What does the research show about sleep trackers?

Sleep trackers

Validation studies against polysomnography report that consumer trackers detect more than 90% of sleep epochs and far fewer wake epochs, so total sleep time is usually overestimated by a margin that differs by device. Sleep stage agreement is lower and varies widely. Two trials reported lower sleep disturbance scores with tracker feedback. Case reports describe preoccupation with tracker data.

Checked 2026-11-13

Status: Sold in the US as general wellness products. FDA's January 2026 guidance lists tracking sleep trends as a general wellness claim and states that the agency does not intend to examine low risk general wellness products for compliance with premarket review requirements [1].
Measures: Sleep and wake timing, total sleep time and sleep stages, estimated from movement and pulse sensors worn on a finger or wrist, or from pressure sensors placed under a mattress.
Also called: sleep tracker, consumer sleep technology, sleep ring, smart ring, sleep watch, sleep mat, under-mattress sleep sensor, Oura Ring, WHOOP, Fitbit sleep, Apple Watch sleep, Garmin sleep, Withings Sleep Analyzer
Used for: sleep tracking, sleep duration, total sleep time, sleep stages, deep sleep, REM sleep, wake time, sleep quality, sleep score, insomnia, recovery

Strong several reviews agree · Moderate one review or consistent trials · Limited one or a few small studies · Untested no study in people, listed with what it rests on

Matched the reference method5

ModerateShown to estimate total sleep time within 3 minutes of reference sleep studies on average across 6 pooled studies, with larger differences for individual nights.
ModerateShown to classify sleep versus wake with 87% pooled accuracy across finger-worn devices, mostly in healthy participants.
LimitedOne study found all three devices estimated total sleep time within 10 minutes of polysomnography on average in 35 healthy adults over one night.
LimitedOne study found six devices agreed with polysomnography on sleep versus wake in 86 to 89% of epochs in 53 healthy young adults.
LimitedOne study found time in light and deep sleep did not differ from polysomnography in 96 healthy adults, while REM sleep was underestimated by 4 to 6 minutes.

Reported risks2

LimitedCase reports: patients sought treatment for sleep problems they inferred from tracker data. One survey classified 3 to 14% of respondents as having orthosomnia.
LimitedOne trial found people with insomnia given sham negative sleep feedback reported more daytime sleepiness and less alertness than those given sham positive feedback.

How it compares with the reference method

Matched the reference method5

Limited

Total sleep time, Apple Watch Series 8, Fitbit Sense 2 and Oura Ring Gen3

One study found all three devices estimated total sleep time within 10 minutes of polysomnography on average in 35 healthy adults over one night.

The numbers and sources

Device: Apple Watch Series 8, Fitbit Sense 2, Oura Ring Gen3
Compared with: Polysomnography, one inpatient night
Effect: Mean total sleep time was 421 minutes on the Oura Ring against 430 on polysomnography, 428 on the Fitbit against 431, and 442 on the Apple Watch against 434. None of the three differences was statistically significant. Intraclass correlation with polysomnography was 0.74 for Oura, 0.56 for Fitbit and 0.85 for Apple Watch [3].
Population: 35 adults aged 20 to 50 without a sleep disorder, one night. Fitbit data came from 33 and Apple Watch data from 29 [3].
Sources: [3]
Caveats: Funded by Oura Ring Inc., and the lead author sits on Oura's medical advisory board. The authors note that nightly estimates for an individual may diverge more widely than the group averages. Healthy sleepers only, one night [3].

Moderate

Total sleep time, Oura Ring

Shown to estimate total sleep time within 3 minutes of reference sleep studies on average across 6 pooled studies, with larger differences for individual nights.

The numbers and sources

Device: Oura Ring, generations pooled [9]. Oura Ring Gen3 with Oura Sleep Staging Algorithm 2.0 [10]. Oura Ring Gen 3 [11]. The ŌURA ring as tested in a study first published in 2017 [12].
Compared with: Polysomnography. The meta-analysis also included studies that used actigraphy as the reference [9].
Effect: Across 6 studies, the ring did not differ significantly from the reference for total sleep time (mean difference -2.97 minutes, 95% CI -10.27 to 4.33), sleep efficiency (-1.32%, 95% CI -2.76 to 0.12) or wake after sleep onset (1.64 minutes, 95% CI -12.57 to 15.86) [9]. In 96 adults measured for up to three nights, the Gen3 ring did not differ from polysomnography for total sleep time and underestimated sleep efficiency by 1.1 to 1.5% [10]. In 41 adolescents and young adults, differences in total sleep time were within 30 minutes for 87.8% of the sample [12]. In 45 sleep clinic patients, the ring overestimated total sleep time by 11.7 minutes on average (SD 37.1), and individual differences ranged from -54.0 to 97.5 minutes [11].
Population: 388 participants in 6 studies [9]. 96 generally healthy Japanese adults aged 20 to 70 [10]. 41 healthy adolescents and young adults, mean age 17.2 [12]. 45 sleep clinic patients aged 23 to 74, with ring data for 31 nights [11].
Sources: [9], [10], [12], [11]
Caveats: The meta-analysis mixed polysomnography and actigraphy references [9]. In sleep clinic patients the ring overestimated more when true sleep time was lower [11]. A separate study in older adults found the 3rd generation ring underestimated total sleep time by 75.5 minutes [7].

