Valerian Root for Sleep: An Old Remedy With a Thin Modern Evidence Base

Valerian (Valeriana officinalis) has been used for insomnia since ancient Greece and Rome — it’s one of the oldest sleep remedies still sold today, marketed on that history as much as on any clinical trial. The traditional-use story is real and long. The modern trial evidence, once you actually weigh it the way a clinical guideline committee does, is thinner than the history suggests — and the field’s own 2024 synthesis of the evidence reached a notably blunter verdict than the individual studies underneath it.

Here’s what the research says, and where it pulls in different directions depending on how you measure “worked.”

The short version

• The largest recent meta-analysis of RCTs (21 trials, 1,433 participants) found valerian had a small-to-moderate effect on self-reported sleep quality (Pittsburgh Sleep Quality Index score improvement, SMD = −1.21) and improved the odds of self-reported better sleep quality and duration. That sounds like a real effect.
• But when the same trials were checked with objective measurements — actigraphy, polysomnography, the tools that don’t rely on what someone remembers about how they slept — valerian showed no significant effect on total sleep time or sleep efficiency, and only one secondary measure (time spent in NREM stage 3 sleep) reached significance.
• A 2024 umbrella review — a review of reviews, pooling 8 systematic reviews and 5 meta-analyses of the whole valerian-for-insomnia literature — concluded there is “no evidence of efficacy” for treating insomnia, despite acknowledging the same subjective-improvement pattern above. That’s the field’s own highest-level synthesis being more skeptical than any single meta-analysis underneath it.
• The American Academy of Sleep Medicine’s 2017 clinical practice guideline explicitly recommends against using valerian for chronic insomnia in adults — the same guideline body that recommends against melatonin for the same use, for similar reasons: the trial base doesn’t clear the bar for a general treatment recommendation.
• Safety looks good short-term, but valerian has a real, well-documented withdrawal syndrome if stopped abruptly after regular use — a fact that sits awkwardly next to how it’s usually marketed as a gentle, non-habit-forming alternative to sleep medication.
• A specific methodological problem, not just a general “more research needed”: valerian has a strong, distinctive odor that multiple reviewers say makes properly blinding placebo-controlled trials difficult — meaning some of the subjective-improvement signal above may reflect participants correctly guessing which pill they got, not a pharmacological effect.

What valerian studies actually test

Valerian is used as dried root/rhizome extract, most commonly at 300–600 mg taken before bed, sometimes for as little as a single night and sometimes for several weeks. Trials vary widely in preparation (aqueous extract, ethanol extract, whole root), dose, and how “sleep quality” gets measured — a heterogeneity the meta-analyses themselves flag as their primary source of statistical noise.

The subjective-vs-objective gap is the central finding

The 2023 meta-analysis (Zhang, Xu, and colleagues, Current Sleep Medicine Reports) pooled 21 RCTs and found valerian improved subjective outcomes clearly: PSQI score (SMD = −1.21, 95% CI −1.92 to −0.51), self-reported sleep-quality improvement (RR = 1.37, 95% CI 1.13–1.68), and self-reported sleep duration (RR = 1.27, 95% CI 1.02–1.57). All three of those are things people reported about their own sleep, not measured directly.

For objective sleep parameters — the subset of included trials that used actigraphy or sleep-lab measurement — the only significant finding was increased time in NREM stage 3 sleep (SMD = 0.89, 95% CI 0.35–1.43), a specific, secondary sleep-architecture measure. Total sleep time and sleep efficiency, the outcomes most people actually care about, did not improve significantly on objective measurement in this pooled analysis.

That gap — real subjective improvement, not confirmed objectively — is exactly the pattern that shows up across herbal sleep aids generally, and it’s why we’re not rating valerian higher than Limited despite the PSQI numbers looking sizable on their own.

The umbrella review: a more skeptical verdict from a higher vantage point

Valente et al.’s 2024 umbrella review (European Neuropsychopharmacology) sits a level above individual meta-analyses — it pooled 8 systematic reviews, 5 of them with their own meta-analyses, to assess the whole body of evidence at once. Its stated conclusion: valerian shows no evidence of efficacy for treating insomnia, even while acknowledging that subjective sleep-quality measures improved in multiple of the underlying meta-analyses. We read this finding via search-result summaries and a research-database entry, not the full original paper, so we’re presenting its topline conclusion with that caveat rather than citing specific numbers from it we haven’t verified ourselves.

This is a genuinely useful case study in how evidence synthesis can disagree with itself depending on the bar being applied: a single meta-analysis pooling raw effect sizes can look favorable, while a review of reviews — weighing study quality, blinding, and consistency across the whole literature — reaches a flatly more skeptical conclusion from the same underlying trials.

A guideline body has already weighed in, and said no

NCCIH’s own fact sheet on valerian states plainly that “the evidence on whether valerian is helpful for sleep problems is inconsistent,” and cites the American Academy of Sleep Medicine’s 2017 clinical practice guideline, which recommends against using valerian for chronic insomnia in adults. This is the same guideline, and the same “recommends against” verdict, we already flagged for melatonin’s general-insomnia use in our melatonin article — a pattern worth noticing: two of the most commonly marketed “natural sleep aid” ingredients both fail to clear the bar for a formal clinical recommendation, for the same chronic-insomnia use case, from the same guideline body.

