9 healthy adults aged 19 to 28: 5 men and 4 women. They lay on a stretcher and breathed a controlled gas mix that took their blood oxygen from about 100% down to about 60% in steps. They wore an Apple Watch Series 7 on the right wrist and a Masimo MightySat Rx, an FDA-cleared fingertip oximeter, on the right middle finger. Each reading was compared with blood taken from an artery at the same time: 308 samples in all.

Both devices tended to read higher than the blood sample. Average gap. The watch read 3.26 points above the blood sample and the fingertip oximeter 1.80. Under 88%, that was 4.99 and 1.96. Within 2 points. 32.14% of watch attempts and 49.03% of fingertip attempts. More than 2 points too high. 56.49% and 44.48% of attempts. Under 88%, 85.33% and 52.44% of readings. No reading. The watch gave none on 24 of 308 attempts, the fingertip oximeter on 3. The error grew as oxygen fell. Under 88%, the root mean square error (Arms) was 5.82 for the watch and 3.52 for the fingertip oximeter.

Apple Watch vs Finger Oximeter in Low Oxygen

Q: What can't it tell you?

Overnight use. These were spot readings on people lying still, with the watch started by hand. Other people. The authors say nine young, healthy adults may not represent older people or people with lung conditions. All skin tones. Nobody in the very light, dark or very dark groups took part. See pulse oximeters and skin tone. Funding. AstraZeneca part-funded it, alongside US public grants. One author sits on a Google consumer health advisory board and consults for Samsung Research America.

A 2026 lab study lowered blood oxygen in 9 adults and compared an Apple Watch Series 7 and a fingertip oximeter with blood samples. Both read high, the watch by more.

PulseOxTracker Team9 October 2026

Infographic. Apple Watch vs fingertip oximeter when blood oxygen is low. Average reading above the blood sample when blood oxygen was under 88%: Apple Watch Series 7 on the wrist plus 4.99 points, Masimo MightySat Rx on the fingertip plus 1.96 points. Who was studied: 9 healthy adults aged 19 to 28, 308 blood samples, blood oxygen lowered in a lab from about 100% to about 60%. What the study found. Within 2 points of the blood sample: watch 32.14% of attempts, fingertip 49.03%. More than 2 points too high: watch 56.49% of attempts, fingertip 44.48%. More than 2 points too high, under 88%: watch 85.33% of readings, fingertip 52.44%. Error size under 88%, called Arms: watch 5.82, fingertip 3.52, both over the FDA limits. Showed above 92% when blood was under 88%: watch 6 of 150 readings, fingertip 3 of 164. No reading: watch 24 of 308 attempts, fingertip 3 of 308. Fast drops of 2 points a minute or more: no statistically significant change for either device. What it cannot show: how either device does overnight. 9 young, healthy people lying still, spot readings only, and no very light, dark or very dark skin tones. — What the study found: when blood oxygen was under 88%, the Apple Watch Series 7 read 4.99 points above the blood sample on average and the fingertip oximeter 1.96.

A smartwatch and a fingertip oximeter both give you a blood oxygen number. Earlier comparisons mostly didn't check how they do when oxygen is actually low. A 2026 study from Duke University lowered people's blood oxygen on purpose and compared both devices with blood samples.

Who was studied?

9 healthy adults aged 19 to 28: 5 men and 4 women.
They lay on a stretcher and breathed a controlled gas mix that took their blood oxygen from about 100% down to about 60% in steps.
They wore an Apple Watch Series 7 on the right wrist and a Masimo MightySat Rx, an FDA-cleared fingertip oximeter, on the right middle finger.
Each reading was compared with blood taken from an artery at the same time: 308 samples in all.

What did it find?

Both devices tended to read higher than the blood sample.

Average gap. The watch read 3.26 points above the blood sample and the fingertip oximeter 1.80. Under 88%, that was 4.99 and 1.96.
Within 2 points. 32.14% of watch attempts and 49.03% of fingertip attempts.
More than 2 points too high. 56.49% and 44.48% of attempts. Under 88%, 85.33% and 52.44% of readings.
No reading. The watch gave none on 24 of 308 attempts, the fingertip oximeter on 3.

The error grew as oxygen fell. Under 88%, the root mean square error (Arms) was 5.82 for the watch and 3.52 for the fingertip oximeter. The FDA recommends under 3.5 for wrist-style sensors and under 3 for fingertip ones, and the authors say both were over.

Did either one miss low oxygen?

When blood oxygen was under 88%, the watch showed above 92% in 6 of 150 readings and the fingertip oximeter in 3 of 164.

Fast drops, of 2 points a minute or more, made no statistically significant difference to the average gap for either device.

What can't it tell you?

Overnight use. These were spot readings on people lying still, with the watch started by hand.
Other people. The authors say nine young, healthy adults may not represent older people or people with lung conditions.
All skin tones. Nobody in the very light, dark or very dark groups took part. See pulse oximeters and skin tone.
Funding. AstraZeneca part-funded it, alongside US public grants. One author sits on a Google consumer health advisory board and consults for Samsung Research America.

What can you do with this?

A wearable's oxygen number is an estimate, and in this study it was less exact when oxygen was low. If you feel unwell, or your readings worry you, talk to a doctor.

Read smartwatch SpO2 vs a pulse oximeter and how accurate pulse oximeters are.

Not a medical device. Does not diagnose, treat or prevent any condition. Discuss results and changes with a clinician.

References

Jiang Y, Spies C, Roghanizad AR, Wang WK, Bhosai SJ, Snyder L, Burke A, MacLeod D, Dunn J. Performance of Wearable Pulse Oximetry During Controlled Hypoxia Induction: Instrument Validation Study. JMIR Formative Research. 2026;10:e85253. doi:10.2196/85253
US Food and Drug Administration. Pulse Oximeters - Premarket Notification Submissions [510(k)s]: Guidance for Industry and Food and Drug Administration Staff. March 2013. fda.gov