Limited

Sleep and wake detection, Apple Watch S6, Garmin Forerunner 245 Music, Oura Ring Generation 2, Polar Vantage V, Somfit and WHOOP 3.0

One study found six devices agreed with polysomnography on sleep versus wake in 86 to 89% of epochs in 53 healthy young adults.

The numbers and sources

Device: Apple Watch S6, Garmin Forerunner 245 Music, Oura Ring Generation 2, Polar Vantage V, Somfit, WHOOP 3.0
Compared with: Polysomnography, one laboratory night with 9 hours in bed
Effect: Agreement on sleep versus wake, with Cohen's kappa, was 88% (0.30) for Apple Watch, 89% (0.35) for Garmin, 87% (0.44) for Polar, 89% (0.51) for Oura, 86% (0.44) for WHOOP and 87% (0.48) for Somfit. Agreement on individual sleep stages was lower, from 50% (kappa 0.25) for Garmin to 65% (kappa 0.52) for Somfit. The authors concluded all six were valid for the timing and duration of sleep and that all six required improvement for specific sleep stages [17].
Population: 53 adults (26 women), mean age 25.4, one night [17].
Sources: [17]
Caveats: Young healthy adults on a single night with a fixed 9 hours in bed. Apple Watch sleep was scored through Sleep Watch, a third-party app, in 5 minute epochs. The authors belong to a research group that receives funding and equipment from WHOOP Inc. and state WHOOP had no role in the study [17].

Moderate

Sleep and wake detection, finger-worn devices pooled

Shown to classify sleep versus wake with 87% pooled accuracy across finger-worn devices, mostly in healthy participants.

The numbers and sources

Device: 11 finger-worn devices pooled across 28 articles, including Oura Ring generations
Compared with: Polysomnography
Effect: Pooled accuracy for classifying sleep versus wake was 87% (95% CI 86 to 89%) [19].
Population: 28 articles published up to April 2025. Sleep staging studies enrolled primarily healthy participants [19].
Sources: [19]
Caveats: The authors describe heterogeneous sensors and algorithms across devices, and call for algorithmic transparency, more diverse populations and standardized methods [19].

Limited

Sleep stages, Oura Ring Gen3

One study found time in light and deep sleep did not differ from polysomnography in 96 healthy adults, while REM sleep was underestimated by 4 to 6 minutes.

The numbers and sources

Device: Oura Ring Generation 3 with Oura Sleep Staging Algorithm 2.0, worn on the dominant and non-dominant hands
Compared with: Ambulatory polysomnography, up to three nights per person
Effect: Time in light sleep and deep sleep did not differ significantly from polysomnography. REM sleep was underestimated by 4.1 to 5.6 minutes. Epoch accuracy by stage ranged from 75.5% (light sleep) to 90.6% (REM sleep). For sleep versus wake, sensitivity was 94.4 to 94.5% and specificity 73.0 to 74.6% [10].
Population: 96 generally healthy Japanese adults aged 20 to 70, contributing 421,045 epochs [10].
Sources: [10]
Caveats: Generally healthy adults in one country. Other studies of the same ring generation reported different results: 53% four-stage accuracy in sleep clinic patients [11], fair agreement (kappa 0.2 to 0.4) in a hospital and clinic sample [20], and a 71.5 minute overestimate of deep sleep in older adults [7].

Mixed results7

Limited

Total sleep time, Apple Watch Series 8, Fitbit Charge 5, Fitbit Sense, Garmin Vivosmart 4, WHOOP 4.0 and Withings ScanWatch

Mixed results: one study found mean bias in total sleep time from 6 to 40 minutes across six wrist devices, with wide limits of agreement for all six.