NCCIH is also explicit that there isn’t enough evidence to draw conclusions about valerian for anxiety, depression, premenstrual syndrome, dysmenorrhea, or stress — the other uses it’s traditionally promoted for beyond sleep. Three small studies suggest a possible benefit for menopause symptoms, but NCCIH says that’s not enough evidence to be certain either.

The withdrawal risk that cuts against the “gentle and natural” pitch

This is the complication we think is easiest to miss in valerian marketing, and worth surfacing directly rather than leaving in the safety section as a minor footnote. NCCIH’s fact sheet states that stopping valerian abruptly after chronic use can produce withdrawal symptoms — anxiety, irritability, heart disturbances, insomnia (the exact thing it’s taken for), and in rare cases hallucinations. That’s a real dependence-adjacent pattern, not a hypothetical one, and it sits uncomfortably next to how valerian is typically positioned in the market: as a mild, “not habit-forming” alternative to prescription sleep medication. We’re not aware of evidence that this withdrawal pattern is common or severe for most users at typical doses, but its existence at all is worth knowing before stopping valerian suddenly after weeks of regular use — taper rather than stop cold, and loop in a healthcare provider if you’ve been using it regularly for an extended period.

Safety and regulatory status

NCCIH describes valerian as generally safe for short-term use by most adults, with apparent safety at 300–600 mg/day for up to 6 weeks in the research reviewed; the safety of longer-term use is not established either way. Reported side effects include headache, stomach upset, mental dullness, excitability, uneasiness, and vivid dreams — a mixed bag that includes both sedation-type effects and, in some people, the opposite (excitability, uneasiness). Some people experience next-day grogginess, particularly at higher doses.

Because valerian may have a sedative effect, NCCIH advises against combining it with alcohol or other sedatives. In very rare cases, liver injury has been reported in people taking valerian, most often as part of combination products with other herbs rather than valerian alone — NCCIH is explicit that valerian’s long-term effect on liver function isn’t established either way. Little is known about safety in pregnancy or breastfeeding. As with every ingredient on this site, valerian is regulated as a dietary supplement, not a drug — it is not FDA-approved or FDA-reviewed for effectiveness before sale.

What we could not check

• We did not read the full text of the 2024 umbrella review (Valente et al.) — its topline conclusion is drawn from search-result summaries and a research-database entry, not the original paper, and is presented with that caveat rather than as a fully independently verified primary-source read.
• We did not independently verify the blinding-difficulty claim against a specific trial’s own methods section — it’s drawn from a clinical review (Taibi et al., Sleep Medicine Reviews, 2007) discussing the literature generally, not confirmed study-by-study.
• We did not test any product. We have no independent data on what’s actually in any specific valerian supplement sold today, including whether it’s standardized to valerenic acid content or what that would mean for comparability to the trial doses above.
• We did not evaluate valerian in combination with other herbs (hops, lemon balm, and similar pairings are common in commercial sleep products) — the evidence above is for valerian on its own, and combination-product evidence is a distinct question we haven’t reviewed here.

Who should talk to someone first

Anyone pregnant, breastfeeding, taking sedatives, benzodiazepines, or alcohol regularly, or managing a liver condition should talk to a healthcare provider before starting valerian, given the safety gaps NCCIH itself flags above. Anyone who has been taking valerian regularly for an extended period should taper rather than stop abruptly, given the documented withdrawal pattern. Nothing here is medical advice, and no one on our team is a clinician — stated plainly on our About page.

Our rating, and why

For sleep quality/insomnia: Limited. Under our evidence scale, this reflects a real, measurable subjective effect in pooled trial data that does not hold up as clearly under objective measurement, is undermined by a documented blinding-difficulty problem specific to this ingredient, and is explicitly rated “no evidence of efficacy” by the most recent umbrella-level synthesis of the field. A formal clinical guideline (AASM 2017) recommends against use for chronic insomnia. We’re not calling this “no effect” — the subjective data is real — but we’re not rating it above the bar the field’s own most careful reviewers have set.

For anxiety, menopause symptoms, PMS, and other traditional uses: Limited-to-Anecdotal. NCCIH is explicit that there isn’t enough evidence for anxiety, depression, PMS, or dysmenorrhea, and only a handful of small studies exist for menopause symptoms specifically. This is squarely the traditional-use-versus-controlled-evidence gap this piece set out to illustrate.

If a well-conducted, larger objective-measures trial changes this picture, or the umbrella review’s full methodology turns out to weight things differently than its topline conclusion suggests, this rating will be revisited.

