Is Your Wearable's Skin Temperature Accurate? Nobody Has Checked
Last updated: October 12, 2026
No independent study has tested the temperature sensor of any consumer wearable against a reference thermometer. That holds for Oura, Apple Watch, Samsung Galaxy Watch and Galaxy Ring, WHOOP, Garmin, Fitbit and Pixel Watch, Ultrahuman and RingConn as of October 2026. The only sensor comparisons that exist are two company write-ups, Oura's from 2020 and Ultrahuman's from 2025, and neither is peer reviewed. What has been studied is whether the temperature signal can help estimate ovulation, and every one of those studies was funded or co-authored by the brand being tested. So wearable skin temperature accuracy is an open question, not a settled one. Here is what the evidence does and does not support.
If you want the wider picture of which wearable metrics have solid evidence behind them and which do not, the wearable accuracy factors tool maps it metric by metric, then read on for the temperature story specifically.
I went looking for a skin temperature validation study expecting to find at least one for the Oura Ring, since temperature is so central to how Oura sells itself. I searched PubMed, preprint servers, the FDA database and every major brand's science pages. The closest thing to an independent test I found was a study that compared a Pixel Watch 2 against another wristband and concluded the two could not be compared reliably.
What your wearable is actually measuring
The first thing to know is that your wearable is not reporting your temperature. Every major brand reports skin temperature overnight, and the ones that explain their method report it as a change from your own baseline. Apple says plainly that its feature is not a thermometer. That design choice is sensible. Skin temperature swings far more than core temperature over a day, so a single absolute reading would tell you very little. A shift against your own recent nights is the useful part.
Device | What it reports | Baseline needed | Published accuracy figure |
Apple Watch (Series 8 or later, all Ultra, SE 3) | Nightly change from baseline, two sensors sampling every 5 seconds during sleep | About 5 nights | None. Apple says it "is not a thermometer" |
Oura Ring | Nighttime skin temperature | Not stated in the sources we read | Company blog only (2020) |
Samsung Galaxy Watch5 or later | Skin temperature from an infrared sensor | Not stated | None |
Google Pixel Watch 2 to 5 | Nightly variation from baseline | 3 nights | None |
Fitbit Charge 6 | Skin temperature variation only, in the Google Health app | Not stated | None. Google notes ambient temperature changes may affect it |
WHOOP | Overnight skin temperature | Not stated | None |
Ultrahuman Ring | Overnight finger temperature | Not stated | Company validation page only (2025) |
Rings carry a temperature sensor almost across the board, and it is often part of the pitch. If you are choosing one partly for that feature, our smart ring comparison of Oura, RingConn, Ultrahuman and CUDIS covers what each one actually measures.
Apple's 0.1 °C figure, from its Series 8 launch, often gets quoted as an accuracy number. It is the smallest change the sensor can detect, which is a different thing. A sensor can register a 0.1 °C change perfectly and still read half a degree off.
The only sensor tests are company tests
Two companies have published a comparison of their temperature sensor against a reference device. Both deserve credit for publishing anything at all, since most brands have not. Neither is independent, and both report their headline result as a share of variation explained rather than as an error in degrees.
Oura (blog, December 2020) | Ultrahuman (science page, January 2025) | |
Who ran it | A task force of Oura employees | Ultrahuman |
Device | Ring generation not named; predates Gen 3 | "Ultrahuman Ring", model not named |
Lab test | Rings on their own, not on fingers, in water baths from 9 to 45 °C, against 7 iButtons | Water bath and open air against a lab data logger |
Lab result | r² above 0.99, changes as small as 0.13 °C detected, within 0.36 °C | 98.7% agreement |
Worn test | 16 people for one week | Worn overnight on the finger, participant count not stated |
Worn result | r² above 0.92 | 97.5% agreement (3,836 paired readings across both tests) |
Peer reviewed | No | No |
The numbers look near perfect, and that is the trap. Ultrahuman defines its 97.5% as the share of the reference thermometer's movement the ring reproduces, and says directly that it is not a percentage error. Oura's r² works the same way. A sensor can track every rise and fall of a reference while sitting consistently above or below it, and both of these figures would still look excellent.
Oura's within 0.36 °C figure is the closest thing to an error number, and it comes from the lab test where nobody was wearing the ring. In the worn test, Oura reports only r².
There is one more wrinkle. Oura's reference device, the iButton, is itself a research skin sensor with known error. That matters for the next section.
What a real temperature validation looks like
For contrast, here is what proper validation of a skin temperature sensor reports. Bach et al. 2026, a Griffith University study in Experimental Physiology, tested two research skin sensors against a wired thermocouple on 26 adults during exercise in 35 °C heat. The figures below are from the paper's abstract. These are not consumer wearables. They show the format a consumer validation would need.