The numbers and sources

Device: Apple Watch Series 8, Fitbit Charge 5, Fitbit Sense, Garmin Vivosmart 4, WHOOP 4.0, Withings ScanWatch
Compared with: Polysomnography, one laboratory night
Effect: Mean bias against polysomnography was 6.3 minutes for Fitbit Sense (p = 0.285), 19.6 for Apple Watch Series 8, 24.5 for WHOOP 4.0, 38.4 for Garmin Vivosmart 4 and 39.9 for Withings ScanWatch, each an overestimate. The Fitbit Charge 5 difference was not statistically significant (p = 0.051). Limits of agreement ran from -38.4 to 77.6 minutes for Apple Watch Series 8 and from -124.7 to 137.3 minutes for Fitbit Sense. The authors treated a bias under 30 minutes as clinically satisfactory and named Apple Watch Series 8, Fitbit Charge 5 and Fitbit Sense as meeting it [2].
Population: 62 adults (52 men, 10 women, mean age 46.0), with and without suspected sleep apnea, each wearing two to four devices for one night. Usable nights per device ranged from 20 (Apple Watch) to 41 (Withings) [2].
Sources: [2]
Caveats: The paper's table and text give different Fitbit Charge 5 biases (11.1 and -5.7 minutes). Apple Watch data were missing or partial for 15 participants and Garmin data for 18. Firmware versions are not reported. Funded by Flanders Innovation and Entrepreneurship, and the authors declared no financial ties [2].

Moderate

Total sleep time, Fitbit wristband models pooled

Mixed results: early Fitbit models without sleep staging overestimated total sleep time by 7 to 67 minutes, while sleep-staging models did not differ significantly from polysomnography.

The numbers and sources

Device: Fitbit wristband models, pooled as early models without sleep staging and later sleep-staging models
Compared with: Polysomnography
Effect: Fitbit models without sleep staging overestimated total sleep time (by about 7 to 67 minutes) and sleep efficiency (by about 2 to 15%) and underestimated wake after sleep onset (by about 6 to 44 minutes). They identified sleep epochs with sensitivity of 0.87 to 0.99 and specificity of 0.10 to 0.52. Sleep-staging models showed no significant difference from polysomnography in total sleep time (P = .29), wake after sleep onset (P = .25) or sleep efficiency (P = .19), underestimated sleep onset latency (P = .03), and had sensitivity of 0.95 to 0.96 and specificity of 0.58 to 0.69 [8].
Population: 22 articles in the systematic review, 8 of them in the meta-analysis [8].
Sources: [8]
Caveats: The review was published in 2019 and covers Fitbit models released before then. Its authors concluded the devices are of limited specificity and are not a substitute for polysomnography [8].

Limited

Total sleep time, WHOOP

Mixed results: small studies in healthy young adults found WHOOP within 14 minutes of polysomnography on average, while a later study of WHOOP 4.0 found a 24 minute overestimate.

The numbers and sources

Device: WHOOP strap, generation not stated in the abstracts [13][14]. WHOOP models pooled in a systematic review [15]. WHOOP 4.0 [2].
Compared with: Polysomnography. The systematic review also included a study using the Sleep Profiler EEG monitor [15].
Effect: Over 86 sleeps in 12 adults, WHOOP overestimated total sleep time by 8.2 minutes (SD 32.9), a difference that was not statistically significant. Bedtimes were entered by the researchers [13]. In 32 healthy adults, bias for sleep duration was 13.8 minutes and precision error 17.8 minutes [14]. A systematic review of 8 articles reported a WHOOP disagreement of -1.4 minutes for total sleep time and 21.0 minutes for REM sleep [15]. In 40 adults wearing WHOOP 4.0, total sleep time was overestimated by 24.5 minutes (p = 0.010), with limits of agreement from -101.1 to 150.0 minutes [2].
Population: 12 healthy adults, mean age 22.9, in a 10 day laboratory protocol [13]. 32 healthy adults, mean age 23.8, one night [14]. 8 articles [15]. 40 adults with and without suspected sleep apnea, one night [2].
Sources: [13], [14], [15], [2]
Caveats: In the 12-person study, sleep was scored by the company using its own algorithms, and its authors state the result depends on bedtimes being entered manually [13]. Samples were small and mostly young and healthy.

Limited

Sleep and wake detection, Apple Watch Series 8, Fitbit Charge 5, Fitbit Sense, Garmin Vivosmart 4, WHOOP 4.0 and Withings ScanWatch

One study found six wrist devices detected more than 90% of sleep epochs and 29 to 52% of wake epochs.

The numbers and sources

Device: Apple Watch Series 8, Fitbit Charge 5, Fitbit Sense, Garmin Vivosmart 4, WHOOP 4.0, Withings ScanWatch
Compared with: Polysomnography, one laboratory night
Effect: Sensitivity for sleep epochs ranged from 91.7% (Fitbit Charge 5) to 96.3% (Apple Watch Series 8). Specificity for wake epochs was 52.2% for Apple Watch Series 8, 48.8% for Fitbit Sense, 47.5% for Fitbit Charge 5, 40.1% for WHOOP 4.0, 31.1% for Withings ScanWatch and 29.4% for Garmin Vivosmart 4. Cohen's kappa ranged from 0.21 (Garmin) to 0.53 (Apple Watch), which the authors describe as fair to moderate agreement [2].
Population: 62 adults (mean age 46.0) with and without suspected sleep apnea, one night [2].
Sources: [2]
Caveats: Usable nights per device ranged from 20 to 41. Firmware versions are not reported [2].

