looks precise, formatted just like a medical device would format it. But precise-looking isn't the same as validated, and the gap between those two things is exactly what a handful of researchers have gone and tested directly.

The tension isn't whether the Apple Watch produces a number. It clearly does. The open questions are how that number compares with arterial blood gas analysis, the reference doctors rely on, and whether it holds up below the high 90s where most healthy readings sit.

3 studies

• Against arterial blood gas under controlled hypoxia induction, Apple Watch Series 7 readings fell within the accepted 2 percent error range 32.14 percent of the time, against 49.03 percent for a clinical-grade Masimo finger oximeter. Below 88 percent saturation the watch recorded an accuracy root mean square of 5.82 percent (95% CI 5.32 to 6.31) against the oximeter 3.52 percent (95% CI 3.18 to 3.86), and both exceeded the US Food and Drug Administration threshold at that level. The authors note the watch mean bias suggests a likelihood of missing hypoxemia just below 88 percent, and that the small sample and limited skin tone diversity restrict how far this generalizes.Instrument validation study · Jiang et al., JMIR Formative Research, 2026
• In 167 patients hospitalized with an acute COPD exacerbation, Apple Watch 6 correlated strongly with medical-grade pulse oximetry (r = 0.89, ICC = 0.940). Bland-Altman analysis put its mean error against arterial blood gas at 0.458 percent, limits of agreement minus 5.912 to 4.996. That is a smaller mean error than the pulse oximeter own 1.015 percent against the same reference, though both carry wide individual spread. Patients breathing fast needed repeated attempts to get a reading at all.Prospective cross-sectional study · Arslan et al., Irish Journal of Medical Science, 2024
• Eighteen healthy adults were each assessed three times wearing Apple Watch 8, Samsung Galaxy Watch 5 or Withings ScanWatch against a Masimo Radical-7, during hypoxemia induced by gas mixtures at 14, 12 and 10 percent oxygen. All three smartwatches met the maximum acceptable root-mean-square deviation of 4 percent, both at normal oxygen levels and during desaturation below 90 percent. Apple Watch 8 reached the highest reliability of the three, with the lowest mean bias and root-mean-square deviation and the highest correlation with the reference.Comparative validation study · Walzel et al., Sensors, 2023

Claim rating: Uncertain · on the Research Desk

Apple Watch Blood Oxygen Accuracy by Model: Which Have Been Tested?

Only the Series 6, Series 7 and Series 8 have published independent tests of their blood oxygen readings. A search of Europe PMC in October 2026 found no independent study testing the Series 9, Series 10, Series 11, Series 12 or any Ultra model against arterial blood or a clinical oximeter. That is a statement about what has been studied, not a finding that the newer models are better or worse.

Apple itself describes the Blood Oxygen app as for wellness purposes only and not for medical use, and its support page says that even under ideal conditions the watch may not get a reliable measurement every time, naming a loose fit, some tattoos, cold skin and motion as things that interfere. The app runs on Series 6 and later and every Ultra; no Apple Watch SE has it.

The Series 12 and Ultra 4 matter most here because Apple redesigned their sensing system, describing more capable optical and electrical sensors and larger green LEDs, and its accuracy claim for that redesign covers heart rate. Apple makes no blood oxygen accuracy claim for either model and does not say whether the blood oxygen measurement changed. A redesigned sensor is a reason not to carry older results across automatically, not evidence that accuracy went up or down.

Where the number is calculated changed too. Apple disabled Blood Oxygen on US-sold Series 9, Series 10 and Ultra 2 units in January 2024 during a patent dispute with Masimo, then restored it in the United States in August 2025 in a redesigned form: the watch captures the signal and the paired iPhone calculates the result, which appears in the Health app. Apple says this applies to watches bought in the US on or after January 18, 2024, and its Series 12 and Ultra 4 pages state that measurements are calculated and viewed on iPhone. Every published study ran on the original on-watch version. The US legal position is still contested, with a further ITC investigation opened in late 2025.

Does My Apple Watch Have Blood Oxygen, and Where?

Apple lists the Blood Oxygen app on Series 6 and later and on every Ultra, which includes the Series 11, Series 12, Ultra 3 and Ultra 4. No Apple Watch SE has it. The table above shows which of those models have a published accuracy test. Most do not.

