The short answer: consumer sleep trackers are good at telling whether you are asleep and poor at telling when you are briefly awake. In peer-reviewed studies against polysomnography, the hospital standard, rings and watches usually identify more than 90% of sleep correctly but catch only between roughly 20% and 75% of the time you spend awake. Their sleep stage estimates, light, deep and REM, agree with the lab only moderately. EEG headbands, which read brain activity directly, come much closer. Most sleep earbuds have no published validation at all.
That does not make trackers useless. It means they are better at showing trends over weeks than at judging a single night, and they tend to make your sleep look slightly better than it was.
What sleep trackers are measured against
Polysomnography (PSG) is an overnight test in a sleep clinic or at home with a technician’s kit. It records brain waves (EEG), eye movements, muscle tone, heart rhythm, breathing and blood oxygen. A trained scorer then labels every 30-second “epoch” of the night as wake, N1, N2, N3 (deep) or REM sleep, following American Academy of Sleep Medicine rules.
Even this gold standard has limits. In the Dreem headband study published in Sleep in 2020, five experienced sleep scorers agreed with a consensus score on average 86.4% of the time. So no consumer device should be expected to reach 100%; the realistic ceiling is roughly what two human experts manage.
Researchers report accuracy in a few ways. Sensitivity is how much true sleep a device labels as sleep. Specificity is how much true wake it labels as wake. Cohen’s kappa measures agreement beyond chance, where 0.4 to 0.6 is usually called moderate and 0.6 to 0.8 substantial.
The key studies at a glance
| Study | Journal, year | Devices | People | Headline result |
|---|---|---|---|---|
| Chinoy et al. | Sleep, 2021 | 7 wrist and bedside trackers (Fitbit, Garmin, others) | 34 healthy adults | Sleep sensitivity at least 93%; wake specificity only 18% to 54% |
| Lee et al. | JMIR mHealth and uHealth, 2023 | 11 wearables, bedside and phone trackers | 75 adults with sleep complaints | Four-stage agreement ranged from slight to moderate; Oura Ring 3 kappa 0.35, Fitbit Sense 2 0.42, Galaxy Watch 5 0.42, Apple Watch 8 0.30 |
| Svensson et al. | Sleep Medicine, 2024 | Oura Ring Gen3 | 96 adults, up to 3 nights each | Sleep sensitivity 94.4%, wake specificity 73% to 75%, stage accuracy 75.5% to 90.6% |
| Robbins et al. | Sensors, 2024 | Oura Gen3, Apple Watch 8, Fitbit Sense 2 | 35 adults | Four-stage kappa: Oura 0.65, Apple 0.60, Fitbit 0.55 (Oura-funded) |
| Schyvens et al. | SLEEP Advances, 2025 | 6 wrist trackers incl. Apple Watch 8, Whoop 4.0, Garmin | 62 adults | Kappa 0.21 to 0.53; wake specificity 29% to 52% |
| Arnal et al. | Sleep, 2020 | Dreem EEG headband | 25 adults | 83.5% staging accuracy vs 86.4% for expert scorers |
| Lanthier et al. | SLEEP Advances, 2026 | Muse-S EEG headband | 56 adults (47 usable) | Kappa 0.76; total sleep within 6 minutes of PSG on average |
Smart rings: strong on sleep, weaker on wake
Rings estimate sleep from your pulse, heart rate variability, temperature and movement. Oura is the most studied by far.
One of the biggest ring studies, by Svensson and colleagues in Sleep Medicine in 2024 (doi.org/10.1016/j.sleep.2024.01.020), compared the Oura Gen3 with its 2.0 algorithm against home PSG in 96 generally healthy Japanese adults aged 20 to 70, over up to three nights each. The ring labelled 94.4% of sleep correctly and 73% to 75% of wake correctly. Accuracy for individual stages ran from 75.5% for light sleep to 90.6% for REM, and the authors concluded it showed good agreement for total sleep and time in light and deep sleep.
A 2024 study in Sensors led by Rebecca Robbins at Brigham and Women’s Hospital (pmc.ncbi.nlm.nih.gov/articles/PMC11511193) found the Oura Gen3 had a four-stage kappa of 0.65, ahead of the Apple Watch Series 8 at 0.60 and Fitbit Sense 2 at 0.55. All three identified at least 95% of sleep. Oura funded that study, and Robbins is a medical adviser to the company.
The picture is less flattering elsewhere. In the 2023 Korean study by Lee and colleagues in JMIR mHealth and uHealth (mhealth.jmir.org/2023/1/e50983), which recruited adults with sleep complaints rather than good sleepers, the Oura Ring 3 scored a kappa of 0.35, behind the Fitbit Sense 2, Galaxy Watch 5 and Google Pixel Watch. Population matters: devices tend to do worse on people who sleep badly, which is precisely who buys them.
Other rings, including RingConn, Ultrahuman and Samsung’s Galaxy Ring, have far less independent PSG research published. That is a gap in the evidence, not proof that they are worse. Our best smart rings guide notes where evidence exists.
Watches and wrist trackers: similar story, wider spread
The 2021 study in Sleep by Chinoy and colleagues at the US Naval Health Research Center (pmc.ncbi.nlm.nih.gov/articles/PMC8120339) tested seven consumer trackers in 34 healthy adults. Every device caught at least 93% of sleep, but specificity for wake ranged from 18% to 54%, and the authors singled out the two Garmin models as the weaker performers. Devices tended to do worse on nights of poorer or deliberately disrupted sleep.