Limited

Sleep and wake detection, EarlySense Live, Fatigue Science Readiband, Fitbit Alta HR, Garmin Fenix 5S, Garmin Vivosmart 3, ResMed S+ and SleepScore Max

One study found seven devices detected sleep epochs with sensitivity of 0.93 or higher and wake epochs with specificity of 0.18 to 0.54.

The numbers and sources

Device: Four wearables (Fatigue Science Readiband, Fitbit Alta HR, Garmin Fenix 5S, Garmin Vivosmart 3) and three nonwearable devices (EarlySense Live, ResMed S+, SleepScore Max)
Compared with: Polysomnography, three consecutive laboratory nights, with a research actigraph (Philips Respironics Actiwatch 2) worn alongside
Effect: Epoch sensitivity for sleep was 0.93 or higher for every device, and specificity for wake ran from 0.18 to 0.54. Bias in total sleep time was -0.3 minutes for ResMed S+, 2.6 for Fitbit Alta HR, 7.5 for SleepScore Max, 13.3 for Fatigue Science Readiband, 13.6 for EarlySense Live, 43.7 for Garmin Fenix 5S and 46.8 for Garmin Vivosmart 3, against 23.9 for the research actigraph. Limits of agreement for Fitbit Alta HR ran from -42.0 to 47.1 minutes and for Garmin Fenix 5S from -47.0 to 134.4 minutes. Sleep stage comparisons were mixed, and devices tended to perform less closely on nights with disrupted sleep [18].
Population: 34 healthy young adults (22 women, mean age 28.1), three nights including one with disrupted sleep [18].
Sources: [18]
Caveats: Funded by the US Office of Naval Research, with no funding or involvement from device companies. Each participant wore a subset of devices, so nights per device ranged from 29 to 51. Healthy young adults only [18].

Limited

Sleep stages, Amazon Halo Rise, Apple Watch 8, Fitbit Sense 2, Galaxy Watch 5, Google Nest Hub 2, Google Pixel Watch, Oura Ring 3, Withings Sleep Tracking Mat and three phone apps

One study found agreement with polysomnography on sleep stages ranged from a macro F1 score of 0.26 to 0.69 across 11 trackers.

The numbers and sources

Device: Google Pixel Watch, Galaxy Watch 5, Fitbit Sense 2, Apple Watch 8, Oura Ring 3, Withings Sleep Tracking Mat, Google Nest Hub 2, Amazon Halo Rise, and the phone apps SleepRoutine, SleepScore and Pillow
Compared with: Polysomnography, in the laboratory
Effect: Macro F1 scores for epoch-by-epoch sleep stage classification ranged from 0.26 to 0.69. The SleepRoutine app scored 0.69 and Amazon Halo Rise 0.62. Cohen's kappa was moderate (0.4 to 0.6) for Google Pixel Watch, Galaxy Watch 5, Fitbit Sense 2, Amazon Halo Rise and SleepRoutine, fair (0.2 to 0.4) for Apple Watch 8, Oura Ring 3, Withings Sleep Tracking Mat and SleepScore, and slight for Google Nest Hub 2 and Pillow [20].
Population: 75 participants at a tertiary hospital and a sleep clinic in Korea, 349,114 epochs [20].
Sources: [20]
Caveats: Several authors are affiliated with Asleep, the maker of SleepRoutine, and hold stock or stock options in it. SleepRoutine had the highest score. Participants were recruited at a hospital and a sleep clinic, and scores varied with body mass index, sleep efficiency and apnea severity [20].

Limited

Sleep stages, Apple Watch Series 8, Fitbit Sense 2 and Oura Ring Gen3

Mixed results: one study found Oura stage durations did not differ from polysomnography, while Fitbit and Apple Watch underestimated deep sleep by 15 and 43 minutes.

The numbers and sources

Device: Apple Watch Series 8, Fitbit Sense 2, Oura Ring Gen3
Compared with: Polysomnography, one inpatient night
Effect: The Oura Ring did not differ from polysomnography in wake, light, deep or REM sleep duration. The Fitbit overestimated light sleep by 18 minutes and underestimated deep sleep by 15 minutes. The Apple Watch overestimated light sleep by 45 minutes and underestimated deep sleep by 43 minutes and wake by 7 minutes. Sensitivity for individual stages was 76.0 to 79.5% for Oura, 61.7 to 78.0% for Fitbit and 50.5 to 86.1% for Apple Watch. Sensitivity for sleep versus wake was 95% or higher for all three [3].
Population: 35 adults aged 20 to 50 without a sleep disorder, one night [3].
Sources: [3]
Caveats: Funded by Oura Ring Inc., and the lead author sits on Oura's medical advisory board. Intraclass correlations for stage durations were 0.13 to 0.52 across the three devices. Healthy sleepers only [3].