---

Sources

1. National Center for Complementary and Integrative Health (NCCIH). Valerian: Usefulness and Safety. Fact sheet, last updated May 2025. https://www.nccih.nih.gov/health/valerian
2. Zhang X, Lu Y, Lv F, et al. Valerian for Insomnia on Subjective and Objective Sleep Parameters: a Meta-analysis of Randomized Controlled Trials. Current Sleep Medicine Reports. 2023;9:211-224. https://link.springer.com/article/10.1007/s40675-023-00259-4
3. Valente V, Machado D, Jorge S, Drake CL, Marques DR. Does valerian work for insomnia? An umbrella review of the evidence. European Neuropsychopharmacology. 2024;82:6-28. https://pubmed.ncbi.nlm.nih.gov/38359657/ (Read via search-result summaries and a research-database entry, not the full original paper — see “What we could not check.”)
4. Sateia MJ, Buysse DJ, Krystal AD, et al. Clinical practice guideline for the pharmacological treatment of chronic insomnia in adults: an American Academy of Sleep Medicine Clinical Practice Guideline. Journal of Clinical Sleep Medicine. 2017;13(2):307-349. (Same guideline cited in our melatonin article; recommends against valerian for chronic insomnia.)
5. Taibi DM, Landis CA, Petry H, Vitiello MV. A systematic review of valerian as a sleep aid: safe but not effective. Sleep Medicine Reviews. 2007;11(3):209-230. (Cited for the blinding/odor methodological limitation discussion.)

L-theanine is the amino acid in tea that supplement marketing credits for a very specific promise: calm focus, without the sedation of a sleep aid or the jitters of caffeine alone. It’s one of the milder ingredients we’ve reviewed on this site — the safety record is genuinely clean — but the newest and largest analysis of the trial evidence, published in July 2026, tells a more specific and slightly different story than “it calms you down.” The best-supported effect isn’t calm. It’s attention.

Here’s what the trials actually measured, including why the calm-focus story and the tested-and-confirmed story aren’t quite the same claim.

The short version

• The largest and most recent meta-analysis, pooling 31 randomized trials and 1,168 participants, found the single most robust, reproducible effect of L-theanine is on attention: a single 200 mg dose taken 30–60 minutes before cognitive testing significantly improved choice reaction time (SMD = 0.51). That’s a moderate-to-large effect by conventional standards.
• The stress-reduction effect that actually drives the “calm” marketing was modest (SMD = 0.31) and — the authors say this plainly — was largely driven by studies at high risk of bias. Strip those out and the calming claim gets considerably shakier.
• Anxiety effects were inconsistent and mostly not statistically significant, with one exception: a single trial in people with psychotic-disorder-related anxiety, at a much higher dose (400 mg/day for 8 weeks) than the typical single-dose studies. That’s a specific clinical population, not a general “feel less anxious” result.
• A possible antidepressant signal turned up (SMD = 0.69) after excluding one outlier study — interesting, but the authors themselves call it something that needs confirmation in better trials, not an established benefit.
• Safety was clean across the board: no serious adverse events reported in any of the 31 trials. This is one of the better safety profiles among the ingredients we’ve covered in this category.

Our evidence rating is split by claim rather than one line — see the ratings section for why.

What “L-theanine” studies actually test

L-theanine is a non-protein amino acid that occurs naturally in tea (Camellia sinensis) and, in much smaller amounts, in some mushrooms. Trials typically test a single oral dose — most commonly 200 mg, sometimes up to 250–400 mg — either alone or paired with caffeine, and measure outcomes anywhere from 30 minutes to a few hours later. That single-dose design matters: it’s a different kind of evidence than a months-long trial of daily supplementation, and most of the strongest L-theanine data is the single-dose kind.

The best-supported finding: attention, not calm

The July 2026 systematic review and meta-analysis by Gerolymos, Saddier, Boyer, Fond and colleagues, published in Molecular Psychiatry, pooled 31 randomized, placebo-controlled trials (n = 1,168) in both healthy and clinical populations. Its primary, best-replicated result: a single 200 mg dose of L-theanine, taken 30 to 60 minutes before cognitive testing, significantly improved choice reaction time — a standard measure of attention — with a standardized mean difference of 0.51 (95% CI: 0.25–0.77).

That’s the finding with the least ambiguity in the whole paper. It’s also, notably, not the claim most L-theanine products actually lead with.

The stress-reduction claim: real, but built on shakier ground

The same meta-analysis’s primary outcome was actually the acute effect of a single dose on stress in healthy adults — and here the result was real but modest (SMD = 0.31), and the authors explicitly note it was largely influenced by studies carrying a high risk of bias. In plain terms: the calming effect that’s central to how L-theanine is marketed is the weakest-supported of the effects this meta-analysis measured, not the strongest.

Anxiety: mostly not there, with one narrow exception

Across the pooled trials, effects on anxiety were inconsistent and non-significant, with a single exception: one trial in people with anxiety related to a psychotic disorder, using a considerably higher and more sustained dose — 400 mg per day for 8 weeks, not a single 200 mg dose (SMD = 0.54). That’s a specific clinical population on a specific dosing schedule, and it shouldn’t be read as evidence that L-theanine reduces general anxiety in a healthy adult taking an occasional 200 mg capsule.

