Fitbit Accuracy: Strong on Steps, Unusual on Blood Oxygen
Fitbit pioneered consumer step tracking and still takes the most different approach to blood oxygen of any brand compared here.
Fitbit's identity was built on step counting before any of today's composite scores existed, and its more recent features often reframe familiar metrics in its own way rather than following the naming conventions other brands settled on. Below is what published research and Fitbit's own help documentation say about each metric it tracks. A few items here rest on device-comparison research that hasn't been verified to the same depth as the rest, and are flagged plainly where that's the case.
Fitbit Air and Fitbit Edge: What Has Been Tested?
Neither has an independent study yet, in a search of Europe PMC in October 2026. Fitbit's sleep and heart tracking has been tested on earlier models such as the Sense 2, Sense and Charge 5, but the Air and the Edge are new hardware, so those results describe the older devices.
What Google says, and what the evidence covers
Fitbit Air. A screenless pebble launched in May 2026 at $99.99, with up to a week of battery. Google lists 24/7 heart rate, resting heart rate, HRV, SpO2, sleep stages and AFib alerts, and states it is not intended for medical purposes.
Fitbit Edge. A tracker with a display, on shelves October 27, 2026 at $179.99, with up to seven days of battery, an ECG app, a skin temperature sensor and built-in GPS that Google says is up to 40 percent more accurate than previous generations, a claim no independent study has tested yet.
Tested on earlier Fitbits. In 35 adults under polysomnography, the Fitbit Sense 2 detected sleep stages with 61.7 to 78.0 percent sensitivity (Robbins et al., Sensors, 2024). In a six-device sleep-lab study, the Fitbit Sense and Charge 5 were among the three devices the authors named for higher stage agreement, kappa 0.42 and 0.41, alongside the Apple Watch Series 8 (Schyvens et al., Sleep Advances, 2025). In 75 adults at two Korean sleep clinics, the Sense 2 was one of three wearables with moderate stage agreement, kappa 0.4 to 0.6, and it and the Pixel Watch scored highest of 11 trackers on deep sleep (Lee et al., JMIR mHealth and uHealth, 2023).
Cost. Both include three months of Google Health Premium, which then renews at $9.99 a month; the core tracking works without it.
How Does Fitbit Measure HRV?
Fitbit calculates HRV using RMSSD, averaged across the full night rather than reported as a single snapshot or tracked continuously through the day. It's overnight-only: Fitbit doesn't offer a daytime HRV reading at all, the same overnight-only design as Oura's ring but the opposite of Whoop's continuous, all-day approach.
What Fitbit calls it
"Heart Rate Variability," shown on the Health Metrics dashboard rather than surfaced as a standalone branded score the way Oura's HRV Balance or Garmin's HRV Status are.
How accurate it is
Two independent studies have tested Fitbit's HRV, both small, and they disagree on which way it errs. In 8 adults over at least three nights, a Fitbit Versa 4's overnight RMSSD was not significantly different from an ambulatory ECG's, with the ECG reading about 92 percent of the Fitbit value on average (Nasr et al., Digital Health, 2025). In 55 adults, 30 of them with COPD, a Fitbit Charge 4 agreed moderately with a Polar H10 chest strap in the healthy group (intraclass correlation 0.69) and strongly in the COPD group (0.87), and read HRV lower than the strap in both (Hermans et al., JMIR mHealth and uHealth, 2024). Neither study tested the Fitbit Air or Fitbit Edge.
KnowYourPrime Evidence Registry: Finding KYP-ER-0003: Wearable HRV Accuracy Against ECG.
How Does Fitbit Measure Resting Heart Rate?
Fitbit averages heart rate during non-movement periods across the full day, including daytime, rather than restricting the measurement to overnight the way Oura does.
What Fitbit calls it
"Resting Heart Rate," the same label as every other brand, with a daytime-inclusive averaging window specific to Fitbit.
