| Quick Facts | |
|---|---|
| PLOS Digital Health study sample | 49 participants, median age 64 |
| Apple Watch Series 7 mean absolute SpO2 error vs. clinical oximeter | 2.2 percentage points |
| Garmin Venu 2s mean absolute SpO2 error vs. clinical oximeter | 5.8 percentage points |
| Occult hypoxemia rate, White patients (pulse oximetry review) | roughly 3.6-4.9% |
| Occult hypoxemia rate, Black patients (pulse oximetry review) | roughly 6.2-6.9% |
| Blood pressure monitors sold globally never formally validated | about 80% (American Heart Association, Dec 2025) |
| FDA blood glucose wearable warning issued | February 21, 2024 |
Can you trust the SpO2 reading on your smart watch?
You can trust a smart watch SpO2 reading as a rough, in-the-ballpark trend indicator, but not as a number precise enough to base a medical decision on by itself.
Shopping for a specific model? See our full smart watches roundup — 10 models compared on the specs that decide it.
Can you trust the SpO2 reading a smart watch gives you the same way you'd trust a device in a hospital? No, and most manufacturers quietly agree, tucking a wellness-use-only disclaimer into the fine print of every health app. The more useful question is narrower: does the number move in the right direction, and is it close enough to flag a real problem?
On that narrower question, the answer leans toward yes, with real caveats attached. A watch that reads 98 percent most nights and suddenly drops to 91 percent is picking up on something worth a second look, even if 91 percent is not exactly what a hospital pulse oximeter would report at that same moment. The trend matters more than any single reading, and a night of scattered high-80s numbers is a better reason to buy a dedicated fingertip pulse oximeter than to panic over one data point.
Where trust breaks down is treating the watch as diagnostic. It was never built, tested, or cleared for that job, and the accuracy research below explains why the margin of error is too wide for anything beyond a nudge to check further.
How a smart watch actually measures blood oxygen
A smart watch shines light through the wrist and reads how much bounces back, a method called reflectance pulse oximetry that is inherently noisier than the fingertip transmittance method hospitals use.
Every pulse oximeter, wrist-worn or fingertip, works on the same basic principle: oxygenated and deoxygenated blood absorb red and infrared light differently, and measuring that difference produces an SpO2 estimate. Where the two device types part ways is geometry. A fingertip clip shines light through a thin, blood-rich part of the body and measures what comes out the other side, a method called transmittance. A smart watch cannot do that at the wrist, so it shines light into the skin and measures what reflects back toward the same sensor, a method called reflectance.
Reflectance has to work harder for a weaker signal. The wrist is mostly tendon and bone rather than the dense capillary bed found in a fingertip, so the ratio of useful blood-flow signal to background noise from skin, motion, and ambient light starts out worse. Wrist temperature, a loose strap, tattoo ink, or even how tightly you're gripping something can shift a reading enough to matter. None of that makes the sensor useless. It just hands a smart watch a harder physics problem than a fingertip clip ever has to solve.
What the accuracy research actually found
A 2023 study testing four consumer smartwatches against a clinical pulse oximeter found average errors between roughly two and six percentage points, with too few low-oxygen readings in the sample to confirm how any of them perform during genuine hypoxemia.
Can you trust a smart watch SpO2 reading to match a hospital device number for number? The most direct evidence comes from a 2023 study published in PLOS Digital Health. Researchers tested four consumer devices, Apple Watch Series 7, Garmin Venu 2s, Garmin Fenix 6 Pro, and Withings ScanWatch, against a Masimo MightySat Rx, a clinical-grade reference oximeter, across 49 participants with a median age of 64. This was not a test run by a device maker's marketing team; it is an independent, peer-reviewed comparison.
The spread between devices was wide. Apple Watch Series 7 came closest to the reference device, with a mean absolute error of 2.2 percentage points. Garmin Venu 2s sat furthest away, at 5.8 percentage points, with Garmin Fenix 6 Pro and Withings ScanWatch landing in between at 3.8 and 4.0 points. That is a real spread for a health metric where clinicians typically care about single-digit percentage changes.