A possible antidepressant signal, honestly caveated

After excluding one outlier study, the meta-analysis found a significant reduction in depressive symptoms following a single dose in healthy individuals (SMD = 0.69, 95% CI: 0.13–1.25, with low heterogeneity once that outlier was removed). The excluded outlier is worth naming: it’s the widely cited Haskell et al. 2008 study that tested L-theanine, caffeine, and their combination together on cognition and mood — precisely the kind of combination-product study this article set out to interrogate. The authors of the 2026 meta-analysis themselves describe the antidepressant finding as warranting confirmation in high-quality trials, not as a settled result — and we’re treating it the same way.

The caffeine-combination question

Many “focus” products pair L-theanine with caffeine rather than selling it alone, and the classic trial behind that pairing — Haskell et al., 2008, in Biological Psychology — specifically tested L-theanine, caffeine, and the combination against each other, not just against placebo. That’s a meaningfully different comparison than most of the trials in the 2026 meta-analysis, which largely tested L-theanine against placebo alone.

We were not able to determine from the meta-analysis’s published abstract how many of the 31 pooled trials tested L-theanine alone versus in combination with caffeine, which matters for a product marketed as delivering “calm focus” through the combination specifically. That’s a real gap in what we can tell you from this research pass — flagged in the notes below rather than guessed at.

Where the dose in your tea falls short of the dose in the trials

A cup of tea is not equivalent to an L-theanine supplement, in either direction of comparison but especially on dose. Peer-reviewed measurements of brewed tea find L-theanine content varies widely by tea type and brewing method — commonly in the range of roughly 5 to 25 mg per 200 mL cup, depending on the tea and how it’s steeped, with brewing time being a major factor in how much theanine actually ends up in the cup. The single-dose trials behind the attention finding above used 200 mg — something like eight to forty cups of tea’s worth in one sitting, not remotely realistic. If a product’s marketing implies “it’s just like the calm you get from tea,” the dose math doesn’t support treating the two as comparable experiences.

Safety and regulatory status

Across all 31 trials in the 2026 meta-analysis — 1,168 participants combined — no serious adverse events were reported for L-theanine compared to placebo, and dropout rates and reasons were comparable between groups. That’s a genuinely reassuring safety signal, and a meaningfully cleaner profile than some of the other ingredients in this category: it doesn’t carry ashwagandha’s documented liver-injury signal or melatonin’s pediatric-ingestion problem.

On regulatory status: a specific branded, patented L-theanine extract (Suntheanine, from Taiyo Kagaku) received an FDA “letter of no objection” to a self-affirmed Generally Recognized As Safe (GRAS) determination for use in foods and beverages, reportedly at levels up to 250 mg per serving. We’re sourcing that GRAS detail from industry and trade-press coverage of the FDA notification, not from a primary FDA document we read directly — worth knowing if you’re citing the specific 250 mg figure elsewhere. GRAS status is a food-additive safety determination, not an evaluation of whether the ingredient works for any of the claims above; the two questions are regulated and assessed completely separately, a distinction worth keeping in mind given how often “FDA-recognized” gets used in marketing as if it answered the efficacy question too.

What we could not check

• We could not determine the alone-versus-combined-with-caffeine breakdown of the 31 pooled trials — flagged above, and relevant to any product that markets the caffeine+theanine pairing specifically rather than L-theanine alone.
• We did not independently read the full text of the 2026 meta-analysis — the figures above come from its published abstract, which is unusually detailed and specific for an abstract, but we have not verified the underlying forest plots or individual trial data ourselves.
• We did not verify the GRAS/FDA “letter of no objection” claim against a primary FDA document. It’s sourced from trade-press and industry coverage of the filing, consistent with how this kind of regulatory correspondence is typically reported, but not confirmed against FDA’s own records directly.
• We did not test any product. We have no independent data on what’s actually in any specific L-theanine product sold today.
• We did not evaluate long-term daily supplementation. Nearly all of the strongest evidence here is single-dose; we found no comparably strong evidence pool for what happens with ongoing daily use over months.

Who should talk to someone first

L-theanine’s safety record in this literature is clean, and NCCIH-style general dietary-supplement caution still applies: talk to a healthcare provider before adding any supplement if you’re pregnant, breastfeeding, or taking medication that affects blood pressure, since amino acid supplements can theoretically interact with blood-pressure-lowering effects at high doses — this is a general caution rather than a documented problem specific to the trials reviewed here, and we want to be clear about that distinction rather than implying a risk we haven’t actually seen evidence for.

Nothing here is medical advice, and no one on our team is a clinician. That’s stated plainly on our About page.

Our rating, and why

For attention and cognitive performance from a single dose: Moderate. Under our evidence scale, this reflects the most consistently replicated finding in the largest available meta-analysis — a specific, moderate-to-large effect size with a reasonably tight confidence interval.

For acute stress reduction — the claim closest to how this ingredient is actually marketed: Limited. The pooled effect is real but modest, and the meta-analysis’s own authors attribute much of it to high-risk-of-bias studies. That’s not nothing, but it’s a considerably weaker foundation than the attention finding, and we’re not going to let the more flattering “calm” framing outrank what the data actually supports best.

For anxiety as a general claim: Limited-to-Anecdotal. The evidence is inconsistent and largely non-significant outside one narrow clinical population at a different dose and duration than most products use.