Did not match the reference method7

Strong

Total sleep time, consumer wrist-worn devices pooled

Not shown to match polysomnography for total sleep time in pooled studies, differing by about 17 minutes on average, with wearables tending to overestimate sleep.

The numbers and sources

Device: Pooled across studies of Fitbit, Jawbone, myCadian watch, WHOOP strap, Garmin, Basis B1, Zulu Watch, Huami Arc, E4 wristband, Fatigue Science Readiband, Apple Watch and Xiaomi Mi Band 5
Compared with: Polysomnography
Effect: A meta-analysis of 24 studies found wrist-worn trackers differed from polysomnography on total sleep time (mean difference -16.9 minutes, 95% CI -26.3 to -7.4), sleep efficiency (-4.7%, 95% CI -7.1 to -2.3), sleep latency (2.6 minutes, 95% CI 0.6 to 4.5) and wake after sleep onset (13.3 minutes, 95% CI 4.5 to 22.0) [4]. A second meta-analysis found wearables overestimated total sleep time and sleep efficiency and underestimated wake after sleep onset, with substantial variability between devices [5].
Population: 798 participants in 24 studies [4]. Healthy adults in 16 studies with simultaneous wearable and laboratory polysomnography recordings, 10 of them pooled [5].
Sources: [4], [5]
Caveats: The abstract of the first review does not state which direction its mean differences run [4]. Pooled figures mix device models and generations. The second review reported methodological heterogeneity and risk of bias, and found no device performed consistently above the others [5].

Limited

Total sleep time, EMFIT QS

One study found the under-mattress EMFIT QS overestimated total sleep time by 178 minutes in sleep clinic patients, with data lost for 47% of participants.

The numbers and sources

Device: EMFIT QS (Quantified Sleep), an under-mattress ballistocardiography sensor
Compared with: Polysomnography, one clinical night
Effect: EMFIT QS overestimated total sleep time by 177.5 minutes (SD 119.4, p < .001) and underestimated wake after sleep onset by 44.7 minutes (SD 68.8). Agreement on sleep versus wake epochs was low (kappa 0.13), and the device did not distinguish sleep stages. Sleep onset latency and average resting heart rate agreed with polysomnography [6].
Population: 62 sleep laboratory patients, of whom 33 had usable data (15 women, mean age 53.7) [6].
Sources: [6]
Caveats: Sleep clinic patients on a single night. Data loss occurred in 47% of participants. The abstract does not report the firmware version [6].

Limited

Total sleep time, Fitbit Sense 2, Oura Ring Gen3, SleepScore Max and Withings Sleep Mat in older adults

One study found three of four devices underestimated total sleep time by 57 to 76 minutes in 19 adults aged 56 to 80.

The numbers and sources

Device: Fitbit Sense 2 (version 194.61), Oura Ring 3rd generation (version 2.8.41), SleepScore Max (version 2.1.0.1), Withings Sleep Mat (version 2481)
Compared with: Polysomnography, one night
Effect: In older adults, bias in total sleep time was -74.5 minutes for Fitbit (p = .012), -75.5 for Oura (p < .0001), -56.5 for SleepScore Max (p = .001) and -45.7 for the Withings mat (p = .083). Deep sleep was overestimated by 71.5 minutes on Oura, 97.4 on Withings and 88.8 on SleepScore Max. Limits of agreement were generally wider in older adults than in young adults across all measures [7].
Population: 19 adults aged 56 to 80 and 13 adults aged 19 to 24, one night each [7].
Sources: [7]
Caveats: Small groups and a single night. The Withings bias did not reach statistical significance. The authors reported no financial disclosures [7].

Limited

Total sleep time, Withings Sleep Analyzer

One study found the under-mattress Withings Sleep Analyzer overestimated total sleep time by 48 minutes on average in 183 people with and without sleep disorders.

