Consumer brain-sensing wearables

Also known as Consumer EEG headbands, EEG earbuds, Consumer neurotech wearables

Headbands and earbuds sold to consumers record brain signals to estimate sleep stages. Small studies show that some of these devices agree reasonably well with lab sleep scoring for the main stages, but evidence for light sleep, clinical groups and long-term home use is thin, and some jurisdictions now treat neural data as sensitive.

Pilot and early commercialReported, unverifiedReviewed 21 Oct 2026
AI illustration · Illustrative scene, not a photograph of the specific device or experiment.

In plain language

Some headbands and earbuds have small electrodes that touch the scalp or the inside of the ear. They pick up tiny electrical signals from the brain, using the same basic idea as the EEG (electroencephalography) tests used in hospitals. Companies use these signals to estimate things people care about, such as which stage of sleep someone is in during the night.

The central question is whether these estimates are accurate enough to trust. The reference test for sleep is polysomnography (PSG), an overnight lab study that records brain waves along with eye movements, muscle activity and breathing. Small studies suggest that some consumer devices agree reasonably well with lab scoring for the main sleep stages, but they agree less well for the lightest stage of sleep. The studies so far mostly involved healthy adults, usually for a single night. Brain signals are also now treated as sensitive personal data in some places: two US states changed their privacy laws in 2024, and UNESCO adopted a global ethics recommendation in 2025.

Technical explanation

Consumer devices use dry electrodes with few channels. The Muse-S headband records four EEG channels at the sites Fp1, Fp2, TP9 and TP10, and the Lanthier study scored its data with InteraXon's automated staging algorithm (Lanthier et al., 2025). The NextSense in-ear system uses a custom-fit earbud device (Berent et al., 2026). With few channels, products depend on onboard algorithms to turn a small number of signals into sleep stages, and signal quality remains a constraint: Lanthier et al. (2025) excluded 16% of recordings for poor signal quality. Agreement is usually reported as epoch-by-epoch accuracy and Cohen's kappa, which corrects for agreement expected by chance.

Lab evidence is encouraging for overall sleep structure but narrow in scope. Lanthier et al. (2025) recruited 56 adults and analysed 47 in a single lab night. They reported a kappa of 0.76 across all epochs for the Muse-S, with per-stage accuracy of 87.7% to 96.4%. The Muse-S ran about 7 minutes high on total sleep time and about 15 minutes high on deep sleep, and about 15 minutes low on light sleep. Arnal et al. (2020) studied the Dreem headband in 25 adults; its automatic staging reached 83.5 ± 6.4% accuracy, compared with 86.4 ± 8.0% for five human scorers, but accuracy for the N1 stage was 47.7 ± 15.6%, against 54.2 ± 16.8% for the scorers. Berent et al. (2026) tested a custom-fit in-ear EEG device in 16 healthy adults and reported a kappa of 0.77 for overnight staging against scalp EEG, with N1 agreement at 54% of epochs. A further eight people with central disorders of hypersomnolence were included in the wider study, and daytime sleepiness testing gave moderate agreement (kappa 0.50). The authors describe the work as a feasibility study and call for larger confirmatory studies.

Privacy rules are catching up with these devices. Colorado's HB 24-1058 (signed 17 April 2024) adds biological data, which covers neural properties, to the sensitive data category of the Colorado Privacy Act when the data is used or intended for identification. California's SB 1223 (Chapter 887, Statutes of 2024) adds neural data to the CCPA's sensitive personal information category. UNESCO's Recommendation on the Ethics of Neurotechnology, adopted on 11 November 2025, calls on frameworks to consider neural data as sensitive personal data, and states that privacy, including mental privacy, is fundamental for personal identity and agency. As a recommendation, it depends on member states to put its provisions into national law.

What the evidence says

Each statement is typed by the strength of its evidence and linked to the source that supports it.

Demonstrated Shown in a published experiment, test or shipping product.

In 47 adults analysed in a single-night lab study, automated Muse-S sleep staging reached a Cohen's kappa of 0.76 against level 1 polysomnography across all epochs, with per-stage accuracy of 87.7% to 96.4%.

Sources: [1] · Results, Table 3 · medium confidence

In the Muse-S study, total sleep time ran about 7 minutes high and deep sleep about 15 minutes high, while light sleep ran about 15 minutes low, and the NREM1 kappa was 0.41.

Sources: [1] · Results · medium confidence

In the Dreem headband study of 25 adults, automatic staging reached 83.5 ± 6.4% accuracy against polysomnography, compared with 86.4 ± 8.0% for five human scorers, and the authors describe the headband as in the range of individual scorers rather than equal to them.

Sources: [2] · Abstract and Table 4; Discussion for the 'range of individual scorers' wording · medium confidence

Several authors of the 2020 Dreem headband validation study were Dreem employees, and the study was supported by Dreem.

Sources: [2] · Disclosures

In 16 healthy adults with overnight recordings, sleep staging from a custom-fit in-ear EEG device agreed with scalp EEG at a Cohen's kappa of 0.77, with wake agreeing in about 90% of epochs and N1 in about 54%.

Sources: [3] · Results (overnight sessions) · medium confidence

Colorado's HB 24-1058, signed on 17 April 2024 and in force from 7 August 2024, adds biological data, which covers neural properties, to the sensitive data category of the Colorado Privacy Act when the data is used or intended for identification.

Sources: [5]

California's SB 1223, chaptered as Chapter 887 of the 2024 Statutes on 28 September 2024, adds neural data to the California Consumer Privacy Act's sensitive personal information category.

Sources: [6]

UNESCO's General Conference adopted the Recommendation on the Ethics of Neurotechnology on 11 November 2025; it calls on frameworks to consider neural data as sensitive personal data, and states that privacy, including mental privacy, is fundamental for personal identity and agency.