The possible antidepressant signal is not rated at all — it’s a single finding pending confirmation, not yet a claim this site is prepared to stand behind either way.

If the research develops further — particularly on the caffeine-combination question and on longer-term daily use — these ratings will change, and we’ll say so on this page when they do.

---

Sources

1. Gerolymos C, Saddier E, Boyer L, Fond G, et al. Cognitive and affective effects of L-Theanine: a systematic review and meta-analysis of 31 randomized trials. Molecular Psychiatry. 2026 (published online July 6, 2026). https://www.nature.com/articles/s41380-026-03727-9
2. Haskell CF, Kennedy DO, Milne AL, Wesnes KA, Scholey AB. The effects of l-theanine, caffeine and their combination on cognition and mood. Biological Psychology. 2008;77(2):113-122. (Cited as the excluded-outlier study in source 1’s depression sensitivity analysis.)
3. FDA “Letter of No Objection” to self-affirmed GRAS status for Suntheanine® (L-theanine), Taiyo Kagaku Co. — reported via industry/trade press, 2007. (Sourced secondhand; not independently verified against a primary FDA document — see editorial notes.)

Melatonin is the most widely used sleep supplement in the United States, sold as tablets, capsules, gummies, and sprays in doses from under 1 mg to 10 mg or more. It’s also one of the more misunderstood ones: it’s a hormone, not a sedative, and the research behind it splits cleanly into two different questions that get marketed as one.

Here’s what we found separating them, including where the evidence for “just take less” is more complicated than it first appears.

The short version

• Melatonin has its best evidence for circadian-timing problems — jet lag and delayed sleep-wake phase disorder (a body clock that runs hours later than a normal schedule). For delayed sleep-wake phase disorder, the American Academy of Sleep Medicine (AASM) actually recommends melatonin, timed to the body clock rather than to bedtime — though it’s a weak recommendation, with acknowledged uncertainty about whether benefits outweigh harms.
• For general chronic insomnia unrelated to circadian timing — which is how melatonin is mostly bought and sold in the US — the same AASM guideline recommends against using it. That’s also a weak recommendation, reflecting evidence the guideline panel judged too thin to support routine use either way.
• The largest meta-analysis of melatonin for sleep pooled 19 trials in 1,683 people and found modest benefits: falling asleep about 7 minutes faster and sleeping about 8 minutes longer than placebo. But that analysis mixed insomnia and circadian-disorder trials together, and — a complication we’re not going to smooth over — it found higher doses and longer treatment duration were associated with larger effects, not smaller ones. That cuts against a simple “lower doses work just as well” story.
• The cleaner case for lower doses comes from a different, narrower body of research: studies measuring what dose of melatonin actually reproduces the body’s own natural nighttime blood levels. That work, done specifically in adults 55 and older, concludes the lowest effective dose is best because higher doses push blood melatonin well above what the body would ever produce on its own.
• Independent of effectiveness, there’s a real quality-control problem: analyses of commercial melatonin products have repeatedly found labeled and actual content don’t match, sometimes by a wide margin, and gummies in particular have driven a sharp rise in accidental pediatric ingestions serious enough to need emergency care.

Our evidence rating for melatonin is not one verdict — this is a case where the indication matters more than the ingredient. See the ratings section near the end.

What melatonin is actually studied for

Melatonin is a hormone the brain produces in response to darkness, largely from the pineal gland, and it’s central to timing the body’s circadian rhythm rather than to sedation itself. Most of the higher-quality trial evidence targets problems with the timing of sleep, not sleep in general.

Jet lag

Crossing multiple time zones disrupts the body clock’s alignment with the local light-dark cycle, causing the cluster of symptoms known as jet lag: poor sleep, daytime fatigue, impaired functioning, and sometimes digestive upset.

According to a synthesis of two mid-sized systematic reviews (2010 and 2014) summarized by the National Center for Complementary and Integrative Health (NCCIH), four studies totaling 142 travelers found melatonin outperformed placebo on jet lag symptoms after eastward flights, and a separate study of 234 travelers found low-quality evidence of a sleep-quality benefit on eastward flights specifically. Two further studies totaling 90 travelers found a benefit after westward flights. We’re relying on NCCIH’s synthesis of the underlying Cochrane review here rather than having independently read the full Cochrane text ourselves — flagged below in the editorial notes.

Delayed sleep-wake phase disorder (DSWPD)

People with DSWPD have body clocks that run persistently later than a conventional schedule — typically unable to fall asleep before 2–6 a.m. and preferring to wake between 10 a.m. and 1 p.m. This is a circadian rhythm disorder, not garden-variety insomnia, and it’s diagnosed rather than self-identified.

This is the specific condition where the evidence is strong enough that a clinical body has weighed in affirmatively. The 2015 AASM clinical practice guideline for circadian rhythm sleep-wake disorders recommends melatonin, timed to the individual’s circadian phase rather than to a fixed bedtime, for DSWPD — a weak recommendation, with the guideline itself flagging uncertainty about whether the benefits clearly outweigh the risks. A 2018 randomized trial of 307 people with DSWPD (published after that guideline) found that melatonin taken an hour before the desired bedtime, combined with a fixed wake time, produced meaningful improvements: falling asleep about 34 minutes earlier on average, along with better sleep in the first third of the night and better daytime functioning.