The numbers and sources

Device: Withings Sleep Analyzer, an under-mattress sensor
Compared with: Polysomnography, in the laboratory
Effect: The sensor overestimated total sleep time by 48 minutes (SD 81), sleep efficiency by 9% (SD 15) and sleep onset latency by 6 minutes (SD 26), and underestimated wake after sleep onset by 54 minutes (SD 78). Epoch accuracy for sleep versus wake was 83%, with sensitivity 95% and specificity 37%. Accuracy was 89% in healthy individuals and 81% in those with sleep disorders [16].
Population: 183 participants (51% male, mean age 45) across several laboratory studies and more than 400 recordings, including people with sleep disorders and daytime sleep opportunities [16].
Sources: [16]
Caveats: Two authors have held an investigator-initiated research grant supported by Withings. The study was supported by Australia's Defence Science and Technology Group. The authors note a tendency to overestimate sleep where wake time is high [16]. In a separate study of older adults, the Withings Sleep Mat bias was -45.7 minutes and not statistically significant [7].

Limited

Sleep stages, Apple Watch Series 7, Fitbit Charge 5 and Polar Vantage M2

One study found weak agreement with polysomnography on sleep stage durations for three wrist devices, with light sleep off by 58 to 74 minutes on two of them.

The numbers and sources

Device: Apple Watch Series 7, Fitbit Charge 5, Polar Vantage M2
Compared with: Home polysomnography, one night
Effect: Fitbit Charge 5 overestimated light sleep by 57.7 minutes (SD 80.5, P = .009). Apple Watch Series 7 underestimated light sleep by 73.8 minutes (SD 61.6, P < .001) and overestimated REM sleep by 30.3 minutes (SD 36.3, P = .003). Polar Vantage M2 overestimated REM sleep by 32.3 minutes (SD 43.6, P = .026). The authors describe agreement with polysomnography as generally weak, with wide limits of agreement [21].
Population: Healthy adults aged 18 to 45, each assigned one device: 20 Fitbit, 19 Apple Watch, 15 Polar [21].
Sources: [21]
Caveats: Each participant wore only one device, so the three were not tested on the same sleepers. Groups of 15 to 20. The abstract does not report firmware versions [21].

Limited

Sleep stages, Circul+ Ring, Oura Ring Gen 3 and SleepOn Go2Sleep in sleep clinic patients

One study found three rings classified sleep stages with 35 to 53% accuracy in 45 sleep clinic patients.

The numbers and sources

Device: Oura Ring Gen 3, SleepOn Go2Sleep, Circul+ Ring
Compared with: Polysomnography, in a university sleep laboratory
Effect: Accuracy across wake, light, deep and REM sleep was 53.2% for Oura, 50.5% for SleepOn and 35.1% for Circul. Stage sensitivity ranged from 0.14 (REM on Circul) to 0.58 (light sleep on SleepOn). Sleep versus wake accuracy was about 85% for Oura and SleepOn and about 65% for Circul. Total sleep time was overestimated by 11.7 minutes on Oura and 50.5 minutes on SleepOn and underestimated by 31.5 minutes on Circul [11].
Population: 45 patients (20 women) aged 23 to 74, mean body mass index 33.5, with sleep apnea, insomnia and other disorders. Usable nights were 31 for Oura, 35 for SleepOn and 31 for Circul [11].
Sources: [11]
Caveats: A patient group with sleep disorders, where trackers have been studied less. The authors conclude that agreement on averages masks large differences for individual nights. They declared no competing interests [11].

Moderate

Sleep stages, finger-worn devices pooled

Not shown to match polysomnography for sleep stages, with pooled accuracy of 65% for light sleep, 81% for deep sleep and 74% for REM sleep.

The numbers and sources

Device: 11 finger-worn devices pooled across 28 articles, including Oura Ring generations
Compared with: Polysomnography
Effect: Pooled accuracy was 0.65 for light sleep, 0.81 for deep sleep and 0.74 for REM sleep, against 0.87 for sleep versus wake. The authors concluded finger-worn devices do not replace polysomnography for detailed sleep architecture [19].
Population: 28 articles published up to April 2025, with sleep staging studied primarily in healthy participants [19].
Sources: [19]
Caveats: Devices used different sensors and algorithms, and the pooled figures combine device generations [19].

Does using one change anything

Shown to help2

Limited

Insomnia symptoms with tracker feedback and guidance

One trial found feedback on tracker data with guidance lowered insomnia severity more than sleep education in 113 adults with insomnia symptoms over 5 weeks.

The numbers and sources

Effect: Compared with sleep education and hygiene, the group given feedback on sensor-based sleep data with guidance on interpreting it had lower Insomnia Severity Index scores (p = .011, d = 0.51) and lower sleep disturbance (p = .036, d = 0.42) after the intervention. Sleep-related impairment, depression, anxiety and the gap between self-reported and sensor-measured sleep did not differ between groups (P > .40) [22].
Population: 113 adults with Insomnia Severity Index scores of 10 or higher (mean age 47.5, 64.9% female), 5 weeks. 103 completed [22].
Sources: [22]
Caveats: The intervention combined a Fitbit and an EEG headband with an individual session and two check-in calls, so the trial does not isolate the tracker from the guidance. Single-blind, with self-reported outcomes [22].