Sensor | Mean bias vs thermocouple | 95% limits of agreement | Concordance (CCC) |
BodyCAP eTemp | −0.04 °C | −0.33 to +0.25 °C | 0.986 |
iButton DS1922L | +0.11 °C | −0.23 to +0.46 °C | 0.974 |
Two things stand out. First, this is the shape of an honest answer: a bias in degrees, a range that covers 95% of readings, and an agreement score. No consumer wearable has published all three.
Second, look at the iButton row. That is the same type of sensor Oura used as its reference. Against a thermocouple, its readings ran up to roughly half a degree off at the edges, and it does not meet the precision recommended by the ISO 9886 standard. So Oura's "within 0.36 °C" was measured against a reference that carries error of a similar size. The figures in this table are mean skin temperature across four body sites, which is the standard research method, so they are a benchmark rather than a like-for-like comparison with a ring.
Where the temperature signal has been tested: cycle tracking
If the sensors have not been tested, what has? The answer is the temperature signal combined with each company's algorithm, almost always for estimating ovulation. These studies tell you whether the overall system is useful for that job. They cannot tell you how many degrees the sensor is off.
Device | Study | Who took part | Key result | Funding and authors |
Oura | Thigpen 2025, J Med Internet Res | 1,155 cycles from 964 users | Detected 96.4% of ovulations; average error 1.26 days vs 3.44 for the calendar method | All authors Oura employees |
Oura | Maijala 2019, BMC Women's Health | 22 women, about 115 days each | Ring vs oral temperature r = 0.563; best algorithm found 83.3% of ovulations in a 6-day window | 2 of 5 authors Oura employees; lead author received an Oura salary grant |
Apple | Wang 2025, Human Reproduction | 260 people, 889 cycles | Estimate given in 80.8% of completed cycles, average error 1.22 days | Apple funded; used a prototype watch, not a retail model |
Samsung Galaxy Watch 5 | Kim 2026, Biosensors and Bioelectronics | 85 women, 225 cycles | Ovulation within 2 days in 64.0% (not statistically significant); within 3 days 77.8% | "Supported by Samsung Electronics" |
WHOOP | Gonzalez 2026, npj Digital Medicine | 2,596 women, 42,759 cycles | Describes how skin temperature varies across the cycle. No accuracy figures | Stanford lead authors; 3 co-authors WHOOP employees holding shares |
The Oura figures are strong, and the 2025 study is large. The Samsung result is the most cautious in the set, and it came from the only study that confirmed ovulation with a blood test rather than a urine kit. None of this is a reason to dismiss the feature. It is a reason to read each result with its funder beside it.
Regulators have looked at a related question. Natural Cycles, an FDA-cleared birth control app, accepts temperature from Oura and Apple Watch, and has since added Samsung Galaxy Watch and Garmin. Its FDA filings compared Oura and Apple Watch against an oral thermometer: 40 women and 223 cycles for Oura, 272 users and 1,918 cycles for Apple Watch. Neither reports a temperature agreement figure, and neither names the device generation. A plan first cleared in 2024, and renewed in July 2026, lets Natural Cycles add new wearables as temperature sources without a new filing. That clearance covers the app's algorithm, not the accuracy of any sensor.
This is where your own data becomes the more useful test. Kygo connects a fast food log (voice, photo, barcode or a plain description) to the metrics your Oura, Apple Watch, Fitbit, Garmin or WHOOP already reports, so you can see how your nights actually move with what you ate and drank instead of reading one number in isolation. Get it on iOS or Android.
Why skin temperature is hard to measure well
Skin is the body's radiator. It warms and cools with room temperature, bedding, blood flow and where on the body you measure. Oura's own blog makes this point: core temperature moves about 0.5 to 1 °C in a day, while skin temperature can swing 15 °C or more. That is exactly why every brand reports change against your baseline rather than a raw value.
It also means placement and conditions matter. Google's Charge 6 spec page warns that significant ambient temperature changes may affect skin temperature tracking. A finger and a wrist sit in different thermal conditions, and no study has compared ring and wrist temperature sensors head to head. Our post on where to wear your fitness tracker covers what the research says about placement for the metrics that have been tested.
Software matters as much as hardware here. The number you see is the sensor reading after the brand's own filtering and baseline math, none of which is published. If your trend line suddenly jumps, goes flat or disappears after an app or firmware update, check for a software change before reading meaning into it.
How to use your temperature data anyway
None of this makes the feature useless. A sensor that reliably tracks change can be valuable even if its absolute reading is unknown. The practical approach is to treat it as a trend line about you, not a measurement you can compare with anyone else.
Compare against yourself only. Your baseline is the point. A deviation of plus 0.4 °C on your ring means something in the context of your last few weeks and nothing next to a friend's watch.
Keep conditions steady. Same bedroom, similar bedding, the device worn the same way. Ambient changes show up in skin temperature, and a hotel room can look like a fever.