What published research found
Emerging, activity-dependentA 2022 evaluation ran the Fitbit Charge 4 and a Samsung Galaxy Watch Active2 through five resting or sedentary phases, two low-intensity phases, and three high-intensity exercise phases in 23 healthy adults, referenced against a medically certified Holter ECG. Across the full protocol, the Charge 4 slightly underestimated heart rate by 1.66 bpm on average (the Samsung overestimated by 3.84 bpm), and both devices kept mean absolute percentage error under 10 percent overall.
The accuracy split by activity rather than by device. The Fitbit had the lower error during resting and sedentary phases (for example, 7.8 vs 9.4 bpm mean absolute error while seated, 8.1 vs 11.6 bpm while typing), while the Samsung performed better during and after higher-intensity activity. Neither device reached what the authors considered sufficient accuracy during seated rest or keyboard typing specifically, the two conditions with the least physical movement.
Nissen et al., JMIR Formative Research, 2022 (doi:10.2196/33635)How Does Fitbit Calculate Its Sleep Score?
Sleep Score combines six sub-metrics, sleep duration, time to sound sleep, sound sleep itself, restlessness, full awakenings, and interruptions, into a single 0-100 number personalized to age and gender, grouped into four bands: Excellent (90-100), Good (80-89), Fair (60-79), and Poor (below 60).
What Fitbit calls it
"Sleep Score," available on the free tier unlike the Premium-gated Daily Readiness Score below. Its sleep-architecture-based inputs overlap with what Oura and Garmin use, but neither brand weighs them identically, so a given score isn't measuring the same combination of things across devices.
Correction, 2026-09-18
This section previously described Sleep Score as a blend of heart rate, total time asleep, and sleep-stage breakdown. Google redesigned the Sleep Score algorithm in 2026 into the six-metric breakdown described above; heart rate now feeds the underlying sleep-stage detection rather than scoring the total directly. The score bands (Excellent 90-100 through Poor below 60) are unchanged. The description above reflects the current algorithm.
What published research found
Established for detectionCorrection, 2026-10-25
This section previously called all seven of these sleep-lab validations independent. The Charge 2 study lists Fitbit as its funder, and the text now says so.
Seven polysomnography validations, one of them funded by Fitbit (the Charge 2 study), spanning five Fitbit generations (Charge HR, Charge 2, Charge 4, Inspire 2) and populations from healthy adults to children to patients with chronic insomnia, converge on the same shape of result the rest of this device category shows: Fitbit is good at telling sleep from wake, and weaker at telling sleep stages apart.
Sensitivity for detecting sleep stayed high across every study: 0.96 in 44 healthy adults wearing a Charge 2, 95.8 percent in 59 children wearing a Charge HR, 93.9 percent in 9 adults wearing an Inspire 2, and 92.6 percent in 37 Chinese patients with chronic insomnia wearing a Charge 4. Specificity for detecting wake told a different story every time: 0.61, 56.3 percent, 13.1 percent, and 35.7 percent in the same four studies, respectively. In the chronic-insomnia study, a research-grade actigraph worn alongside the Fitbit caught wake more reliably (62.2 percent), evidence that this gap is a Fitbit-specific limitation rather than one shared by every wrist-worn device.
Stage classification carries the same asymmetry. The Charge 2 study overestimated light sleep by 34 minutes and underestimated deep sleep by 24 minutes, with deep-sleep detection accuracy of only 0.49. The Charge 4 study found the same direction of error at larger magnitude: light sleep overestimated by 37.7 minutes, deep sleep underestimated by 41.4 minutes. A separate free-living study of sleep-stage transition probabilities, not just time spent in each stage, found systematic bias ranging from 0 to 60 percent, with Fitbit tending to overestimate the chance of staying in whatever stage it had just detected.
