Sleep trackers: how accurate they are, what they miss and what sleep doctors say about them

A watch, a ring or a pad under the mattress now hands its owner a score every morning: seven hours twelve minutes, 81 out of 100, an hour and a half of deep sleep. The figures look precise. Whether they are correct is a separate question, and it is one that sleep laboratories have been testing for several years.

That research shows trackers to be better than their critics assume and worse than their apps imply. They are good at telling that someone is asleep. They are poor at noticing when someone is lying awake, and their breakdown of the night into stages is unreliable. Professional guidance is blunt about the limit: the American Academy of Sleep Medicine holds that these devices cannot be utilized for the diagnosis and/or treatment of sleep disorders at this time.

How is a sleep tracker’s accuracy tested?

The reference is polysomnography, the overnight recording used in sleep clinics. Sensors on the scalp, face and body track brain waves, eye movements, muscle tone and breathing, and a trained scorer labels each short slice of the night as wake or as a particular stage of sleep. A tracker is tested by wearing it during such a recording and comparing the two, slice by slice.

The Naval Health Research Center in San Diego ran one of the most thorough comparisons. Evan Chinoy and colleagues brought in 34 healthy young adults for three consecutive nights, one of which was deliberately disturbed. Each volunteer was wired for polysomnography while using a selection of seven consumer devices: four worn on the wrist and three that sat on the bed or the nightstand. A research-grade wrist monitor called an actigraph, the tool sleep scientists have used for decades outside the laboratory, was included for comparison.

What do sleep trackers get right, and what do they miss?

On the simplest task the devices did well. Consumer sleep-tracking devices exhibited high performance in detecting sleep, correctly labeling at least 93% of the slices in which the volunteer really was asleep. Most of them matched or beat the research actigraph.

The weakness was on the other side. When the volunteer was awake in bed, the devices recognized it between 18% and 54% of the time, depending on the model. Lying still and awake looks like sleep to a motion and pulse sensor. The result is a flattering total: three of the seven devices overstated total sleep time against the laboratory recording.

That error grows exactly where it matters. The study found that devices tended to perform worse on nights with poorer/disrupted sleep. Someone who buys a tracker because they suspect they sleep badly is the user for whom it is least accurate.

Stages were the weakest area. The authors’ summary is short: device sleep stage assessments were inconsistent. All of the devices that reported stages misjudged light sleep, most by overestimating it.

A larger and newer study points the same way. Researchers in South Korea put 11 current products, including recent smartwatches and a ring, through the same kind of comparison in 75 people. For stage-by-stage agreement, the highest score any device reached was 0.69 on a scale where 1 is a perfect match, while the lowest macro F1 score was 0.26. That score is a standard measure of how well labels agree. Which device did best depended on the stage being measured.

Experts who follow the field take a measured view. A 2024 review led by Massimiliano de Zambotti of SRI International in California concluded that the available evidence suggests that consumer-grade devices exceed the performance of traditional actigraphy. It then listed the persistent problems, beginning with the misclassification of wakefulness during the sleep period, and warned that physical characteristics such as skin color or obesity can interfere with the sensors.

Why can two sleep trackers disagree about the same night?

Each manufacturer turns raw sensor signals into sleep estimates with its own software, and that software is secret and changes. An earlier review from the same California group named the factors researchers have to contend with: proprietary algorithms, device malfunction, firmware updates. A device validated in one year may behave differently after an update the next.

Testing methods have been uneven too. In proposing a common protocol, the group wrote that because of the lack of a standardized framework to evaluate the performance of sleep trackers, their accuracy and reliability in measuring sleep remains largely unknown. A published accuracy figure applies to one model, running one version of its software, in the kind of people who were tested.

What do sleep medicine bodies say about trackers?

The American Academy of Sleep Medicine issued a position statement in 2018 that still frames clinical practice. It accepts that the devices are everywhere and that patients bring their data to appointments. It also draws a line. Consumer products lack validation and regulatory clearance as medical devices, so they are not a basis for diagnosing or treating a sleep disorder.

The statement is not dismissive. It says such technologies may be utilized to enhance the patient-clinician interaction when presented in the context of an appropriate clinical evaluation. A chart showing weeks of irregular bedtimes can be a useful starting point for a conversation, even if the nightly totals are off.

Can tracking sleep make sleep worse?

Clinicians have described a group of patients for whom it seems to. In 2017, Kelly Glazer Baron and colleagues at Rush University Medical Center in Chicago reported a series of cases and gave the pattern a name, orthosomnia. These were people seeking treatment for sleep problems they had diagnosed from their own tracker data. The pursuit of better numbers, the authors wrote, may become a perfectionistic quest for the ideal sleep.

One detail from the cases is telling. To the patients, sleep tracker data often feels more consistent with their experience of sleep than validated techniques such as polysomnography. A laboratory result showing normal sleep did not always persuade them. The report describes a handful of patients, so it shows that the problem exists without saying how common it is.

How many people use sleep trackers?

Fewer than the marketing suggests. A National Sleep Foundation survey published in 2026 questioned random samples of about 1,000 adults each in the United States and South Korea. Majorities in both the U.S. (86.1%) and South Korea (91.2%) did not use sleep tracking technologies.

Attitudes differed sharply between the two. A majority of Americans expressed concerns about misuse of collected data from sleep tracking devices, and roughly two-thirds doubted that technology would help their sleep. Most South Koreans were unworried about data and optimistic about benefits.

Healthy young volunteers and fast-changing devices: where tracker testing falls short

The San Diego volunteers were aged 18 to 35, not obese and free of any sleep disorder. That is the standard first step for a validation study and a poor match for the people most interested in their sleep. The authors themselves called for testing in other populations and settings.

Products also move faster than research. The seven devices were current when the study was run, and their successors use different sensors and software. No study has yet shown that owning a tracker improves sleep or health over the long run. Persistent trouble sleeping is assessed by a clinician using validated methods, with tracker data as background at most.

A sleep tracker gives a reasonable estimate of when someone was asleep, tends to overstate how long, and guesses at the stages.