Look for clusters, not single nights. One warm night is noise. Several in a row alongside a higher resting heart rate and lower HRV is a pattern worth noticing. Your temperature reading is also one of the inputs to most recovery scores, which we cover in can you trust your recovery score.
Do not use it as a thermometer. If you think you have a fever, use a thermometer. Apple and Google both say their features are not built for that.
Common questions
How accurate is wearable skin temperature?
Nobody knows yet, because no independent study has tested a consumer wearable's temperature sensor against a reference thermometer. Oura and Ultrahuman have published their own comparisons showing the sensor tracks a reference closely, but both report variation explained rather than error in degrees, and neither is peer reviewed.
Is the Oura Ring temperature sensor accurate?
Oura's 2020 internal test found the ring detected changes as small as 0.13 °C and stayed within 0.36 °C of a research sensor in the lab, with the rings not being worn. In a one-week worn test with 16 people, it reported r² above 0.92 and no error figure. The ring generation tested is not named and predates Gen 3. Nothing has been published for Ring 4 or Ring 5.
Is Apple Watch wrist temperature accurate?
Apple has not published an accuracy figure for its wrist temperature sensor. The "0.1 °C" often quoted is the smallest change the sensor can detect, not its accuracy. Apple's support page says the feature is not a thermometer and reports nightly change from your baseline after about five nights.
Which wearable has the most accurate temperature sensor?
There is no evidence to answer this. No study has compared temperature sensors across brands against a reference, so any ranking you see online is not based on data.
Can a smart ring or watch detect a fever?
None of them has been validated for fever detection, and Apple and Google both say their temperature features are not thermometers or for medical use. A sustained rise against your baseline can be a useful prompt to check with an actual thermometer.
Why is my skin temperature different on my ring and my watch?
Fingers and wrists sit at different temperatures and respond differently to the room and blood flow, and each brand uses its own baseline method. Two devices disagreeing is expected and does not tell you which is closer to the truth.
Is wearable temperature accurate enough for cycle tracking?
Brand-funded studies suggest the combined system can be useful: Oura's 2025 study detected 96.4% of ovulations, Apple's prototype study averaged 1.22 days of error, and Samsung's study was more modest at 64.0% within two days. Natural Cycles accepts wearable temperature data under FDA clearance, though that clearance covers its app, not sensor accuracy.
The bottom line
Every major wearable reports skin temperature, and not one sensor has been independently tested against a reference thermometer. The only sensor numbers come from Oura and Ultrahuman themselves, and they measure how closely the ring tracks a reference, not how far off it is. The cycle tracking research is real and some of it is large, but all of it is brand-funded and tests the algorithm and sensor together.
That leaves a clear way to use the feature: as a personal trend line, read against your own baseline, under steady conditions, alongside your other metrics. Read that way, a temperature shift can be a useful signal. Read as a thermometer, it is a number nobody has checked.
Want to see what actually moves your nights? Kygo lines up your food log with your sleep, HRV and resting heart rate so the patterns show up in your own data rather than in a population average. Get it on iOS or Android.
Key sources
Oura's How Accurate Is Oura's Temperature Data? (company blog, December 2020, not peer reviewed); Ultrahuman's temperature sensor accuracy page (January 2025, company published); Bach et al. 2026, Experimental Physiology, n=26, independent; Thigpen et al. 2025, J Med Internet Res, Oura employees; Maijala et al. 2019, BMC Women's Health, n=22, Oura-affiliated; Wang et al. 2025, Human Reproduction, Apple-funded; Kim et al. 2026, Biosensors and Bioelectronics, Samsung-supported; Gonzalez et al. 2026, npj Digital Medicine, WHOOP co-authors; Johnson et al. 2026, npj Women's Health, independent scoping review; Brandebusemeyer et al. 2026, Sensors, n=20; FDA 510(k) records K202897, K231274 and K253694 (Natural Cycles); and manufacturer pages from Apple, Google, Samsung and WHOOP, read October 2026.
A note on what this post leaves out
Several "most accurate wearable for temperature" rankings circulate online. We could not trace any of them to a study, so none appear here. We have also left out a 2026 SSRN preprint sometimes described as a long-term skin temperature comparison: it contains no temperature data, studies one person, and was written by the maker of one of the devices. The Pixel Watch 2 study by Brandebusemeyer and colleagues is cited above but not tabled, because it compared the watch against another wristband rather than a reference and reported no agreement figures. Finally, Oura's blog lists its top test temperature as "45 °C (81 °F)". 45 °C is 113 °F, so we quote the Celsius figure only.
Disclaimer: Kygo is a personal data aggregation and insights platform designed for informational purposes only. The information provided by Kygo, including correlations, patterns, and trends identified in your data, does not constitute medical advice, diagnosis, or treatment. Always consult a licensed healthcare provider with any questions regarding medical conditions.
If you have checked your ring or watch's temperature trend against a thermometer during an illness, I would like to hear how closely the two moved together.