Two reviews put Fitbit beside other brands. One, of eight studies, found the Charge 4 had the smallest REM error of the three devices it compared, 4.0 minutes against polysomnography, and higher sensitivity for deep sleep (75 percent) and REM (86.5 percent) than the Garmin Vivosmart 4 or WHOOP. A meta-analysis of 24 studies and 798 people found that wake after sleep onset on Fitbit did not differ significantly from the sleep lab, while Fitbit recorded longer time to fall asleep than the other devices pooled.
de Zambotti et al., Chronobiology International, 2018 (doi:10.1080/07420528.2017.1413578 [funding listed from Fitbit]) · Godino et al., PLOS ONE, 2020 (doi:10.1371/journal.pone.0237719) · Lim et al., Nature and Science of Sleep, 2023 (doi:10.2147/nss.s391802) · Dong et al., PLOS ONE, 2022 (doi:10.1371/journal.pone.0275287) · Liang & Chapa Martell, JMIR mHealth and uHealth, 2019 (doi:10.2196/13384) · Schyvens et al., JMIR mHealth and uHealth, 2024 (doi:10.2196/52192) · Lee et al., Journal of Clinical Sleep Medicine, 2025 (doi:10.5664/jcsm.11460)What holds up outside the lab
Group-level accurateA naturalistic validation against portable home polysomnography in 59 police officers and paramedics working rotating shifts, a harder real-world test than an overnight lab stay, found sleep onset, offset, total sleep time, REM duration, and combined light-sleep duration were all unbiased at the group level, but with limits of agreement wide enough that individual nights varied considerably. The same algorithm overestimated REM latency by 29.4 minutes and wake after sleep onset by 37.1 minutes, and per-stage accuracy ranged from 0.55 for light sleep up to 0.86 for REM.
A 2018 systematic review of 67 studies reached a broader verdict that lines up with the studies above: Fitbit devices were about as accurate as research-grade accelerometers for total sleep time and time in bed, but markedly overestimated time spent in higher-intensity sleep activity and, separately, were unlikely to measure energy expenditure accurately in any condition tested.
Stucky et al., Journal of Medical Internet Research, 2021 (doi:10.2196/26476) · Feehan et al., JMIR mHealth and uHealth, 2018 (doi:10.2196/10527)KnowYourPrime Evidence Registry: Finding KYP-ER-0007: Wearable Sleep Tracking Accuracy Against a Sleep Lab.
Does Fitbit Have an ECG Feature?
Yes, on the Sense, Sense 2, Charge 5, and Charge 6, which offer AFib monitoring through an ECG app similar in concept to Apple's, though narrower in device availability than Apple's since older Fitbit models (Charge 4 and earlier) lack the required hardware. Devices with the feature pair it with a second, passive feature, Irregular Heart Rhythm Notifications, which uses the same optical (PPG) sensor as everyday heart rate tracking to flag possible AFib in the background rather than requiring an on-demand spot check.
What Fitbit calls it
Officially the "Fitbit ECG App," branded the "Google ECG App" in newer Google-authored materials. Both it and Irregular Heart Rhythm Notifications carry FDA clearance and a CE mark, but they work differently: the ECG app is a one-time spot check a user runs on demand, while notifications monitor passively over time.
What testing found
The background notifications were tested at scale in the Fitbit Heart Study: of 455,699 US adults without known AFib, 1 percent received an irregular rhythm alert over a median of about four months, and 4 percent of those aged 65 and over. When an alert fired again while a participant wore an ECG patch, the ECG showed AFib 98.2 percent of the time; across everyone alerted who then wore the patch for a week, 32.2 percent had AFib during it. Several of the authors worked for Fitbit (Circulation, 2022 [author affiliations include Fitbit]). The on-demand ECG app on a Fitbit Sense was tested in 201 cardiology patients against a physician-read 12-lead ECG: 66 percent sensitivity and 79 percent specificity for AFib, with 21 percent of tracings inconclusive, against 85 and 75 percent for an Apple Watch 6 in the same patients (JACC: Clinical Electrophysiology, 2023 [funding listed from Abbott]).
Correction, 2026-09-17
This section previously stated the ECG app was available only on the Sense 2. Fitbit's own support documentation now lists it on the original Sense, Sense 2, Charge 5, and Charge 6. The device list has been corrected above.