General chronic insomnia — the use case it’s mostly sold for

This is where the marketing and the guidance diverge most. Two major clinical practice guidelines have weighed in on melatonin for ordinary chronic insomnia not tied to circadian timing: the 2017 AASM guideline and the 2016 American College of Physicians guideline. Both concluded there isn’t enough strong evidence on melatonin’s effectiveness or safety for chronic insomnia to recommend its use — the AASM guideline’s language is a recommendation against using it, again graded weak.

The largest single meta-analysis on melatonin and sleep, published by Ferracioli-Oda, Qawasmi and Bloch in PLoS ONE in 2013, pooled 19 randomized placebo-controlled trials covering 1,683 subjects — 14 trials in insomnia, 4 in DSWPD, and 1 in REM sleep behavior disorder, combined into a single analysis rather than reported separately by condition. Pooled across that mixed population, melatonin reduced sleep onset latency by 7.06 minutes (95% CI: 4.37–9.75, p<0.001), increased total sleep time by 8.25 minutes (95% CI: 1.74–14.75, p=0.013), and modestly improved a composite sleep-quality measure (standardized mean difference 0.22, 95% CI: 0.12–0.32, p<0.001). The authors’ own conclusion: the effects are real but modest, and don’t seem to fade with continued use — but they’re smaller than what’s typically seen from prescription insomnia medications. One author disclosed research fellowship funding tied to Eli Lilly, a pharmaceutical company that sells prescription sleep and psychiatric medications; the paper’s own funding statement says none of its funders, including that one, had a role in study design, analysis, or the decision to publish. We mention it because our Editorial Policy says we will, not because it changes the numbers above.

The complication we want to be direct about: this same meta-analysis found, via meta-regression, that trials using higher doses and longer treatment duration reported larger effects on sleep latency (trend-level, p=0.05) and total sleep time (p=0.007) — not smaller ones. Dose and duration had no significant effect on the sleep-quality measure either way. That finding runs against a simple “lower doses are just as effective for insomnia” narrative, and we’re not going to pretend it doesn’t exist because it complicates this article’s own title.

The dose-mismatch question — where the evidence for “less is more” actually comes from

The idea that commercial melatonin doses run far above what’s useful is common in health journalism, and it isn’t baseless — but the strongest version of that case isn’t about insomnia outcomes. It’s about matching the hormone’s own physiology.

A 2014 systematic review by Vural, van Munster and de Rooij in Drugs & Aging set out specifically to determine what dose of exogenous melatonin best reproduces natural nighttime melatonin levels in adults aged 55 and older. Its conclusion: clinicians should use the lowest possible dose of an immediate-release formulation, because that best mimics the body’s own physiological circadian rhythm — and because higher doses risk prolonged, supra-physiological blood levels that go well beyond what the body would ever produce naturally. That’s a real and citable basis for “less is often enough,” but it’s a conclusion about matching physiology in an older-adult population specifically, not a head-to-head finding that higher doses fail to help with sleep symptoms generally — the Ferracioli-Oda meta-regression above suggests the opposite for insomnia-latency outcomes.

Set against that physiological argument, retail melatonin in the US commonly comes in doses at or above what circadian-timing protocols actually use. A 2023 study by Cohen and colleagues, published in JAMA, measured the actual melatonin content of 25 commercially available gummy products in the US and found actual per-serving content ranging from 1.3 mg to 13 mg — and in 22 of the 25 products, that measured amount didn’t match the label, most often running higher (74% to 347% of the labeled amount). For comparison, the jet lag and DSWPD protocols described above mostly used doses in the roughly 0.5–5 mg range.

So the honest version of the dose story has two separate strands: a physiological argument (from research in older adults) that lower doses better match the body’s own hormone levels and avoid unnecessary supra-physiological exposure, and a completely separate, unrelated problem that a meaningful share of commercial products don’t reliably deliver the dose printed on the label in the first place — which makes “just take a lower dose” harder to act on than it sounds, since the label may not be trustworthy regardless of what number is printed on it.

What we could not check

• We did not independently read the full Cochrane jet lag review. The jet lag figures above come from NCCIH’s published synthesis of that review, not from our own read of the Cochrane text — a lower bar of verification than we held the magnesium article to, and worth naming rather than hiding.
• We did not test any product. We have no independent data on what’s actually in any specific melatonin product sold today; the label-accuracy figures above are from the cited studies’ own product samples (Erland & Saxena’s 2017 sample of 31 products; Cohen et al.’s 2023 sample of 25 gummies), not a survey we ran.
• We could not resolve the dose question into a single number. The circadian-timing literature and the general-insomnia meta-analysis point in different directions on dose, and we don’t think a single “ideal mg” figure can honestly be extracted from the evidence as it currently stands.
• We did not evaluate children’s dosing. NCCIH’s own review flags substantial uncertainty about melatonin’s effects on children’s hormonal development; that’s a distinct topic from the adult-focused claims in this piece and would need its own dedicated treatment, not a subsection here.

