Supplements for Sleep: The Honest Evidence Ranking

The table above is the answer, and it is deliberately unflattering. The best-evidenced option on it — plain magnesium — bought older adults with insomnia about 17 minutes of faster sleep onset, on a pooled estimate drawn from two small trials and 55 people, which the reviewers themselves graded low certainty.¹ Ashwagandha's effect on sleep is real, statistically significant, and also small.² Nothing here is a sedative, and nothing here will move a body clock that is three hours out of position.

Most "best supplements for sleep" pages list a dozen options and grade every one of them "may help." This page grades four, gives three of them the same middling C, and prints a D in public. Below is the reasoning behind each grade — and the two levers that beat all four.

The two things that beat everything on this list

Behavior first. The American College of Physicians recommends that all adults with chronic insomnia disorder receive cognitive behavioral therapy for insomnia (CBT-I) as the initial approach — a strong recommendation on moderate-quality evidence, which is a full tier above anything any supplement on this page has earned.³ It is unglamorous, it is not a pill, and there is nothing to put in a cart. It also works better than what you were about to buy.

Then caffeine timing. In a double-blind study of 12 adults sleeping at home, 400 mg of caffeine taken six hours before bed cut self-reported sleep by 41 minutes — short of significance at that sample size (p = 0.08) — while the EEG monitor they wore recorded significantly less total sleep at all three timings tested (bedtime, three hours before, six hours before), by 1.1 to 1.2 hours.⁴ Set even the smaller of those two numbers against magnesium's 17 minutes¹ and the priority order gets easy: moving your last coffee from 4 p.m. to noon is plausibly a bigger intervention than anything below.

That does not make supplements pointless. It makes them the last ten percent, not the first.

How these grades work

A means multiple well-run human trials agree. B is moderate: consistent, but limited. C is limited — a real signal built on small, short, or low-quality trials. D means the human evidence for this goal is not there.

Sleep is a hard field to grade. Most trials are small, most primary outcomes are self-reported questionnaires, and the placebo response to "how did you sleep?" is enormous. Nothing on this page reaches A or B, and we are not going to invent one to make the page look more useful.

Magnesium

The number worth memorizing comes from a 2021 systematic review that identified three randomized trials in 151 older adults — but read what it actually pools. Sleep onset latency was 17.36 minutes shorter on magnesium than placebo (95% CI −27.27 to −7.44), and that estimate rests on two of the three trials, 55 participants in total, which is why the reviewers graded it low certainty for imprecision. Total sleep time improved by about 16 minutes but not significantly, all three trials carried moderate-to-high risk of bias, and the review concludes the quality of the literature is "substandard" for making recommendations.¹ That is the whole case, stated at full strength.

Here is a detail no sleep listicle mentions: at the time of writing, the NIH Office of Dietary Supplements' health-professional fact sheet on magnesium — the most thorough government summary of the mineral there is — has no sleep section and does not mention insomnia anywhere.⁵ Magnesium's sleep reputation is bigger than its evidence base, and the most authoritative reference on the mineral simply doesn't engage with it.

So why does it still lead this ranking? Because the plausible mechanism is the boring one — you are topping up a nutrient a lot of people are short on. NHANES data suggest 48% of Americans of all ages take in less magnesium from food and beverages than their estimated average requirement, with older men and adolescents at highest risk.⁵ The trials that found the biggest effects were run in older adults,¹ and the most recent randomized trial found the improvement concentrated in participants who started with the lowest dietary intake.⁷ The pattern points one way: magnesium looks less like a sleep aid and more like a deficiency correction that happens to show up in sleep scores.

One more honest wrinkle. The pooled trials dosed 320, 500, and 729 mg of elemental magnesium per day¹ — two of the three above the Tolerable Upper Intake Level for supplemental magnesium, which the Food and Nutrition Board sets at 350 mg for adults because higher supplemental doses cause diarrhea.⁵ A sensible starting point is 200 mg of elemental magnesium in the evening with that 350 mg ceiling in view; our magnesium dosage guide works through the RDA-versus-upper-limit math, and magnesium for sleep is the full workup of the trials. The ingredient hub is magnesium.

Ashwagandha

A 2021 systematic review and meta-analysis of five randomized trials in 400 participants found a small but significant effect on overall sleep, with a standardized mean difference of −0.59 (95% CI −0.75 to −0.42) and substantial heterogeneity between studies (I² = 62%); effects were more prominent in adults diagnosed with insomnia, at doses of 600 mg/day or more, and at durations of eight weeks or longer.² The review's own word for the effect is "small," and the same analysis found improvements in mental alertness on rising and in anxiety, but none in quality of life.²

That is comparable evidence to magnesium's — arguably a touch better, since it rests on more trials and more participants. It sits below magnesium here for a different reason: the safety file is thicker. NIH's fact sheet describes ashwagandha as well tolerated for up to about three months, notes that evidence on safety over many months or years is lacking, and documents liver injury and thyroid effects alongside a recommendation against use in pregnancy.⁶ Magnesium's downside at a sensible dose is loose stools.⁵

The sedative interaction is the one that matters most on a sleep page: the people most likely to try a calming supplement are also the people most likely to already be taking something calming.

Where ashwagandha genuinely earns its place is stress-driven sleeplessness — the case where the supplement is aimed at the cause rather than the symptom. Read ashwagandha side effects before you start, not after; the hub is ashwagandha.