One limitation in that study needs stating plainly: only three participants had true SpO2 readings below 90 percent, the range where accuracy matters most for catching a genuine problem. The researchers said so themselves, and it means this study, solid as it is, cannot show how any of these watches perform during real hypoxemia. That gap is exactly why a watch reading in the low 90s deserves a second check with a real pulse oximeter rather than a decision made on the spot.
Skin tone changes the accuracy picture, and not just for smart watches
Research on pulse oximetry, including a 2023 review in Physiological Measurement, has repeatedly found that pulse oximeters overestimate oxygen levels more often in people with darker skin, a bias that predates wearables and likely carries over to wrist sensors.
Skin pigmentation absorbs light too, and that can distort the same red-and-infrared measurement a pulse oximeter depends on, clinical or consumer. A 2023 review in the journal Physiological Measurement, led by researchers at City, University of London, pulled together the existing evidence and found a consistent pattern: occult hypoxemia, a falsely normal reading despite genuinely low blood oxygen, showed up in roughly 3.6 to 4.9 percent of White patients across the studies reviewed, compared with 6.2 to 6.9 percent of Black patients.
The reviewers were specific about the direction of the error: darker skin more often causes pulse oximeters to overestimate oxygen saturation, missing a real problem rather than inventing a false alarm. That is arguably the worse failure mode, since it is the direction that leads someone to skip a call they should have made.
Most of this research comes from clinical, fingertip-style devices used in hospitals rather than wrist-worn consumer electronics specifically. But the underlying optical problem, melanin interfering with a light-absorption reading, applies to any reflectance or transmittance sensor. There is little reason to expect a smart watch is somehow immune to a bias documented across such a wide range of pulse oximeter types.
What about blood pressure and blood sugar readings on a smart watch?
Cuffless blood pressure features on smart watches remain largely unvalidated according to the American Heart Association, and the FDA has told consumers outright not to use smart watches or smart rings to measure blood glucose.
SpO2 is not the only wrist metric that gets more marketing confidence than the evidence supports. Smart watches blood pressure features, often sold under names like blood pressure trends or cuffless estimation, work by inferring pressure from pulse wave timing rather than measuring it directly the way an inflatable cuff does. Does a smart watch measure blood pressure the way a doctor's office cuff does? Not according to the American Heart Association's December 2025 scientific statement on cuffless devices, which found that many personal wearables, smart watches included, have not been proven accurate or reliable during exercise, sleep, or after taking medication that affects blood pressure. The statement adds a detail that surprises most people: FDA clearance for a blood pressure device does not require the standardized accuracy testing most buyers would assume it does, and roughly 80 percent of blood pressure monitors sold worldwide, cuffed and cuffless alike, have never been through formal validation at all.
Blood sugar tracking is the more clear-cut case. Can smart watches measure blood sugar without drawing blood? Not currently, and not legally as a marketed medical claim. In February 2024 the FDA issued a direct warning that no smart watch or smart ring is authorized to measure or estimate blood glucose, and using one for that purpose could produce a reading inaccurate enough to affect an insulin dose or another diabetes management decision. If a listing implies otherwise, that claim is ahead of what regulators have actually cleared.
Getting real use out of the health sensors you already have
A smart watch's health sensors work best as an early-warning trend line, not a replacement for a pulse oximeter, blood pressure cuff, or glucose meter, and pairing the watch with one of those dedicated tools closes the gap.
None of this makes wrist-worn sensors pointless. It means the honest use case is narrower than the marketing suggests: track the trend, notice when something moves outside your normal range, and confirm with a purpose-built device before treating a number as fact. A budget-tier fingertip pulse oximeter closes the accuracy gap the research above describes, and a validated upper-arm blood pressure cuff does the same job for blood pressure that a smart watch estimate cannot yet do on its own.
If you're shopping for a watch mainly because of its health-tracking claims, our smart watches buying guide walks through what each sensor can and cannot actually tell you before you pay for a feature you might not need. For a side-by-side look at current models across price tiers, our best smart watches roundup is a better starting point than any single spec sheet, and if you're narrowing things down by fit and features aimed at women's health tracking specifically, our best smart watches for women guide covers models with more complete sensor arrays.
So, can you trust smart watch SpO2 readings enough to skip owning a real pulse oximeter entirely? Not if a genuine health question is on the table. Trust the trend, verify the outlier.