Safety and quality-control issues, separate from effectiveness

Product content doesn’t reliably match the label. A 2017 analysis by Erland and Saxena in the Journal of Clinical Sleep Medicine tested 31 melatonin products from 16 brands and found that more than 70% fell outside a 10% margin of their labeled content, with actual content ranging from 83% below to 478% above the label claim; lot-to-lot variability within the same product reached as much as 465%. The same study found unlabeled serotonin in 26% of tested products, mostly those combining melatonin with herbal extracts.

Pediatric ingestion has risen sharply, and gummies are a specific driver. Per two CDC MMWR reports, annual pediatric melatonin ingestions reported to US poison control centers rose from 8,337 in 2012 to 52,563 in 2021 — a 530% increase — with hospitalizations and serious outcomes also increasing over that period. A more recent CDC report estimated roughly 11,000 emergency department visits during 2019–2022 for unsupervised melatonin ingestion by children aged 5 and under, disproportionately involving flavored gummy products; in more than a third of solid-dose visits, the child had ingested 10 or more units.

Regulatory status. In the US, melatonin is regulated as a dietary supplement, meaning it faces substantially less premarket scrutiny from the FDA than a prescription or over-the-counter drug — it does not need to demonstrate effectiveness or a validated manufacturing standard before going on shelves. NCCIH notes that in several other countries, melatonin is available only by prescription and is regulated as a drug.

Who should talk to someone first

Melatonin interacts with some medications and carries specific caution for a few groups. Per NCCIH, people taking blood thinners or with epilepsy should be under medical supervision if using melatonin supplements. There’s a documented lack of research on melatonin’s safety in pregnancy and breastfeeding. Older adults may retain melatonin longer than younger adults, raising the risk of daytime drowsiness, and the 2015 AASM circadian-rhythm guideline specifically recommends against melatonin use in people with dementia.

For children, NCCIH’s own guidance is to talk to a pediatrician before giving a child melatonin for sleep — not because short-term use at normal doses appears unsafe in the studies available, but because there are meaningfully fewer studies, and because melatonin is a hormone with theoretical (not yet resolved) implications for puberty and other hormonal development.

Nothing here is medical advice, and no one on our team is a clinician. That’s stated plainly on our About page.

Our rating, and why

We’re not giving melatonin one rating, because the evidence genuinely differs by what it’s being used for — collapsing that into a single badge would misrepresent both halves.

For circadian-timing indications (jet lag, delayed sleep-wake phase disorder): Moderate. Under our evidence scale, this reflects a real body of trial evidence and, for DSWPD specifically, an affirmative (if weak) recommendation from a major clinical guideline body. It is not “Strong” — the AASM’s own recommendation is graded weak, reflecting real uncertainty about the balance of benefit and harm, and we relied on a secondary synthesis rather than a first-hand read of the full Cochrane jet lag review.

For general chronic insomnia unrelated to circadian timing — the use case behind most retail marketing: Limited. The largest pooled analysis shows a real but modest effect, and two major clinical guidelines (AASM 2017, ACP 2016) recommend against routine use given the overall strength of the evidence. This is also the rating that best reflects the article’s own most important caveat: the same meta-analysis that shows a modest benefit also shows that benefit trending larger, not smaller, at higher doses — which undercuts the tidy “you just need less” framing this piece set out to test.

If the research on either question develops further, these ratings will change, and we’ll say so on this page when they do.

---

Sources

1. National Center for Complementary and Integrative Health (NCCIH). Melatonin: What You Need To Know. Last updated May 2024. https://www.nccih.nih.gov/health/melatonin-what-you-need-to-know
2. Ferracioli-Oda E, Qawasmi A, Bloch MH. Meta-Analysis: Melatonin for the Treatment of Primary Sleep Disorders. PLoS ONE. 2013;8(5):e63773. https://doi.org/10.1371/journal.pone.0063773
3. Auger RR, Burgess HJ, Emens JS, et al. Clinical practice guideline for the treatment of intrinsic circadian rhythm sleep-wake disorders: advanced sleep-wake phase disorder (ASWPD), delayed sleep-wake phase disorder (DSWPD), non-24-hour sleep-wake rhythm disorder (N24SWD), and irregular sleep-wake rhythm disorder (ISWRD). Journal of Clinical Sleep Medicine. 2015;11(10):1199-1236.
4. Sateia MJ, Buysse DJ, Krystal AD, et al. Clinical practice guideline for the pharmacologic treatment of chronic insomnia in adults: an American Academy of Sleep Medicine clinical practice guideline. Journal of Clinical Sleep Medicine. 2017;13(2):307-349.
5. Qaseem A, Kansagara D, Forciea MA, et al. Management of chronic insomnia disorder in adults: a clinical practice guideline from the American College of Physicians. Annals of Internal Medicine. 2016;165(2):125-133.
6. Vural EMS, van Munster BC, de Rooij SE. Optimal dosages for melatonin supplementation therapy in older adults: a systematic review of current literature. Drugs & Aging. 2014;31(6):441-451.
7. Erland LAE, Saxena PK. Melatonin natural health products and supplements: presence of serotonin and significant variability of melatonin content. Journal of Clinical Sleep Medicine. 2017;13(2):275-281. https://jcsm.aasm.org/doi/10.5664/jcsm.6462
8. Cohen PA, Avula B, Wang Y-H, et al. Quantity of Melatonin and CBD in Melatonin Gummies Sold in the US. JAMA. 2023;329(16):1401-1402.
9. Lelak K, Vohra V, Neuman MI, et al. Pediatric melatonin ingestions — United States, 2012–2021. MMWR Morbidity and Mortality Weekly Report. 2022;71(22):725-729.
10. Freeman DI, Lind JN, Weidle NJ, et al. Notes from the field: emergency department visits for unsupervised pediatric melatonin ingestion — United States, 2019–2022. MMWR Morbidity and Mortality Weekly Report. 2024;73(9):215-217.

