Wearable sensors explained: what your device measures

A light flashes against your skin, a step total changes, and a dashboard gives you a sleep score. Those numbers begin with different kinds of signals. Sensors collect the signals; software turns them into readings, estimates and interpretations.

Here are six sensor categories you may encounter, how they work and where their limits matter. The combination depends on the device. A watch does not necessarily have every sensor described here, and its measurements may run only during particular activities or recordings.

Optical sensors: light and pulse

Photoplethysmography, usually shortened to PPG, uses light to detect changes associated with blood volume near the skin. LEDs illuminate the skin, and a light-sensitive detector records the returning light. Changes across heartbeats help the device estimate pulse rate. Apple's heart-sensor explanation describes this process at the wrist.

LED colors have different uses. For example, Oura's live-heart-rate guide describes green LEDs for that feature. Its blood-oxygen guide describes red and infrared light used during sleep. Check the supported features and instructions for your own model.

Blood-oxygen sensing estimates peripheral oxygen saturation, or SpO2. It does not establish that every watch or ring has medical authorization for an oxygen-related claim. The FDA's pulse-oximeter information describes accuracy limitations, including concerns related to skin pigmentation. A reading needs context; symptoms still matter.

Contact and movement affect optical signals. Apple's measurement guidance names fit, skin perfusion, cold, some tattoos and movement among the influences on its heart-rate readings. A skin-tone concern should prompt a closer look at evidence for the exact device and intended use, rather than an assumption that one accuracy figure covers everyone.

Electrical sensors: recording an ECG

The heartbeat involves electrical activity. Electrodes touching the skin can record differences in electrical potential over time, producing an electrocardiogram, or ECG. This is a different signal from the light changes used for PPG.

Many consumer ECG features record a single lead, which provides one electrical view of the heart. A clinical 12-lead ECG provides additional views. Apple's ECG instructions describe a recording similar to Lead I, started by the user on a supported device.

A supported ECG app can classify certain rhythms within its intended use, including possible atrial fibrillation. Its result applies to that recording. It cannot detect a heart attack or rule out every heart condition.

Follow the device's instructions for contact and positioning. Movement, moisture and poor contact can impair a recording. A watch that requires you to start an ECG and touch an electrode is not continuously recording that ECG as you go about your day.

Motion sensors: acceleration, rotation and elevation

A wearable usually infers steps from movement patterns. It does not directly observe every foot touching the ground. Several kinds of sensor can contribute:

• An accelerometer measures acceleration along its axes. Software can use those signals to recognize patterns associated with steps or other activities.
• A gyroscope measures rotation. Combining rotational and acceleration signals can help describe movement and orientation.
• An altimeter estimates altitude. A barometric-pressure sensor can contribute to estimates of elevation changes, such as climbing stairs.

These are three different measurements. TDK's description of a combined motion-and-pressure sensor gives stair counting as an application. It does not establish that every tracker contains that component or performs equally well.

The CADENCE-Adults study compared step counts with observed treadmill steps in 258 adults. Accuracy depended on walking speed and wear location. Its devices were tested in 2015–2019, so the results cannot rank today's models or every pattern of daily movement.

Activity labels, calorie estimates and sleep stages can combine motion with other inputs. More sensors do not automatically make every estimate accurate. Placement, the algorithm and the circumstances in which it was validated all matter.

Temperature sensors: changes at the skin

A skin-temperature reading concerns the skin at the measurement site. It is not a direct measurement of core body temperature. Environment, activity and physiology can influence the reading.

For example, Apple's wrist-temperature feature looks at nighttime changes from a baseline established after about five nights. Apple says the feature is not an on-demand thermometer and is not intended for medical diagnosis or treatment.

Temperature patterns can contribute to cycle-tracking features on supported devices. Apple's documentation describes retrospective ovulation estimates and improved period predictions. Those functions have their own requirements; a temperature change alone does not diagnose infection, fever or a hormonal condition.

Read a trend alongside its collection conditions. A loose fit or a different sleep environment can affect what gets recorded. Use an appropriate thermometer when you need a temperature measurement for a medical decision.

Chemical sensors: analyzing sweat

Sweat contains water, electrolytes and other compounds. Chemical sensors can target particular substances in a sample. The measured substance and the method determine what the result means.

In a 2016 research prototype, Gao and colleagues combined sensors for sweat glucose, lactate, sodium and potassium with temperature sensing for calibration. That demonstrates specific sweat-analysis channels in a research system. It does not mean an ordinary watch measures those substances or that a sweat result replaces a blood test.

Hydration and electrolyte-loss estimates need additional context. A local sweat measurement is not automatically a measurement of whole-body fluid balance. Ask where the sample is collected, whether the method needs active sweating, and what population and conditions were used to validate any wider estimate.

Sample handling also matters. A 2026 optical sweat-glucose prototype used manually collected and filtered sweat from three healthy men after cycling. The authors state that the system cannot dynamically monitor blood glucose while worn. A promising research sensor can still be far from the everyday use suggested by an advertisement.

Electrodermal activity: a response with several possible causes

Electrodermal activity, or EDA, concerns changes in the skin's electrical conductance related to sweat-gland activity. It is different from chemically measuring a substance in sweat.

An EDA response can contribute to a device's estimate of physiological arousal. It does not identify a person's thoughts or establish why the response happened. The Google Health guide to EDA explains that movement, noise and temperature can influence responses as well as stress.

Other signals and your own account of the moment help give the reading context. A 2025 physiological-data study gathered structured stress and exercise recordings from healthy volunteers. Results from a defined research dataset do not independently validate a consumer diagnosis of anxiety in daily life.

Combining signals and checking new claims

Movement helps interpret another signal. A raised pulse during a run has a different context from a raised pulse while sitting quietly. Software may combine motion, optical and other data to build an estimate. The interpretation still needs evidence for the particular use.

Researchers are exploring cuffless blood-pressure methods and ways to estimate glucose without piercing the skin. A prototype, a wellness score and an authorized medical feature have different evidentiary requirements.

The FDA warns against unauthorized blood-pressure devices, including watch and ring software claiming to measure or estimate blood pressure. Check the exact device and feature in the FDA's authorization records. The warning does not mean every cuffless approach has the same regulatory status.

For glucose, the FDA's watch-and-ring safety communication says it has authorized no smartwatch or smart ring to independently measure or estimate blood glucose. A watch displaying data from a separate authorized continuous glucose monitor is a different arrangement. Do not treat an unverified nonpiercing claim as a basis for medication decisions.

Making the readings useful

When a number catches your attention, check what produced it:

• Which sensor and signal were used?
• Is the displayed value a measured signal, an estimate or a classification?
• What fit, recording conditions and software requirements apply?
• What evidence supports that feature for someone in your circumstances?

Keep records that help you describe a pattern or ask a useful question. If a score conflicts with symptoms or advice from your care team, the score should not decide what is safe. A wearable record can support a conversation; it cannot replace the examination or testing a health concern may need.

For the broader evidence on activity, weight loss and measurement limits, read Do fitness trackers work? What the research shows.