How Accurate Is Fitbit's VO2 Max Estimate?
VO2 max is estimated by tracking the relationship between running pace and heart rate during an outdoor GPS run, comparing that efficiency against resting heart rate, then benchmarking the result against age, sex, and weight norms. A peer-reviewed validation of the Fitbit Charge 2, from before the feature's rename, found the score gave a consistent, unbiased read on relative fitness but overestimated actual lab-measured VO2 max, and that a simple non-exercise prediction equation using only age, resting heart rate, and body composition matched it for accuracy without requiring a run at all.
What Fitbit calls it
"VO2 max," the same direct term Garmin uses; Google Health renamed the feature from "Cardio Fitness Score" to match. Apple is now the outlier still branding the equivalent estimate "Cardio Fitness."
Correction, 2026-09-18
This section previously described the feature as "Cardio Fitness Score," calling it Fitbit's distinct name for the estimate versus Garmin's "VO2 Max" and Apple's "Cardio Fitness." Google Health has since renamed it "VO2 max" (Google's own support documentation now says "tap VO2 max, formerly known as cardio fitness score"), matching Garmin's label directly. The name above has been corrected.
KnowYourPrime Evidence Registry: Finding KYP-ER-0011: Wearable VO2 Max Estimate Accuracy.
How Does Fitbit Measure Blood Oxygen?
Fitbit reports two separate things under two separate names. "SpO2" is an actual nightly average blood-oxygen percentage with a 7-day range, sourced from sleep data. "Estimated Oxygen Variation" (EOV) is a different, additional relative overnight graph showing how much blood oxygen fluctuates, with a red, yellow, or green pattern flagging possible breathing disturbances. Neither is an on-demand daytime spot check the way Apple's or Garmin's on-demand modes are; both are sleep-derived.
What Fitbit calls it
"SpO2" for the percentage, "Estimated Oxygen Variation" for the separate variation graph. It is easy to conflate the two since both come from the same sensor and the same night's data, but they answer different questions: SpO2 gives a number, EOV flags a pattern.
Correction, 2026-09-17
This section previously stated that Fitbit deliberately doesn't report a raw SpO2 percentage at all. That was wrong: Fitbit's own support documentation describes a separate "SpO2" feature that does report a percentage, distinct from Estimated Oxygen Variation. The claim has been corrected above.
KnowYourPrime Evidence Registry: Finding KYP-ER-0001: Smartwatch Blood Oxygen (SpO2) Accuracy Against Arterial Blood (SaO2); Finding KYP-ER-0002: Pulse Oximeter Accuracy and Skin Tone.
What Does Fitbit's Daily Readiness Score Measure?
Daily Readiness Score weighs sleep quality, HRV, and resting heart rate against a personal baseline built from a week of device wear, alongside recent activity load.
What Fitbit calls it
"Daily Readiness Score," free for all compatible-device users the same way Whoop's Recovery Score and Oura's Readiness Score are. Fitbit Premium adds proactive coaching advice built on top of the score rather than gating the score itself.
Correction, 2026-09-18
This section previously stated Daily Readiness Score was a Premium-only feature that Fitbit gated behind a subscription, unlike Whoop's and Oura's free composite scores. That was true when written and is no longer true: Google Health's own support documentation confirms the score is now available to all compatible-device users, with Premium adding coaching on top rather than unlocking the score itself. The description above has been corrected.
Worth knowing
A real peer-reviewed evaluation found composite recovery scores across brands, Fitbit's included, are not methodologically equivalent to each other despite similar-sounding names.
KnowYourPrime Evidence Registry: Finding KYP-ER-0005: Readiness and Recovery Score Validity.
How Does Fitbit's Stress Management Score Work?
Stress Management Score is calculated once daily, after waking, from more than 10 inputs grouped into three categories: Responsiveness (heart rate and electrodermal activity, EDA, signals of stress), Exertion Balance (recent physical activity load), and Sleep Patterns (how last night's sleep affects predicted resilience). It scores 0-100, higher meaning fewer physical signs of stress.