The most recent wrist study, by Schyvens and colleagues in SLEEP Advances in 2025 (doi.org/10.1093/sleepadvances/zpaf021), compared six wrist devices with PSG in 62 adults at Antwerp University Hospital’s sleep centre, including people referred with suspected sleep apnoea. The Apple Watch Series 8 did best (kappa 0.53, wake specificity 52%), followed by the Fitbit Sense and Charge 5. Whoop 4.0, Withings ScanWatch and Garmin Vivosmart 4 trailed, with the Garmin catching only 29% of wake. All six overestimated sleep efficiency, four of the six differed significantly from PSG on total sleep time, and the authors concluded that none can replace PSG and that their algorithms need further work.
The Lee study also found that most of the wearables overestimated sleep efficiency by counting wake as sleep, with the Oura the exception. Put simply: your watch probably thinks you slept better than you did.
Sleep earbuds: comfort first, measurement second
Sleep earbuds such as the Ozlo Sleepbuds 2 and Soundcore Sleep A30 produce sleep reports from motion sensors in the buds and microphones and environmental sensors in the bedside charging case. We found no peer-reviewed study validating either against PSG. Their main job is blocking or masking noise, and they do that well; treat their sleep charts as a diary rather than data. Our best sleep earbuds guide covers what they are good at.
EEG headbands: the closest you can get at home
Headbands that read brain waves through dry electrodes on the forehead and behind the ears measure the same signal a sleep scorer uses, so they should be more accurate in principle. The research agrees.
Arnal and colleagues, publishing in Sleep in 2020 (doi.org/10.1093/sleep/zsaa097), found the Dreem headband staged sleep with 83.5% accuracy in 25 adults, against 86.4% for five human experts. The lead author was affiliated with Dreem, the device’s developer, which is worth bearing in mind.
In 2026, Lanthier and colleagues in SLEEP Advances (doi.org/10.1093/sleepadvances/zpaf089) tested the Muse-S headband against full in-lab PSG in 56 adults, 17% of whom had a sleep-related breathing disorder. Usable data came from 47 people. Agreement was substantial overall (kappa 0.76), near-perfect for REM and wake, substantial for N2 and deep sleep, and only fair for N1, the lightest stage. Across stages, accuracy ran from 88% to 96%. Total sleep time was on average 6 minutes higher than PSG. Interaxon, Muse’s maker, supplied the headbands and the automated scoring but did not fund the study. That is markedly better than any ring or watch study above. Whether the newer Muse S Athena performs identically has not been separately published; we cover it in our Muse S Athena review.
The price of that accuracy is comfort. You wear a band on your head all night, and if it slips, data is lost: 16% of recordings in the Muse study were not usable.
Accuracy by device type
| Device type | What it reads | Sleep vs wake | Sleep stages | Best published evidence |
|---|---|---|---|---|
| Smart ring | Pulse, HRV, temperature, movement | Good on sleep, moderate on wake | Moderate | Oura Gen3, Sleep Medicine 2024 |
| Watch | Pulse, HRV, movement | Good on sleep, weak to moderate on wake | Moderate at best | Six-device study, SLEEP Advances 2025 |
| Sleep earbuds | Movement, room noise, light, temperature | Not validated | Not validated | None found |
| EEG headband | Brain waves, plus pulse and movement | Near-perfect for wake | Substantial | Muse-S, SLEEP Advances 2026 |
How to use a sleep tracker sensibly
Watch the trend, not the night. A single score is noisy. Average sleep time, resting heart rate and heart rate variability over two to four weeks are far more reliable than any one morning’s number.
Do not chase deep sleep. Stage estimates are the least reliable part of the data. A night labelled with little deep sleep may simply be a scoring miss.
Check how you feel. If the score says 85 and you feel dreadful, believe yourself. Trying hard to improve a score can itself make sleep worse, a pattern researchers have called orthosomnia.
Know what a tracker cannot do. No ring, watch or earbud can diagnose insomnia, sleep apnoea, narcolepsy or restless legs. If you snore loudly with pauses in breathing, fall asleep during the day, or have struggled to sleep for months, see your GP. They can refer you for a sleep study, and persistent insomnia has effective treatments, including cognitive behavioural therapy for insomnia.
Mind the medication. Beta blockers and some other medicines change heart rate and HRV, which can shift tracker scores. Your pharmacist can explain what to expect.
Sleep tracker accuracy FAQs
What is the most accurate sleep tracker?
Among consumer devices, EEG headbands such as the Muse S come closest to a hospital sleep study, with a kappa of 0.76 in a 2026 study. Among rings and watches, Oura, Apple Watch and recent Fitbits have the best published results, though rankings vary between studies.
Do sleep trackers overestimate sleep?
Usually, yes. Because they are better at detecting sleep than wake, quiet time lying awake is often counted as sleep, which inflates total sleep time and sleep efficiency.
How accurate is deep sleep tracking?
Moderately at best for rings and watches. In the 2025 SLEEP Advances study of six wrist devices, overall stage agreement with PSG ranged only from fair to moderate. EEG headbands are considerably more accurate, though the Muse-S still over-counted deep sleep by about 15 minutes a night on average.
Can a sleep tracker diagnose sleep apnoea?
No. Some devices flag signs, such as blood oxygen dips or breathing changes, but diagnosis needs a sleep study. Speak to your GP if you are worried.
Is a smart ring more accurate than a watch for sleep?
The evidence is mixed. An Oura-funded 2024 study put the Oura ahead of an Apple Watch and Fitbit, while a 2023 Korean study put the Oura Ring 3 behind several watches. Comfort may matter more: the tracker you wear every night gives better trend data than the one in the drawer.
Are sleep earbuds accurate trackers?
We found no peer-reviewed validation of sleep tracking in Ozlo or Soundcore earbuds. Use them for noise, not numbers.