It is back in the United States. For US units bought on or after January 18, 2024 with part numbers ending in LW/A or LS/A, Apple says the analysis runs on the paired iPhone and the result appears in the Health app. Outside the US, the app is on by default only in certain countries and regions; Apple's availability list includes Canada.

Apple names a loose fit, the watch sitting on the wrist bone, an arm hanging down or a clenched hand, movement, low blood flow to the skin such as in the cold, some tattoos, and a resting heart rate above 150 beats per minute, and says that even under ideal conditions the watch may not get a reliable reading every time.

Overnight readings come from background measurements, which only run during sleep when the Track Sleep with Apple Watch setting is on. A low overnight number can reflect the sensor or a real dip, and the two are hard to tell apart from the watch alone: what a reading near 90 while asleep can and cannot tell you, and why oxygen readings drop overnight.

Where the numbers actually diverge

The clearest pattern across this small body of research is that accuracy isn't a single fixed number, it shifts depending on how low oxygen saturation actually goes. The controlled hypoxia study found that both a clinical pulse oximeter and the Apple Watch overestimated saturation, but the clinical device was still more likely to land inside the accepted error range than the watch was. That's a meaningful distinction: it's not that the watch is randomly noisy, it's that its error runs in a consistent direction, toward reading higher than the true value, right when accuracy matters most.

The COPD-flare study found something a little more reassuring for everyday, less extreme readings: a strong statistical correlation between the watch and medical-grade pulse oximetry, and a small average error against arterial blood gas. But that same study reported a wide spread around that average, meaning the watch could be very close for one person and considerably further off for another, even though the overall correlation looked strong.

Neither of these findings contradicts the other. They're describing different parts of the same picture, one from a lab pushing oxygen levels down deliberately, one from a hospital population with real respiratory illness. For a broader look at how the underlying measurement is even taken from the wrist in the first place, how wearables actually measure blood oxygen is a useful companion piece.

How Apple Watch blood oxygen compares with medical pulse oximeters

Medical pulse oximeters are held to an international standard, ISO 80601-2-61, which allows a root-mean-square deviation of no more than 4 percent from a reference measurement. Tested against that standard during deliberately induced low oxygen, in 18 healthy adults breathing gas mixtures of 14, 12 and 10 percent oxygen, the Apple Watch 8, Samsung Galaxy Watch 5 and Withings ScanWatch all stayed within the 4 percent limit against a Masimo Radical-7 clinical oximeter, both at normal oxygen levels and with saturation below 90 percent. The Apple Watch 8 had the lowest mean bias and deviation and the highest correlation of the three (Sensors, 2023).

In a hospital population the comparison held at ordinary levels. Across 167 patients admitted with a COPD flare, the Apple Watch 6 correlated strongly with a medical-grade pulse oximeter (r = 0.89) and differed from arterial blood by a mean of 0.458 percent (Irish Journal of Medical Science, 2024). Where the comparison breaks down is deep desaturation. Under controlled hypoxia, both a Series 7 and a Masimo fingertip oximeter overestimated saturation against arterial blood, and the clinical device was the more likely of the two to land within the accepted range (JMIR Formative Research, 2026).

The largest hospital study of the watch cited here points the same way on the reading that matters most. In 200 patients hospitalized with COVID-19, Apple Watch Series 7 flagged only 34.8 percent of the clinically significant hypoxia that simultaneous ward oximetry detected, while correctly clearing 97.5 percent of readings without it, and it correlated only moderately with the ward oximeter (r = 0.61). The authors conclude its accuracy may not be sufficient to replace standard oximetry for detecting hypoxia (Mayo Clinic Proceedings: Digital Health, 2024). A Withings ScanWatch tested alongside it caught more of the low readings, 68.5 percent, at the cost of more false alarms.

So on the tested models, the watch performs like a medical oximeter at normal and moderately reduced oxygen, and worse than one when oxygen falls far, which is the situation where the reading matters most. The reference device in two of these studies was made by Masimo, the same company whose patent dispute removed the feature from US watches in 2024.

What every Apple Watch blood oxygen study adds up to

On average the watch lands almost exactly on the reference, and any single reading can still be a few points off. A living systematic review of 82 Apple Watch validation studies across 14 health metrics pooled nine blood oxygen studies with 969 participants and found a mean difference of 0.04 percentage points below the reference (95 percent CI 0.42 below to 0.35 above), with limits of agreement from 4.00 points below to 3.94 points above (npj Digital Medicine, 2026).