Magnesium glycinate — and the form question

Everyone will tell you glycinate is the magnesium for sleep. The trials disagree, or rather, they never tested it. Of the three randomized trials behind magnesium's sleep grade, two used magnesium oxide — the cheap, poorly absorbed form the internet tells you to avoid — and the third used magnesium citrate.¹ Not one used glycinate. Whatever the pooled 17 minutes is evidence for, it is evidence for the mineral, not for the compound on the label.

Glycinate does have one direct trial. In 2025, 155 healthy adults reporting poor sleep took 250 mg of elemental magnesium as bisglycinate (with 1,523 mg of glycine) or placebo for four weeks; Insomnia Severity Index scores fell 3.9 points on magnesium versus 2.3 on placebo (p = 0.049), an effect the authors classed as small (Cohen's d = 0.2), and the improvement correlated inversely with baseline dietary magnesium intake.⁷ Worth noting on funding, since so much supplement research is sponsored: that trial was funded by an academic institute rather than a manufacturer, though one co-author discloses directing a contract research organization that receives funding from nutraceutical companies.⁷

The practical read: pick glycinate for your gut, not for its evidence. Citrate absorbs well but its signature effect is drawing water into the bowel — genuinely useful if constipation is also on your list, an odd thing to introduce at bedtime; oxide is the cheapest and among the forms most often reported to cause diarrhea.⁵ Our guide to the types of magnesium compares them properly, and magnesium glycinate has the form hub.

Creatine

Creatine does not improve your sleep, and it has never claimed to. It is on this page because "creatine and sleep" is a question people genuinely ask, and it has a real but narrow answer worth stating instead of ignoring.

In a 2024 crossover study, 15 participants kept awake for 21 hours took a single 0.35 g/kg dose of creatine or placebo; the creatine condition showed measurable changes in cerebral high-energy phosphates along with improvements in processing speed and performance on memory, language, and numeric tasks.⁸ That is interesting neuroscience about brain energetics under sleep deprivation. It is also fifteen people, a single dose several times the size of a standard serving, in a group that was not sleeping at all.

None of that is evidence that creatine helps you sleep. D — insufficient evidence for this goal, and it stays a D until somebody runs a trial that actually measures sleep. If you take creatine for training reasons, that is a separate and much better-supported case: see creatine.

What about melatonin, valerian, L-theanine, and glycine?

Melatonin is the obvious omission here, and the fastest-growing one: US adult use rose from 0.4% in 1999–2000 to 2.1% in 2017–2018, more than a fivefold increase.¹⁰ The awkward part is what the sleep specialists say about it. The American Academy of Sleep Medicine's clinical practice guideline suggests clinicians not use melatonin for sleep onset or sleep maintenance insomnia in adults, a weak recommendation — and reaches the same "suggest against" conclusion for valerian, L-tryptophan, and diphenhydramine, the antihistamine in many over-the-counter sleep aids.⁹

Read that in context rather than as a blanket dismissal. The guideline is specifically about chronic insomnia. Melatonin's better-supported role is circadian: nudging a delayed body clock, jet lag, shift work. It is a timing signal, not a sedative — which is why taking 10 mg because 3 mg "didn't work" misunderstands the job it does.

L-theanine, glycine, and apigenin are absent from the table above for a duller reason: we have not done the evidence review on them yet, and assigning a grade we have not earned is the exact behavior this page exists to avoid. When those hubs are researched they get the same A–D treatment, including a D where a D is warranted.

That is the honest gap, and naming it beats papering over it. A ranking that quietly leaves melatonin out isn't a ranking — but neither is one that hands out grades it hasn't earned.

When it isn't a supplement problem

How to test one of these on yourself without fooling yourself

1. One variable at a time. Starting magnesium, ashwagandha, and a new wind-down routine in the same week tells you nothing about any of them.
2. Give it four weeks. The trials behind these grades ran four to twelve weeks, not four days.⁷²
3. Write down a number before you start. Minutes to fall asleep, number of night wakings, how you feel at 10 a.m. Track two weeks of baseline, then compare. "I think it's helping" is precisely the signal a placebo produces.
4. Respect the ceiling. For supplemental magnesium that ceiling is 350 mg/day for adults.⁵
5. Fix the free things in parallel — consistent wake time, dark cool room, last coffee before noon.⁴³
6. If nothing changes in a month, it isn't your fix. That is a real result, and it saves you the next twelve months of subscriptions.

The bottom line

If you are going to buy one thing, buy magnesium — glycinate if your gut is sensitive — take it in the evening, keep the 350 mg/day supplemental limit in view, and expect a modest improvement that is most likely if your diet runs low in it.¹⁵⁷ If stress is what keeps you awake, ashwagandha at 600 mg/day for at least eight weeks has evidence on par with magnesium's² and safety reading to do first.⁶ Skip creatine for this purpose. And do the behavioral work regardless, because it is the only intervention on this page with a strong recommendation behind it.³

The most honest headline available: the best-supported supplement on this page is worth roughly a quarter of an hour of faster sleep onset in the population it was tested in,¹ and moving your last coffee earlier may be worth more than that.⁴

There are no affiliate links or product picks on this hub, by design. This page's job is the ranking; choosing between specific products belongs where you can see them side by side.

Frequently asked questions