Sources: [7] · Paragraphs 49 and 85

Reported, unverified Stated by a company or team but not independently confirmed.

NextSense describes Smartbuds as having six EEG sensors and reports that in a 106-night beta, slow-wave activity increased and nearly 50% of participants reported better or much better sleep (self-reported); the release cites no peer-reviewed or independent study for these claims.

Sources: [4] · low confidence

Limitations and open problems

  • The validation studies are small and mostly cover healthy adults in single lab nights: 47 analysed adults for the Muse-S (Lanthier et al., 2025), 25 adults for the Dreem headband (Arnal et al., 2020) and 16 healthy adults for in-ear EEG (Berent et al., 2026).
  • Signal quality limits real use. In the Muse-S study, 16% of recordings were excluded for poor signal quality. The authors report that signal quality degraded after about 14 to 20 nights of use, but this comes from their field experience rather than from their single-night data (Lanthier et al., 2025).
  • The lightest stage, N1, is the weakest result in all three validation studies, so claims about light sleep or sleep onset need particular care.
  • Company involvement is common. InteraXon provided the Muse-S in kind but did not fund its study, several authors of the Dreem study were Dreem employees and the study was supported by Dreem, and several in-ear study authors are affiliated with NextSense Inc., whose funding is also listed in the paper.
  • Many product claims come from manufacturers. The NextSense Smartbuds release cites a 106-night beta and no peer-reviewed or independent study, and it gives no beta methods, participant count or comparison group.
  • Privacy coverage is uneven. Colorado's rule depends on whether biological data is used or intended for identification, California's rule sits within the CCPA, and UNESCO's recommendation takes effect only through national action.

Possible future uses

Speculative

Ideas, not predictions. None of these has been demonstrated.

  • Home sleep screening for older adults and people with sleep disorders, if validation extends to these groups.
  • Closed-loop sound timed to slow-wave sleep, using earbuds that detect sleep stages in real time.
  • Workplace focus tracking, which would need strict consent rules and clear limits on how neural data may be used.

Timeline

  1. 20 May 2020Arnal and colleagues publish a study in the journal Sleep comparing the Dreem headband with polysomnography in 25 adults. Sources: [2]
  2. 17 Apr 2024Colorado's governor signs HB 24-1058, which adds biological and neural data to the sensitive data category of the Colorado Privacy Act. Sources: [5]
  3. 28 Sept 2024California's SB 1223 is chaptered as Chapter 887 of 2024, adding neural data to the CCPA's sensitive personal information category. Sources: [6]
  4. 11 Nov 2025UNESCO's General Conference adopts the Recommendation on the Ethics of Neurotechnology in Samarkand, Uzbekistan. Sources: [7]
  5. 12 Dec 2025SLEEP Advances publishes Lanthier and colleagues' single-night comparison of the Muse-S headband with level 1 polysomnography. Sources: [1]
  6. 9 Feb 2026NextSense launches Smartbuds, which it describes as consumer EEG earbuds with six EEG sensors. Sources: [4]

Futures this could help bring about

AI illustration
How far away?

Better deep sleep on demand

Pilots · 9 of 12 steps done

AI illustration
How far away?

Brain data protected by law

Pilots · 5 of 9 steps done

Glossary

EEG (electroencephalography)
A method that records the brain's electrical activity using electrodes placed on the scalp, in the ear or on the surface of the brain.
Polysomnography (PSG)
An overnight sleep study, usually in a lab, that records brain waves and other body signals and is the standard reference for scoring sleep stages.
Cohen's kappa
A measure of agreement between two raters or methods that corrects for agreement expected by chance. A value of 1 means perfect agreement.
Neural data
Information gathered by measuring the activity of the brain or nervous system, which some privacy laws now treat as sensitive personal data.

Sources

  1. peer reviewed paper
    Assessing the performance of a portable electroencephalographic sleep monitor against level 1 polysomnography
    Malika Lanthier, Michael-Christopher Foti, Karina Fonseca et al.. SLEEP Advances (Oxford Academic), 12 Dec 2025. DOI: 10.1093/sleepadvances/zpaf089. Accessed 21 Oct 2026.
  2. peer reviewed paper
    The Dreem Headband compared to polysomnography for electroencephalographic signal acquisition and sleep staging
    Pierrick J. Arnal, Valentin Thorey, Eden Debellemaniere et al.. Sleep, 20 May 2020. DOI: 10.1093/sleep/zsaa097. Accessed 21 Oct 2026.
  3. peer reviewed paper
    A novel, wearable, in-ear EEG technology to assess sleep and daytime sleepiness
    Jonathan Berent, Prabhjyot Saini, Andrew X Stewart et al.. Bioelectronic Medicine, 22 May 2026. DOI: 10.1186/s42234-026-00207-x. Accessed 21 Oct 2026.
  4. company announcement
    NextSense Launches Smartbuds, the World's First Truly Wireless EEG Earbuds
    Business Wire (syndicated on Yahoo Finance), 9 Feb 2026. Accessed 21 Oct 2026.
  5. government report
    HB24-1058: Protect Privacy of Biological Data
    Colorado General Assembly, 17 Apr 2024. Accessed 21 Oct 2026.
  6. other
    SB 1223: Consumer privacy: sensitive personal information: neural data
    CalMatters Digital Democracy, 28 Sept 2024. Accessed 21 Oct 2026.
  7. standard
    Recommendation on the Ethics of Neurotechnology
    UNESCO, 11 Nov 2025. Accessed 21 Oct 2026.

Revision history

3 changes since this profile was created.

Changes on 21 Oct 2026
Published on owner instruction (content loop iteration 51)095cfd1
Sources and claims re-checked against originalsfbb434c
Initial profile drafted from primary sources7e66249