SpO2 on wearables, explained: what wrist oxygen readings actually mean

SpO2 is the number on your watch that you probably do not think about until it shows up red. The reading on a wrist wearable shares a name, a unit, and a general idea with the value a nurse reads off a fingertip pulse oximeter in a hospital, but it is not the same measurement, and the difference matters when you start making decisions about your own body.

We will cover what the percentage represents, how the sensor estimates it from light bouncing through your wrist, why those estimates drift, the Apple-Masimo patent fight that explains why US Apple Watches lost SpO2 in 2024, and what sleep apnea features can and cannot do.

What SpO2 is, in one paragraph

SpO2 stands for peripheral oxygen saturation. It is the estimated percentage of hemoglobin in your arterial blood that is currently carrying oxygen. The “peripheral” part is important: it is measured at the skin, not from arterial blood directly. The gold-standard measurement is SaO2, drawn from an arterial blood gas, which is invasive and only done in clinical settings. SpO2 is a non-invasive proxy. In a healthy adult at sea level it sits between 95% and 100%; below 90% is generally treated as hypoxemia and is a clinical concern.

The technique is called pulse oximetry. Takuo Aoyagi, an engineer at Nihon Kohden in Tokyo, worked out the basic principle in the early 1970s. He noticed that oxygenated and deoxygenated hemoglobin absorb red and infrared light differently, and that you could read the ratio through a fingertip if you separated the pulsatile arterial signal from the steady-state tissue background. Modern fingertip oximeters still use Aoyagi’s two-wavelength method. So do wrist wearables. The hardware just has a much harder job.

Why wrist SpO2 is the hard version of this problem

A fingertip oximeter clips onto a fingertip or an earlobe. Light passes through about a centimeter of tissue from one side to the other, the finger is well-perfused with arterial blood, and the sensor sits in a stable mechanical position. This is transmission oximetry. The signal-to-noise ratio is high enough that FDA-cleared accuracy for medical-grade devices is typically ±2% across the 70-100% range.

A wrist wearable cannot do transmission oximetry. The sensor on the back of a watch shines red and infrared light into the skin and reads the reflected signal that bounces back to a photodiode a few millimeters away. This is reflectance oximetry. The optical path goes through skin, subcutaneous fat, tendons, sometimes bone shadow, and back. Arterial perfusion at the wrist is poorer than at the finger. The sensor moves relative to the skin every time you bend your hand. The watchband is often loose. Ambient light leaks in.

The result is that wrist SpO2 is a noisier estimate of the same underlying value. Independent validation work published over the past few years has consistently found that wrist devices read within about 2-4 percentage points of a clinical finger oximeter on average, with the bias getting worse as true saturation drops below 90% and worse again under motion. A 2021 study by Pipek and colleagues in Scientific Reports tested several wrist devices against arterial blood gas and found bias and limits of agreement that would not pass FDA pulse-oximeter requirements. None of the major wrist wearables are sold as medical pulse oximeters; they are sold as wellness devices. The fuller story of how the green and red LEDs on the back of a watch translate into the numbers in the app is in our wearable sensor accuracy primer.

The skin-tone problem, and why it is worse on the wrist

In December 2020, Michael Sjoding and colleagues published a research letter in the New England Journal of Medicine showing that clinical fingertip pulse oximeters were almost three times as likely to miss occult hypoxemia in Black patients as in white patients across two hospital cohorts. The mechanism is that melanin in the skin absorbs some of the red and infrared light the oximeter relies on, biasing the saturation estimate upward in patients with darker skin. The bias was small in absolute terms but clinically important at the threshold where supplemental oxygen decisions get made.

That paper was about clinical fingertip oximeters, the gold standard of non-invasive monitoring. The same physics applies to wrist devices, with the optical path longer and the signal weaker. Follow-up studies since 2021 suggest the skin-tone bias on consumer wrist wearables is at least as large as the clinical-grade bias Sjoding described, and probably larger. The FDA convened an advisory committee on the issue in November 2022 and has been working on updated guidance ever since; none of it has been finalized.

The practical takeaway: if you have darker skin and your wrist wearable is reading 96% overnight, the true value could easily be a couple of points lower. Treat the number as a rough wellness signal and not as a clinical measurement, regardless of who is wearing it.

How wearables actually measure SpO2: on-demand vs continuous

There are two patterns in consumer wearables, and they have very different reliability characteristics.

On-demand spot checks. This is the Apple Watch model in regions where SpO2 is enabled. You open the Blood Oxygen app, rest your arm flat, hold still for 15 seconds, and the watch gives you a number. The 15-second still-arm requirement exists because the sensor is averaging a lot of noisy samples; motion blows up the math. In our hands these spot checks still fail to read about one time in five even when you sit perfectly still.

Continuous overnight tracking. This is what Fitbit, Garmin, Samsung, Whoop, and most others do by default. The device samples SpO2 in the background while you sleep, when motion is minimal and the wrist is most stable. Most apps report a graph of overnight oxygen variation rather than a single value. You cannot get a snapshot of your saturation right now; you get a trend across the night.