In practice, about 95 in 100 readings fall within roughly 4 points of the reference, so a reading of 95 percent could reflect anything from about 91 to 99. Single readings can land further out than that: an earlier review of five Series 6 studies covering 973 patients found 95 percent limits of agreement of 2.7 to 5.9 points either side, with individual outliers as far as 15 points from the reference (Cureus, 2023). In low-oxygen ranges the spread was far wider: an exploratory subgroup analysis found limits of agreement from 8.35 points below to 9.21 points above, and all five studies that tested both ranges found closer agreement at normal oxygen levels. The authors concluded the watch may in some cases classify someone with low oxygen as normal.

Every blood oxygen study in that review tested a Series 6, 7 or 8. The review was independent, funded by Science Foundation Ireland, with no competing interests declared. Those figures describe those models; they do not measure the Series 12 or Ultra 4.

Is Apple Watch blood oxygen reliable? Consistency is a separate question from accuracy

Accuracy asks how close a reading is to the true value. Reliability asks whether the device behaves consistently, and the evidence on that is thinner. What it does show is that the main error has a direction: under controlled hypoxia the watch overestimated saturation rather than scattering randomly around the true value, which is a systematic bias and, in that narrow sense, a predictable one (JMIR Formative Research, 2026). In the ISO-standard comparison of three smartwatches, the authors described the Apple Watch 8 as the most reliable of the three on bias, deviation and correlation (Sensors, 2023).

Two findings cut the other way. The COPD study found a wide spread around its small average error, a standard deviation of 2.78 percent, so the watch was close for some patients and noticeably further off for others. That study also excluded patients whose watch could not produce a reading within two minutes or eight attempts, which means the published accuracy figure applies only to the readings the watch managed to take (Irish Journal of Medical Science, 2024). None of the studies on this page measured the same person repeatedly over days, so how consistent a single watch is with itself has not been tested here.

Why the comparison device matters

Part of what makes this hard to summarize in one line is that even the reference devices these studies compare against aren't perfectly interchangeable. Research on pulse oximetry broadly has long noted that accuracy depends on the specific instrument, and that from one arterial blood sample to the next, saturation can't always be reliably predicted from a single oximeter reading, even in a controlled ICU setting. That's not a smartwatch-specific problem, it's a pulse oximetry problem, and it's part of why researchers keep re-testing new devices against blood gas rather than just against each other.

This is also why the same review of pulse oximetry principles points to a list of separate factors, motion, poor circulation, and certain blood conditions among them, that can throw off any optical oxygen sensor regardless of what it's built into. None of the Apple Watch-specific studies here isolated those individual factors, so it isn't possible to say from this evidence exactly which one explains the gap seen in the hypoxia study. For how these same accuracy questions play out specifically overnight, blood oxygen accuracy during sleep covers that angle in more depth, and the underlying distinction between the two measurements being compared is unpacked in the difference between SaO2 and SpO2.

2 studies

• In a 26-bed ICU, the average difference between pulse oximetry and arterial oxygen saturation was small, but sample-to-sample fluctuation meant a single blood gas reading couldn't reliably predict future oximeter accuracy, and accuracy varied by the specific oximeter type used.Observational prospective study · Van de Louw et al., Intensive Care Medicine, 2002
• A review of pulse oximetry's underlying principles outlines several separate factors, including motion artifact, poor circulation, and certain blood conditions, that limit accuracy across pulse oximeter devices generally.Review · Sinex, The American Journal of Emergency Medicine, 1999

Claim rating: Uncertain · on the Research Desk

Common questions

KnowYourPrime Evidence Registry: Finding KYP-ER-0001: Smartwatch Blood Oxygen (SpO2) Accuracy Against Arterial Blood (SaO2) · Finding KYP-ER-0002: Pulse Oximeter Accuracy and Skin Tone · Finding KYP-ER-0023: Apple Watch Blood Oxygen Accuracy Against Medical Devices · Finding KYP-ER-0017: Can Blood Oxygen Monitoring Rule Out Sleep Apnea? · Finding KYP-ER-0019: Blood Oxygen Readings With Poor Circulation