Sleep343 members
Does magnesium actually improve your sleep?
On magnesium nights the poorest sleepers spent 4.39 fewer minutes in light sleep and 3.42 fewer awake, on a night no longer than usual. The worse your nights, the bigger the difference.
Ultrahuman set 2,190 magnesium-logged nights against each member's own nights that began at the same clock time, across 343 members wearing the Ultrahuman Ring. A magnesium night ran the same length but was built differently — more deep sleep and REM, less light sleep and time awake — and every one of those readings grows larger the worse the member's baseline sleep is.
Hover or tap anywhere on the figures for the numbers behind them.
For members whose sleep was poorest to begin with, a magnesium-logged night carried the same sleep duration but a different shape. Deep sleep ran 2.02 minutes longer and REM 2.85 minutes longer, while light sleep — the shallow stage you drift through on the way down — came out 4.39 minutes shorter and time awake 3.42 minutes shorter. Sleep efficiency, the share of the night spent asleep, rose 0.76 percentage points.
The numbers
Each reading sets a magnesium-logged night against the same member's own nights that began at the same clock time. These are the readings where baseline sleep runs poorest; they shrink steadily as baseline sleep improves.
- Deep sleep ran 2.02 minutes longer, on a range from 0.95 to 3.09 minutes longer (343 members).
- REM sleep ran 2.85 minutes longer, on a range from 0.91 to 4.79 minutes longer (343 members).
- Light sleep came out 4.39 minutes shorter, on a range from 0.12 to 8.67 minutes shorter (343 members).
- Time awake fell 3.42 minutes, on a range from 1.23 to 5.61 minutes shorter (343 members).
- Sleep efficiency rose 0.76 percentage points, on a range from 0.36 to 1.17 percentage points higher (343 members).
What changes on a magnesium night?
The night keeps its length and changes shape. Sleep efficiency rose 0.76 percentage points (p = 0.00025), total sleep did not move at a matched start time, and overnight heart rate and heart rate variability moved too little to call. Light sleep is the only stage a placebo-controlled ring trial moved in minutes: its authors write that the magnesium group “maintained good sleep quality and daytime functioning, while placebo declined” over three weeks in adults with self-reported sleep problems [1]. Figure 1 puts the four stage readings side by side.
Figure 1 · Interactive
Deep sleep and REM rose, light sleep and time awake fell
The worst-sleeping third: magnesium nights against their own nights that began at the same clock time.
343 members · the Ultrahuman Ring
ALT: Diverging bar chart of the change in four sleep stage readings on the nights magnesium was logged, each set against the same members' own nights that began at the same clock time. Deep sleep sits 2.02 minutes above zero and REM 2.85 minutes above it, while light sleep sits 4.39 minutes below zero and time awake 3.42 minutes below.
Who does this show up for?
The people already sleeping poorly. Ranked by their usual sleep score, members' readings grow steadily larger as the underlying sleep gets worse. Deep sleep rose 0.36 minutes where baseline sleep was best (an average sleep score of 84.0), 0.40 in the middle of the range, and 2.02 where it was poorest (an average score of 72.9). REM traced the same slope: 0.71 minutes longer, then 0.90, then 2.85. Figure 2 sets the groups beside each other. The poorest sleepers are also the population the trials recruit: Schuster and colleagues gave magnesium bisglycinate to adults reporting poor sleep and logged a questionnaire improvement that only just cleared what chance could produce [2].
Figure 2 · Interactive
Deep sleep and REM rose most in the third who slept worst
Minutes of change on magnesium nights, ranked by usual sleep score.
Ask Our Data
Answered on this page — nothing you type leaves it343 members · the Ultrahuman Ring
ALT: Grouped bar chart of the change in deep sleep and REM on the nights magnesium was logged, across three groups of members ranked from poorest to best baseline sleep. Deep sleep is 2.02 minutes higher in the poorest-sleeping group, 0.40 in the middle and 0.36 in the best, while REM is 2.85 minutes higher, then 0.90, then 0.71.
Does a magnesium night give you more sleep?
Yes — but we cannot conclusively say the magnesium did it. On the nights magnesium was logged, sleep began 18.94 minutes earlier than on these members' other nights, and those same nights ran 12.22 minutes longer. The Ring records when sleep started, so we cannot separate magnesium bringing sleep on sooner from a magnesium tag marking a night someone was already turning in early.
The light sleep and time awake readings above come from holding the sleep start time fixed, and once it is held the extra length goes: 0.73 minutes, a gap chance alone could produce (p = 0.77). Pooling three small trials in older adults with insomnia, Mah and Pitre found sleep onset 17.36 minutes shorter on magnesium, on evidence they grade low to very low [3].
What this analysis can't tell you
The magnesium log is a tag on a night. It records that a dose was taken, not how much or which form, and untagged nights may be magnesium nights too. Members who log what they take are attentive, and attentive members may run different nights anyway, so a direct causal link to magnesium supplementation is hard to establish. Deep sleep and REM gains run larger on nights tagged after sleep began, so treat their size as approximate. The baseline-sleep gradient is drawn on the Ring's own sleep score rather than a clinic's threshold. And the Ultrahuman Ring records the start of sleep; time in bed before that is invisible to it.
What this means for you
These readings run largest where baseline sleep is poorest, so that is where a magnesium night is most worth watching. If you take magnesium, judge it by how settled the night is rather than how long it runs: compare your light sleep and awake minutes across a week of magnesium nights against a week without, and keep your sleep start time steady, because these readings come from nights that began at the same clock time. Expect a difference of a few minutes rather than an hour. The Ultrahuman Ring tracks deep sleep, REM, light sleep and time awake every night, which is all the week-against-week comparison above needs.
Frequently asked questions
What does less light sleep mean for how you feel?
Light sleep is the shallow stage between the deeper ones. Against nights that began at the same clock time, these nights ran the same length while light sleep and time awake both fell. More of the same hours went on sleep rather than drifting, and sleep efficiency rose 0.76 percentage points (343 members, p = 0.00025). What you would notice is a night that feels less broken; its length does not change.
How can a night hold less light sleep without being shorter?
Light sleep is the flexible part of a night. It stretches when you settle slowly or surface often, and it gives ground when you do neither. That is these nights: light sleep came out 4.39 minutes shorter while the night held its length, and deep sleep and REM took the minutes it gave up. Abbasi and colleagues saw more of the night spent asleep on a sleep diary in older adults with insomnia over eight weeks [4].
Who does this not apply to?
Anyone already sleeping well: where baseline sleep was best, sleep efficiency moved 0.44 percentage points. The comparison also needs a marked night, so it says nothing about members who do not log what they take, and it cannot untangle medication that touches sleep or the nervous system. It is not a recommendation either: the magnesium trials do not report enough in common to pool [5], how much and in what form are unsettled [6], and nothing here is a reason to start or stop.
References
- Hausenblas HA, et al. Sleep Medicine: X 2024. PMID: 39252819.
- Schuster J, et al. Nature and Science of Sleep 2025. PMID: 40918053.
- Mah J, Pitre T. BMC Complementary Medicine and Therapies 2021. PMID: 33865376.
- Abbasi B, et al. Journal of Research in Medical Sciences 2012. PMID: 23853635.
- Chan V, Lo K. Postgraduate Medical Journal 2022. PMID: 33441476.
- Yeom JW, Cho CH. Psychiatry Investigation 2024. PMID: 39086164.
This is an observational analysis of Ultrahuman member data. It describes associations, not causation, and it is not a diagnosis or a substitute for medical advice.