Ring wearables (Oura, the Samsung Galaxy Ring) sit in their own bucket. The finger has a shorter optical path and better perfusion than the wrist, so a ring is mechanically a better place to read SpO2 than a watch. Both major rings sample only at night, because the finger moves too much during the day for a clean signal. Our skin temperature wearables explainer covers the same form-factor advantages for a different sensor.

The Apple-Masimo dispute, and why your US Apple Watch lost SpO2

Apple added an SpO2 sensor to the Apple Watch Series 6 in September 2020. Masimo, the medical-device company that has dominated hospital pulse oximetry for two decades, sued Apple shortly after, alleging that Apple had hired away Masimo engineers and infringed several patents covering reflectance pulse oximetry. The dispute ran through three years of trials and parallel International Trade Commission proceedings.

In October 2023, the ITC ruled that Apple had infringed two Masimo patents and ordered an import ban on Apple Watch Series 9 and Ultra 2 units containing the disputed feature. The ban took effect December 26, 2023, was briefly paused on appeal, and took effect again on January 18, 2024. Apple’s response was to ship software-modified Apple Watches in the US with the Blood Oxygen feature disabled. The hardware sensor is still under the watch. The app icon is gone.

Apple Watches sold outside the US, and US units purchased before January 18, 2024, are unaffected. A used Series 9 with its original software will still measure SpO2; a brand-new Series 9 from the Apple Store today will not. As of mid-2026 the dispute is unresolved, with Apple pursuing appeals and reportedly working on a hardware-level workaround. Our Apple Watch vs Fitbit health tracking comparison covers where this leaves Apple Watch buyers.

Sleep apnea: trend monitoring vs FDA-cleared detection

Most of the marketing around overnight SpO2 on wearables points, eventually, at sleep apnea. The pitch is intuitive: obstructive sleep apnea causes repeated drops in blood oxygen overnight, your watch is on your wrist overnight, therefore your watch should be able to flag the condition.

It is not that simple. Clinical sleep medicine diagnoses obstructive sleep apnea using an Apnea-Hypopnea Index (AHI), the average number of breathing pauses per hour of sleep, measured during in-lab polysomnography or a prescribed home sleep apnea test. AHI requires a respiratory effort belt, an airflow sensor, and oxygen monitoring; you cannot derive it from wrist SpO2 alone. What most consumer wearables report is oxygen variation or breathing disturbance trends, patterns that may correlate with apnea but are not a diagnostic.

The current exception is the Samsung Galaxy Watch sleep apnea feature, which received FDA De Novo authorization in February 2024 (DEN230030). It is the first wrist wearable cleared for sleep apnea detection in the US. The feature requires two nights of monitoring across a 10-day window, looks for moderate-to-severe obstructive sleep apnea in adults 22 and older, and explicitly does not provide an AHI value or replace a sleep study. A positive result is a flag to get evaluated, not a diagnosis.

Apple announced a similar sleep apnea notification feature for the Apple Watch in late 2024, and Fitbit’s Sleep Profile includes a related “breathing disturbances” signal that has not received FDA clearance. Our sleep stages explainer covers the polysomnography vs wearable comparison in more depth. The short version: a wearable can sometimes notice a pattern that looks like apnea; only a sleep study can confirm one.

A real comparison: clinical, wrist, and ring

The differences are easier to see in a table than in prose.

The table makes a point the marketing copy will not: there is no consumer wearable currently sold as a pulse oximeter in the US. Every device on the right side is a wellness product that happens to estimate a number you also see in a hospital. That gap is intentional and regulatory, and it is the right way to think about every reading the device produces.

What to do with the number on your wrist

A few rules of thumb we use ourselves.

Single overnight average above 94%? Ignore it. That is well within the noise floor of the sensor and almost certainly fine.

Sustained drop in your personal baseline over weeks or months? Worth a conversation with a doctor, especially if it tracks with new snoring, daytime fatigue, or weight changes. The trend is more useful than any single number, which is why the better apps emphasize the chart and not the digit.

One-off reading below 90% with no symptoms? Probably a fit issue. Tighten the band, wait two minutes, re-read. If it stays low and you have shortness of breath, chest tightness, or blue extremities, the watch stops being relevant and you call someone.

Reading from a wrist or ring at altitude? Expect lower numbers. A reading of 88-92% in Denver is not the same thing as 88-92% in Boston. Acclimatized residents at altitude can comfortably read in the high 80s.

The honest framing of consumer SpO2 is that it is most useful as a personal trend monitor for a healthy adult who wants to notice when something changes. It is least useful in the moments people most want a diagnostic tool, when they are sick, anxious, or trying to decide whether to go to urgent care at 11 PM. The watch can tell you the number is lower than it was last week. It cannot tell you whether that matters. The ECG story on these same devices, covered in our ECG smartwatch explainer, has gone through the same arc: useful trend tool, occasional genuine alert, never a substitute for a 12-lead.