The sleep supplement aisle is one of the most overpromoted corners of the wellness industry. Search any major retailer for "sleep aid" and you get thousands of products, most of them stacking six or eight ingredients with names that sound impressive and evidence bases that fall apart on inspection. This is a triage.
Of the roughly two dozen ingredients that show up regularly in sleep formulas, four have real clinical evidence behind them. Modest evidence — none of them is a sedative, and none will fix chronic insomnia. The other twenty or so range from "one small trial, mostly indirect" to "essentially no controlled data." That is what the literature actually shows once you strip out the marketing.
This article walks through both piles honestly, then covers the supplement-quality problem that makes any of this hard to trust without third-party testing.
The four with real evidence
1. Magnesium (glycinate or bisglycinate)
Magnesium is the most defensible sleep supplement in this category, and it is one of the cheapest. The clinical evidence is not huge, but it is consistent across mechanism, epidemiology, and small trials.
The most cited randomized trial is Abbasi 2012, a small double-blind study of 46 elderly adults with insomnia. Participants who received 500 mg of elemental magnesium daily for 8 weeks showed statistically significant improvements in sleep time, sleep efficiency, and self-reported insomnia severity compared to placebo. The trial is small, the population is narrow, and the effect is modest — but it is a real controlled signal.
Cao 2018 is an observational study of over 1,400 adults that found higher dietary magnesium intake was associated with better sleep quality and duration. Observational data cannot prove causation, but it lines up with the trial data.
Mechanistically, magnesium is a cofactor in GABAergic signaling and helps regulate the HPA axis. Most Americans get less magnesium than the RDA. Correcting a mild deficiency is plausibly why the trial effect showed up in the elderly cohort more clearly than it does in already-replete populations.
The form matters. Magnesium glycinate and bisglycinate have the best tolerability and the fewest gastrointestinal side effects. Magnesium oxide is common in cheap products and is barely absorbed. Magnesium citrate is a mild laxative at doses above about 200 mg, which is a feature if you need it and a bug if you do not.
2. Low-dose melatonin (0.3 to 1 mg)
Melatonin is one of the most misunderstood supplements on this list. The evidence is real but narrow, and the doses on most bottles are wildly wrong.
The Buscemi 2005 meta-analysis pooled data across trials and found melatonin shortened sleep onset by about 7 minutes on average. That is not a lot. It is not a sedative. What melatonin actually does well is shift circadian timing — jet lag, shift work, delayed sleep phase syndrome. For chronic primary insomnia in adults, the effect is small.
The dose problem is real. Most retail melatonin comes in 3, 5, or 10 mg tablets. Physiological nighttime levels are extremely low, and doses above about 1 mg produce blood levels that overshoot the receptor and can produce next-morning grogginess. The effective dose for a circadian shift is 0.3 to 1 mg. Higher is not more.
Melatonin is best used as a targeted tool — jet lag on the direction you are traveling, a temporary shift to earlier bedtime, or short-term help with sleep timing during a schedule change. It is not something to take every night indefinitely.
3. L-theanine
L-theanine is an amino acid found in tea. The evidence base is small but reasonably clean.
Rao 2015 and follow-up work found that 200 mg of L-theanine reduced subjective stress and improved measures of relaxation without sedation. In trials pairing L-theanine with caffeine, it consistently blunts the jitter and heart-rate response of the caffeine without killing the alertness benefit.
For sleep specifically, the direct evidence is thin — a handful of small trials show modest improvements in sleep quality, especially in participants with pre-bed stress or high caffeine intake. Mechanistically, L-theanine increases alpha wave activity, the pattern associated with relaxed wakefulness. That plausibly translates to easier sleep onset in people whose pre-bed anxiety is what is keeping them up.
Dose in the literature is typically 100 to 200 mg. Side effects are essentially none. If you drink caffeine late in the day and think it is affecting your sleep, L-theanine 30 to 60 minutes before bed is one of the better-supported options in this list.
4. Glycine
Glycine is the least-known of the four, and the case is not overwhelming, but it is real. Bannai 2012 and adjacent small Japanese trials found that 3 g of glycine taken before bed improved subjective sleep quality, particularly in participants reporting unsatisfactory sleep at baseline. Follow-up measures showed improvements in sleep onset and self-reported daytime fatigue the following day.
The trials are small and mostly from a single research group, which limits how much weight to put on the effect. Mechanistically, glycine is a fast inhibitory neurotransmitter and appears to modestly lower core body temperature, which lines up with the sleep-onset signal. Side effects at 3 g are essentially none.
The twenty (or so) without strong evidence
The rest of the aisle. In rough order of how often they show up in retail formulas.
Valerian. The most-studied herbal sleep aid, and the results are still messy. The Bent 2006 systematic review and later meta-analyses found the trials were heterogeneous and results inconsistent. Some show modest improvement in subjective sleep quality, others show nothing. The signal-to-noise is not compelling.
Chamomile. Traditional and pleasant, but the controlled trial data is weak. A few small trials show marginal effects on sleep quality, and none are powered to be convincing.
Passionflower. Very weak. A handful of tiny trials, no consistent effect.
GABA (oral). The neurotransmitter itself, sold as a supplement. Oral GABA does not meaningfully cross the blood-brain barrier. Any reported effects are almost certainly indirect via the gut. Skip.