What Fitbit calls it
"Stress Management Score." The EDA sensor, which measures micro-sweat responses on the palm, debuted on Fitbit Sense as the first electrodermal-activity sensor in a smartwatch and can also run a standalone guided EDA "spot check" separate from the daily composite score.
What testing found
In 55 university students doing a social stress task and a reading task, the Fitbit Charge 5 agreed only moderately with a research-grade chest device: correlations of 0.45 to 0.58 for heart rate and 0.42 to 0.50 for EDA, with a pattern of underestimating heart rate and overestimating EDA. The authors advise caution in interpreting its readings (Psychophysiology, 2025). No study cited here has tested the daily Stress Management Score itself.
How Accurate Is Fitbit's Step and Calorie Tracking?
Fitbit was the original consumer step-counting wearable, and step count remains one of the more reliable metrics across this entire device category. Calorie tracking is the weaker point, shared with every other brand compared here.
Worth knowing
A systematic review of 65 studies across 72 devices and 29 brands found energy-expenditure error exceeded 30% for every brand tested, even though step count and heart rate performed comparably better in the same review; across 20 of its studies the Fitbit Charge and Charge HR stayed under 25 percent error for steps (Journal of Medical Internet Research, 2022). A meta-analysis of 41 Fitbit validation studies found recent models read slightly low on all three measures: 2.99 beats per minute for heart rate, 2.77 kcal per minute for energy and 3.11 steps per minute, with individual comparisons ranging from 13.07 steps per minute under to 6.86 over (JMIR mHealth and uHealth, 2022). This is a category-wide limitation, not something specific to Fitbit.
What device-specific research found
Emerging, steps solidA 2021 free-living comparison of the Fitbit Charge 2, Fitbit Alta, and Apple Watch 2 against a Yamax pedometer, an ActiGraph accelerometer, and a Polar H7 chest strap in 48 adults found step counts within single digits of the pedometer reference (11,734 for the Charge 2 and 11,922 for the Alta, against 10,906 for the Yamax), with correlations of 0.84 to 0.95 across all three wrist devices and mean absolute percentage error of 17.1 to 35.5 percent. Heart rate tracked the Polar chest strap closely for the Charge 2, with 4 to 16 percent error, better during sedentary periods than during vigorous activity.
Moderate-to-vigorous activity minutes were the weak point: the Charge 2 recorded 76.3 minutes against the ActiGraph reference's 47.8 minutes, a substantial overestimate shared by the Fitbit Alta and the Apple Watch alike. A separate free-living study of older adults found the same overestimation pattern for moderate-to-vigorous minutes, while rating the Fitbit's calorie tracking as the most accurate of the trackers it tested (12.25 percent mean absolute percentage error) even though it still underestimated calories burned.
Bai et al., PLOS ONE, 2021 (doi:10.1371/journal.pone.0251975) · Tedesco et al., JMIR mHealth and uHealth, 2019 (doi:10.2196/13084)KnowYourPrime Evidence Registry: Finding KYP-ER-0009: Wearable Heart Rate Accuracy; Finding KYP-ER-0010: Wearable Calorie and Energy Expenditure Accuracy.
What Fitbit Doesn't Do
One real absence relative to its two closest competitors on price and positioning.
No CGM integration
Unlike Oura's live Dexcom Stelo integration, Fitbit has no continuous glucose monitoring feature as of October 2026. Abbott and Google announced in August 2026 that glucose data from Abbott's Lingo sensor will appear in the Google Health app, with a rollout planned for later in 2026 and tied to the paid Google Health Premium tier; no launch had been announced by October 2026.
If step and activity tracking is the priority and a lower price point matters, Fitbit's original strength still holds up. If a native ECG or a free (not Premium-gated) recovery score matters more, that shapes which specific Fitbit model, if any, actually fits.
KnowYourPrime Evidence Registry: Finding KYP-ER-0016: Wearable Sleep Stage Accuracy