Limited

Sleep quality in healthy adults

One small trial found a week of wearing a tracker with feedback lowered self-reported sleep disturbance in 32 healthy adults, with total daily sleep unchanged.

The numbers and sources

Effect: In a randomized crossover trial, one week of wearing a WHOOP device that gave sleep feedback, plus sleep logs, was compared with one week of sleep logs alone. PROMIS sleep disturbance scores were lower with the device (B = -1.69, 95% CI -3.11 to -0.27, P = .021). Self-reported daytime naps rose slightly (B = 3.2, SE 1.4, P = .023), and total daily sleep did not change (P = .43) [14].
Population: 32 healthy adults (21 women, mean age 23.8) without medical conditions or sleep disorders [14].
Sources: [14]
Caveats: One week per condition, a self-reported outcome and no blinding. Young healthy volunteers at one center [14].

Risks

Reported risks2

Limited

Preoccupation with sleep data (orthosomnia)

Case reports: patients sought treatment for sleep problems they inferred from tracker data. One survey classified 3 to 14% of respondents as having orthosomnia.

The numbers and sources

Effect: A 2017 case report described patients seeking treatment for self-diagnosed sleep disturbance based on tracker data, and named the pattern orthosomnia, a perfectionistic pursuit of ideal sleep. The patients often trusted tracker data over polysomnography or actigraphy [23]. In a cross-sectional survey of 523 people, 176 regularly used a sleep tracker, and an algorithm written by the authors classified 3.0% (95% CI 1.6 to 4.5), 8.6% (95% CI 6.2 to 11.0) or 14.0% (95% CI 10.9 to 16.9) of the sample as having orthosomnia under conservative, moderate or lenient cutoffs. Those classified had higher insomnia scores [24].
Population: Patients seeking treatment for sleep complaints [23]. 523 adults in a general population survey [24].
Sources: [23], [24]
Caveats: Case reports cannot give a rate. The survey was cross-sectional and required insomnia symptoms as part of its definition, so it cannot show that tracker use causes the pattern. The classification algorithm was devised by the survey's authors [24].

Limited

Daytime symptoms after negative sleep feedback

One trial found people with insomnia given sham negative sleep feedback reported more daytime sleepiness and less alertness than those given sham positive feedback.

The numbers and sources

Effect: Participants were randomized to receive sham feedback that their sleep efficiency had been high or low. Those given negative feedback reported lower alert cognition (d = 0.79) and more sleepiness and fatigue (d = 0.55) in the evening than those given positive feedback. Tests of sleep-related attentional bias (d = 0.20) and psychomotor vigilance (d = 0.12) did not differ [25].
Population: 63 adults meeting DSM-5 criteria for insomnia disorder, one day [25].
Sources: [25]
Caveats: The feedback was sham and delivered through a research actigraphy watch made to simulate a wearable, so the trial did not test a consumer tracker. One day, in people with insomnia only [25].