5-HTP. A serotonin precursor. Evidence for sleep is indirect and mostly extrapolated from mood-disorder trials. Not directly supported.
Ashwagandha. Mostly studied for anxiety and cortisol, where the evidence is stronger. Salve 2019 and a few adjacent trials show a small sleep signal at 300 to 600 mg of a standardized extract, but the primary use case is anxiety, not sleep. If pre-bed anxiety is your issue, it is defensible. If it is not, the direct sleep case is weaker.
Tart cherry. Contains small amounts of endogenous melatonin. Trials show small subjective sleep improvements, but the effect is likely equivalent to a low-dose melatonin capsule at a much higher cost.
Lavender oil (oral, Silexan). Kasper 2010 and follow-ups showed benefit for generalized anxiety disorder. Sleep improvement is marginal and mostly secondary to the anxiety effect.
CBD. The most-hyped ingredient in the category. Evidence for sleep specifically is thin. Shannon 2019 was an open-label case series with no control — not proof of anything. Most reported effects are indirect via anxiety reduction. Product-quality problems are severe: independent lab testing has repeatedly found retail CBD products under-dosed, over-dosed, or contaminated relative to label.
Reishi (Ganoderma lucidum). Traditional sleep aid. Direct controlled human trial data for sleep is essentially absent.
Kava. Some evidence for anxiety. Hepatotoxicity concerns are real — case reports of severe liver injury led to bans in several countries. Not worth the risk profile for sleep.
Lemon balm. A handful of small trials in combination with valerian. Nothing convincing on its own.
Hops. Nearly always combined with valerian in trials. Independent effect unknown.
Magnolia bark. Marketed as anxiolytic. Human trial data is very limited.
Kanna (Sceletium tortuosum). Very few controlled trials, none focused on sleep.
Apigenin. Popularized in the podcast biohacking space. There are zero direct human trials of apigenin as a sleep aid. Mechanistic plausibility, no clinical evidence.
Phosphatidylserine. Some evidence for cortisol reduction post-exercise. No direct sleep trial data.
Taurine. Interesting biochemistry, no direct human sleep trials.
Tryptophan. Some older data suggesting mild sleep benefit at gram-scale doses. Superseded in practice by 5-HTP and melatonin, neither of which is a slam dunk either.
Inositol. Some anxiety data. No direct sleep data.
Theobromine. A cocoa alkaloid. Included in some "relaxation" formulas. No sleep evidence, and it is a mild stimulant.
That is roughly the field. Some of these may eventually show something in a well-powered trial. As of today, none of them have.
The supplement quality problem
Here is the part that gets glossed over. Even the four ingredients with evidence are only as useful as the pill actually contains what the label says. Supplement manufacturing is not FDA-regulated the way pharmaceuticals are.
Erland 2017 tested 31 melatonin supplements from Canadian retailers and found that 71 percent did not match label claim by more than 10 percent. Some contained as little as 17 percent of the labeled dose. Others contained more than 4 times the labeled dose. A subset contained measurable serotonin as a contaminant — which is not something you want.
That pattern is not unique to melatonin. Independent testing across multivitamin, herbal, and amino acid categories consistently finds meaningful discrepancies between label and content. If you cannot trust the dose, you cannot trust anything you infer from taking the product.
Three third-party certifications materially reduce this risk.
- USP Verified — the United States Pharmacopeia tests for identity, potency, purity, and manufacturing quality. The tightest of the three.
- NSF Certified for Sport — tests for label accuracy and banned substances. Used by professional athletes and generally worth the premium.
- ConsumerLab — an independent testing lab that publishes reports on retail supplements. Not a certification the brand applies for, but a useful reference for what actually matches its label.
If your supplement does not carry one of these marks and the brand does not publish a Certificate of Analysis from a real ISO-accredited lab, you are trusting a marketing claim.
How to actually test whether a supplement is doing anything
The one thing worse than a supplement that does not work is a supplement you cannot tell is not working. Given how modest the effect sizes are on the four defensible ingredients, running a real self-test matters more here than in almost any other category.
The protocol is standard: pick one supplement at a time, take it for at least two weeks nightly, measure something objective, then discontinue for two weeks and remeasure. Subjective sleep quality is unreliable — expectation effects are strong and last for weeks. Objective measurement is what makes the comparison honest.
The honest bottom line
Sleep supplements are a small tool in the sleep stack, and only four of them have evidence worth the shelf space. Magnesium glycinate is the safest and most defensible general-purpose pick. Low-dose melatonin is a circadian tool, not a nightly sedative, and the doses on most bottles are five to fifty times what the literature supports. L-theanine is the best-supported option if pre-bed stress or caffeine is your issue. Glycine is a reasonable low-risk try. Everything else on the shelf is either weakly supported, indirectly extrapolated, or unsupported by controlled human data.
Even the four are modest. They will not fix a sleep problem that is really about temperature, light, consistent bedtime, or an underlying condition. They will not out-perform the boring interventions that actually run this category — a dark, cool, quiet room and a bedtime you keep on weekends. Supplement money spent before those things are locked in is money spent in the wrong order.
If you do try any of these, buy from a brand with USP, NSF, or ConsumerLab documentation. Run a real two-week on, two-week off test with an objective tracker. And keep your expectations calibrated to the size of the effect the literature actually shows — a small nudge, not a transformation.
