At least once a week, a patient shows me their Apple Watch or Oura Ring sleep data and asks me to interpret it. The data looks impressive — colourful graphs showing sleep stages, sleep scores, and trends over time. The question they really want answered is: can I trust this?

The honest answer is complicated.

What Clinical Sleep Studies Measure

A polysomnography (PSG) — the gold standard clinical sleep study — uses electrodes placed directly on the scalp to record brain electrical activity (EEG), sensors near the eyes to track eye movements (EOG), and sensors on the chin and legs for muscle activity (EMG). It also records airflow, blood oxygen, chest movement, heart rhythm, body position, and snoring.

Sleep staging is determined from the EEG. Different sleep stages produce distinct brain wave patterns: the alpha waves of relaxed wakefulness, the theta waves and sleep spindles of N2, the slow delta waves of N3, and the mixed-frequency low-amplitude patterns with rapid eye movements of REM sleep. A trained scientist reviews the recording in 30-second segments and classifies each one.

This is direct measurement. You’re recording the brain’s electrical output and using established criteria to determine what’s happening.

What Wearables Measure

Consumer wearables don’t have EEG sensors. Instead, they primarily use:

Accelerometry — a motion sensor that detects movement. The assumption is that you move less during deeper sleep, which is broadly true but far from precise.

Photoplethysmography (PPG) — an optical heart rate sensor. Heart rate and heart rate variability change across sleep stages. REM sleep tends to have a higher, more variable heart rate than deep sleep.

Some newer devices add additional sensors. The Oura Ring measures skin temperature, which fluctuates across sleep stages. A few research-grade wearables include single-channel EEG, though these haven’t hit the mainstream consumer market in a reliable form.

From these indirect signals, the device’s algorithm estimates sleep stages. It’s an inference, not a direct measurement.

How Accurate Are They?

This depends on what you’re measuring and which device you’re using. Independent validation studies — where people wear both the consumer device and full PSG simultaneously — tell a mixed story.

Total sleep time: Most modern wearables estimate total sleep time within 20-30 minutes of PSG measurements. That’s reasonably good for a consumer device. However, they tend to overestimate total sleep time because they can’t reliably distinguish quiet wakefulness (lying still in bed with your eyes closed) from light sleep.

Sleep efficiency: Related to the above, wearables typically overestimate sleep efficiency (the percentage of time in bed that you’re actually asleep) because they miss periods of quiet wakefulness.

Deep sleep (N3): Accuracy is moderate. A 2023 study published in Sleep found that the Apple Watch and Oura Ring both identified N3 periods with roughly 60-70% accuracy compared to PSG. That’s better than chance but far from clinical grade.

REM sleep: Similar accuracy range — 60-70% agreement with PSG for the better devices. REM detection benefits from the heart rate variability signal, which is a reasonably reliable marker.

Light sleep (N1 and N2): Poor discrimination. Most wearables lump N1 and N2 together, and their accuracy for detecting transitions between light sleep and wakefulness is low.

Where Wearables Are Useful

Despite their limitations, wearables serve a legitimate purpose for certain use cases.

Long-term trends. A single night’s data from a wearable is unreliable. But weeks or months of data can reveal meaningful patterns. If your sleep tracker consistently shows you’re getting less sleep on workdays vs weekends, or that your sleep quality degrades during stressful periods, those trends are probably real even if the absolute numbers aren’t precise.

Sleep schedule awareness. Many people don’t realise how irregular their sleep schedule is until they see the data. A wearable that shows your bedtime varying by two hours from night to night is providing useful information, regardless of how accurately it classifies sleep stages.

Motivation and accountability. Some people sleep better simply because they’re paying attention to their sleep. If wearing a tracker motivates you to maintain a consistent schedule and prioritise sleep duration, the behavioural change matters more than the data accuracy.

Screening (with caveats). Some wearables now include pulse oximetry that can detect patterns suggestive of sleep apnea. This isn’t diagnostic, but it might prompt someone to seek formal evaluation who otherwise wouldn’t have. The Australian Sleep Health Foundation acknowledges the potential screening role of consumer devices while emphasising they don’t replace clinical assessment.

Where Wearables Fall Short

Diagnosing sleep disorders. No consumer wearable can diagnose sleep apnea, narcolepsy, periodic limb movement disorder, or any other clinical sleep condition. They don’t have the sensors, and their algorithms aren’t validated for clinical diagnosis. Using wearable data to self-diagnose is unreliable and potentially harmful if it delays proper evaluation.

Guiding treatment decisions. If you’re on CPAP therapy, your CPAP machine’s built-in data (AHI, mask leak, usage hours) is far more useful than anything your wearable provides. Treatment decisions should be based on clinical-grade data, not consumer device estimates.

Anxiety-prone individuals. This is an underappreciated problem. Some people become anxious about their sleep data — obsessing over sleep scores, worrying when the tracker shows a “bad” night, and developing what researchers have termed “orthosomnia” (an unhealthy preoccupation with achieving perfect sleep data). For these individuals, wearable tracking actually worsens sleep by increasing performance anxiety around sleep.

If checking your sleep data first thing in the morning makes you feel anxious rather than informed, the tracker is doing more harm than good. Stop wearing it.

The Clinical Perspective

When I see patients, I look at wearable data with interest but treat it as supplementary information, not clinical evidence. If someone’s Oura Ring shows consistently poor sleep quality, I take that seriously as a symptom report — but I don’t use the specific numbers to guide treatment.

For organisations exploring how AI and data tools can support healthcare delivery, working with teams that have genuine AI consulting help can make the difference between useful innovation and expensive mistakes. The gap between consumer-grade and clinical-grade health data is one area where expert guidance matters.

If I need objective sleep data, I order a clinical sleep study. There’s no shortcut for that. The wearable might be what prompts the referral, but it doesn’t replace the test.

The Future

Wearable technology is improving rapidly. EEG-capable headbands, more sophisticated PPG algorithms, and multi-sensor fusion approaches are narrowing the gap between consumer and clinical devices. Within the next few years, we might see wearables that achieve 80-85% sleep staging agreement with PSG, which would make them genuinely useful as clinical screening tools.

But the fundamental limitation remains: without direct brain measurement, sleep staging from a wrist or finger sensor will always be an approximation. The question is whether the approximation is close enough to be clinically useful.

For now, think of your sleep tracker as a rough guide — better than nothing, not as good as a proper assessment. If you’re sleeping well and feel rested, the tracker’s data is interesting but not particularly important. If you’re struggling with sleep, the tracker might flag a problem, but the solution involves seeing a clinician, not tweaking your sleep score.