Sources

US Food and Drug Administration. General Wellness: Policy for Low Risk Devices. Guidance for Industry and Food and Drug Administration Staff. January 6, 2026. https://www.fda.gov/regulatory-information/search-fda-guidance-documents/general-wellness-policy-low-risk-devices
Schyvens AM, Peters B, Van Oost NC, et al. A performance validation of six commercial wrist-worn wearable sleep-tracking devices for sleep stage scoring compared to polysomnography. Sleep Advances. 2025. https://doi.org/10.1093/sleepadvances/zpaf021
Robbins R, Weaver MD, Sullivan JP, et al. Accuracy of Three Commercial Wearable Devices for Sleep Tracking in Healthy Adults. Sensors. 2024. https://doi.org/10.3390/s24206532
Lee YJ, Lee JY, Cho JH, et al. Performance of consumer wrist-worn sleep tracking devices compared to polysomnography: a meta-analysis. Journal of Clinical Sleep Medicine. 2025. https://doi.org/10.5664/jcsm.11460
Agostinho M, Borges M, Pereira T, et al. Are Wearable Sleep-Tracking Devices Reliable Alternatives to Polysomnography? A Systematic Review and Meta-Analysis. Behavioral Sleep Medicine. 2026. https://doi.org/10.1080/15402002.2026.2673893
Kholghi M, Szollosi I, Hollamby M, et al. A validation study of a ballistocardiograph sleep tracker against polysomnography. Journal of Clinical Sleep Medicine. 2022. https://doi.org/10.5664/jcsm.9754
Searles ME, Licata A, Cucinotta M, et al. Performance evaluation of consumer sleep-tracking wearables and nearables in healthy young and older adults. Sleep Advances. 2026. https://doi.org/10.1093/sleepadvances/zpag006
Haghayegh S, Khoshnevis S, Smolensky MH, et al. Accuracy of Wristband Fitbit Models in Assessing Sleep: Systematic Review and Meta-Analysis. Journal of Medical Internet Research. 2019. https://doi.org/10.2196/16273
Khan S, Ibrahim AF, Vasudevan SS, et al. The Oura Ring Versus Medical-Grade Sleep Studies: A Systematic Review and Meta-Analysis. OTO Open. 2025. https://doi.org/10.1002/oto2.70181
Svensson T, Madhawa K, NT H, et al. Validity and reliability of the Oura Ring Generation 3 (Gen3) with Oura sleep staging algorithm 2.0 (OSSA 2.0) when compared to multi-night ambulatory polysomnography: A validation study of 96 participants and 421,045 epochs. Sleep Medicine. 2024. https://doi.org/10.1016/j.sleep.2024.01.020
Herberger S, Aurnhammer C, Bauerfeind S, et al. Performance of wearable finger ring trackers for diagnostic sleep measurement in the clinical context. Scientific Reports. 2025. https://doi.org/10.1038/s41598-025-93774-z
de Zambotti M, Rosas L, Colrain IM, Baker FC. The Sleep of the Ring: Comparison of the ŌURA Sleep Tracker Against Polysomnography. Behavioral Sleep Medicine. 2019. https://doi.org/10.1080/15402002.2017.1300587
Miller DJ, Lastella M, Scanlan AT, et al. A validation study of the WHOOP strap against polysomnography to assess sleep. Journal of Sports Sciences. 2020. https://doi.org/10.1080/02640414.2020.1797448
Berryhill S, Morton CJ, Dean A, et al. Effect of wearables on sleep in healthy individuals: a randomized crossover trial and validation study. Journal of Clinical Sleep Medicine. 2020. https://doi.org/10.5664/jcsm.8356
Schyvens AM, Van Oost NC, Aerts JM, et al. Accuracy of Fitbit Charge 4, Garmin Vivosmart 4, and WHOOP Versus Polysomnography: Systematic Review. JMIR mHealth and uHealth. 2024. https://doi.org/10.2196/52192
Manners J, Kemps E, Lechat B, et al. Performance evaluation of an under-mattress sleep sensor versus polysomnography in > 400 nights with healthy and unhealthy sleep. Journal of Sleep Research. 2025. https://doi.org/10.1111/jsr.14480
Miller DJ, Sargent C, Roach GD. A Validation of Six Wearable Devices for Estimating Sleep, Heart Rate and Heart Rate Variability in Healthy Adults. Sensors. 2022. https://doi.org/10.3390/s22166317
Chinoy ED, Cuellar JA, Huwa KE, et al. Performance of seven consumer sleep-tracking devices compared with polysomnography. Sleep. 2021. https://doi.org/10.1093/sleep/zsaa291
Jin Y, Xu J, Yue H, et al. Performance evaluation of finger-worn devices for sleep stage classification and sleep apnea detection: a systematic review and meta-analysis. Journal of Translational Medicine. 2026. https://doi.org/10.1186/s12967-026-08214-x
Lee T, Cho Y, Cha KS, et al. Accuracy of 11 Wearable, Nearable, and Airable Consumer Sleep Trackers: Prospective Multicenter Validation Study. JMIR mHealth and uHealth. 2023. https://doi.org/10.2196/50983
Alhejaili F, Alshumrani R, Alamoudi W, et al. Performance of three consumer sleep-tracking devices compared with Actigraphy and Polysomnography. Sleep Medicine. 2026. https://doi.org/10.1016/j.sleep.2026.109059
Spina MA, Andrillon T, Quin N, et al. Does providing feedback and guidance on sleep perceptions using sleep wearables improve insomnia? Findings from "Novel Insomnia Treatment Experiment": a randomized controlled trial. Sleep. 2023. https://doi.org/10.1093/sleep/zsad167
Baron KG, Abbott S, Jao N, et al. Orthosomnia: Are Some Patients Taking the Quantified Self Too Far? Journal of Clinical Sleep Medicine. 2017. https://doi.org/10.5664/jcsm.6472
Jahrami H, Trabelsi K, Husain W, et al. Prevalence of Orthosomnia in a General Population Sample: A Cross-Sectional Study. Brain Sciences. 2024. https://doi.org/10.3390/brainsci14111123
Gavriloff D, Sheaves B, Juss A, et al. Sham sleep feedback delivered via actigraphy biases daytime symptom reports in people with insomnia: Implications for insomnia disorder and wearable devices. Journal of Sleep Research. 2018. https://doi.org/10.1111/jsr.12726

Everything in the library