Sleep trackers went mainstream. The sleep-stage evidence did not follow.

The sleep tracker has quietly become one of the most widely consulted pieces of health technology in American life. In a national survey of 1,280 US adults presented at the SLEEP 2026 meeting, 32.4% reported tracking their sleep with a digital device — 415 people out of the sample [s4]. That is roughly the same share of the population that owns a dog.

What those 415 people are looking at each morning is a number the industry has never been required to prove.

The gap between the pitch and the measurement

Consumer sleep trackers do not measure sleep the way a sleep laboratory does. Polysomnography — the clinical reference standard — reads brain activity directly through EEG electrodes, along with eye movement, muscle tone, airflow, and cardiac signals. A wrist-worn tracker or a ring infers all of that from movement, pulse waveform, and skin temperature, then runs the result through a proprietary model to produce a hypnogram that looks, on screen, exactly as authoritative as the clinical one.

A validation study published in SLEEP Advances in March 2025 put six of those devices against simultaneous polysomnography in 62 adults — 52 men and 10 women, mean age 46.0 ± 12.6 years — and scored them epoch by epoch [s3]. The pattern it found is the one worth carrying around.

The devices are good at the easy question. Sensitivity for detecting sleep ranged from 91.68% to 96.27% across the six trackers [s3]. If you were asleep, the device generally knew it.

They are poor at the hard question. Specificity — correctly identifying the epochs when a person was actually awake — ranged from 29.39% to 52.15% [s3]. At the low end, a device is missing roughly seven of every ten awake moments during the night. That is precisely the failure mode that matters to someone with insomnia, whose complaint is not that they cannot fall asleep but that they cannot stay asleep.

The consequence shows up in total sleep time. Five of the six devices overestimated it by a statistically significant margin: Whoop 4.0 by 24.46 minutes (p=0.010), Apple Watch Series 8 by 19.60 minutes (p=0.004), Garmin Vivosmart 4 by 38.44 minutes (p=0.002), and Withings ScanWatch by 39.87 minutes (p<0.001) [s3]. Only Fitbit Sense (+6.31 minutes, p=0.285) and Fitbit Charge 5 (+11.12 minutes, p=0.051) did not differ significantly from polysomnography [s3].

On staging agreement, measured as Cohen's kappa against the reference scoring, the six devices ran from 0.53 for the Apple Watch Series 8 down to 0.21 for the Garmin Vivosmart 4, with Fitbit Sense at 0.42, Fitbit Charge 5 at 0.41, Whoop 4.0 at 0.37, and Withings ScanWatch at 0.22 [s3]. None of these are numbers a clinical instrument would ship with.

Two limits on reading too much into this single study: it tested six specific device generations, several of which have since been superseded, and its sample skewed heavily male. Device firmware and staging algorithms change without notice, which is itself part of the problem — a validation published today describes a product that may no longer exist in the same form.

The second-order effect

The same national survey that counted the trackers also counted the anxiety. Among those who tracked their sleep, 30.9% screened positive for orthosomnia risk on a two-item measure adapted from a validated anxiety screener, which asked whether respondents felt "nervous, anxious or on edge" about their sleep tracker data or could not stop worrying about it [s4]. Scores of 3 or higher on the 0–3 scale marked positive risk. The authors report that this is approximately twice the rate found in prior estimates [s4].

Orthosomnia — preoccupation with sleep data to the point that it degrades sleep — has an obvious circularity to it. A person worried about their sleep checks a score. The score is generated by a model that systematically overstates how long they slept and understates how often they woke. Either direction of error can feed the worry: a bad score confirms the fear, and a good score that contradicts how the person actually feels undermines trust in their own experience.

This was a conference abstract, not a peer-reviewed full paper, and the orthosomnia measure was an adapted two-item screen rather than a diagnostic instrument. The prevalence figure should be read as a signal, not a settled number.

What the position statement actually says

In March 2026, the National Sleep Foundation published a position statement on consumer sleep technology in its journal Sleep Health, the field's most direct attempt yet to say where these devices belong [s1]. According to the University of Miami Miller School of Medicine, whose faculty member Azizi Seixas — Associate Professor of Psychiatry and Behavioral Sciences and Director of the school's Media and Innovation Lab — helped shape the statement, its core positions are that the technology can help users see patterns and stay engaged with sleep health, that results vary meaningfully by device, that the tools should support rather than replace basic sleep behaviors, and that device data should not substitute for clinical evaluation when a sleep disorder is suspected [s2]. The statement also flags data governance and privacy as unresolved, and names obsessive pursuit of perfect metrics as a recognized harm [s2].

"Sleep technology has made sleep visible and visibility is the first step to better health and sleep health," Seixas said [s2].

Visibility is not the same as accuracy, and the statement's framing acknowledges the difference.

The one place the line is drawn

There is a regulatory boundary here, and it is narrower than most users assume. Sleep-stage estimates are sold as general wellness features and are not reviewed by the FDA. A small number of specific screening functions are.

In February 2024, the FDA granted De Novo authorization to Samsung's Galaxy Watch sleep apnea feature — the first of its kind — to detect signs of moderate to severe obstructive sleep apnea in adults 22 and older [s5]. The authorization is tightly scoped. The feature requires the user to track sleep twice for more than four hours within a 10-day window, and it is explicitly not intended to replace diagnosis and treatment by a clinician, not intended to help clinicians diagnose sleep disorders, and not intended for people already diagnosed with obstructive sleep apnea [s5].

That is what a cleared claim looks like: one condition, one population, one narrow question, with the limits written into the authorization. Everything else on the same screen — the sleep score, the stage breakdown, the readiness rating — carries no such review.

What to watch

Three things. Whether device makers publish independent validation against polysomnography for current firmware rather than superseded generations. Whether the orthosomnia prevalence signal survives a full peer-reviewed study with a diagnostic instrument rather than a two-item screen. And whether the boundary between cleared screening features and unreviewed wellness scores stays legible to users as more functions cross it.

The trackers are not useless. Night-to-night trends in bedtime, wake time, and duration are the kind of pattern they capture reasonably well, and that pattern is where most sleep improvement actually lives. The stage pie chart is the part the evidence does not yet support.

Sources

• [s1] Consumer sleep technology: A position statement from the National Sleep Foundation, Sleep Health: Journal of the National Sleep Foundation, published online 2026-03-05.
• [s2] Consumer Sleep Technology Can Support Better Sleep, With Limits, University of Miami Miller School of Medicine (InventUM), 2026-03-25.
• [s3] Schyvens et al., A performance validation of six commercial wrist-worn wearable sleep-tracking devices for sleep stage scoring compared to polysomnography, SLEEP Advances, 2025-03-22.
• [s4] Goel et al., 0553 Exploring the Prevalence of Sleep Tracking and Orthosomnia in a National Survey of Adults in the US, SLEEP, Vol. 49, Supplement 1, May 2026, p. A246.
• [s5] Samsung Galaxy Watch sleep apnea feature receives FDA authorization, American Academy of Sleep Medicine, 2024-02-12.