Ashwagandha (Withania somnifera) has gone from a centuries-old Ayurvedic herb to one of the fastest-growing ingredients in US stress-relief supplements, and unlike some of the ingredients we’ve covered on this site, it actually has a reasonably consistent body of small human trials behind the stress claim. That doesn’t mean the picture is simple — the trial base is thin by the standards of a mainstream medication, dominated by short studies, and there’s a real, recently documented gap between what it does to a stress hormone and what people actually report feeling.

Here’s what we found, including a safety issue — rare liver injury — that gets far less attention than the stress claim does.

The short version

• The best-quality synthesis of the evidence, a 2022 meta-analysis pooling 12 trials in 1,002 people, found ashwagandha significantly reduced both stress scores (SMD −1.75) and anxiety scores (SMD −1.55) compared to placebo — large effect sizes by conventional statistical standards.
• But NCCIH, reviewing the same broad literature, describes the evidence as reasonably supportive for stress, while calling the evidence on anxiety specifically “unclear.” Those are different claims, and the marketing usually doesn’t distinguish them.
• A newer 2025 meta-analysis of 8 trials found something worth taking seriously: ashwagandha produced a statistically significant drop in measured cortisol, but no significant improvement in how stressed people said they felt on a standard perceived-stress questionnaire. A hormone moving in the right direction didn’t reliably translate into the subjective outcome the supplement is sold on.
• Several of the most-cited trials use a specific patented extract (KSM-66), manufactured and commercially promoted by the ingredient’s own supplier — a funding pattern worth knowing about, which we were only able to partially verify ourselves.
• Ashwagandha carries a real, if rare, safety signal that’s easy to miss in stress-relief marketing: documented cases of clinically apparent liver injury, including a small number of fatal cases and liver transplants, tracked by the NIH’s own drug-induced liver injury registry.

Our evidence rating is Moderate for stress as a broad outcome, but drops toward Limited for anxiety treated as its own distinct claim — see the ratings section for why we’re not collapsing these into one line.

What the trials actually measured

Most modern ashwagandha trials for stress use a standardized root extract, dosed once or twice daily, over 8 to 12 weeks, in adults who are healthy but self-report elevated stress — not people with a diagnosed anxiety disorder. That population detail matters for reading the results honestly.

The pooled evidence

A 2022 systematic review and meta-analysis by Akhgarjand and colleagues, published in Phytotherapy Research, pooled 12 randomized controlled trials covering 1,002 participants aged 25 to 48. Compared to placebo, ashwagandha supplementation was associated with a significant reduction in stress scores (standardized mean difference −1.75, 95% CI: −2.29 to −1.22) and anxiety scores (SMD −1.55, 95% CI: −2.37 to −0.74), both reported at p = .005. Those are large effect sizes as statistics go — but the same paper reported very high heterogeneity between trials (I² in the 83–94% range), meaning the individual studies varied a great deal in design, population, and result, which tempers how much confidence a single pooled number deserves.

NCCIH’s own review of the broader literature lands in a similar but more cautious place: it states research shows some ashwagandha preparations “may be effective for insomnia and stress,” while separately noting the evidence for anxiety specifically “is unclear.” That’s a meaningful distinction the Akhgarjand pooled anxiety number tends to flatten — a systematic review can show a significant pooled effect on an anxiety scale while the underlying evidence for anxiety as a clinical claim still doesn’t clear the bar a regulatory-adjacent body considers convincing.

The complication we want to be direct about

A newer 2025 systematic review and meta-analysis (Albalawi et al.), pooling 8 randomized trials in 488 adults, measured both cortisol — a bodily stress-hormone marker — and scores on the Perceived Stress Scale (PSS), a standard self-report questionnaire. Ashwagandha produced a statistically significant reduction in cortisol (roughly −1.16 µg/dL, 95% CI: −1.64 to −0.69, p < 0.001). But the same analysis found no significant effect on PSS scores (SMD −0.355, 95% CI: −1.188 to 0.47, p = 0.40) — people’s own reported sense of stress didn’t move in a way that cleared statistical significance, even though their cortisol did.

That’s a real tension worth sitting with rather than resolving in the supplement’s favor: a hormone marker moving favorably is not the same thing as a person feeling less stressed, and the second is what someone buying this product actually wants. We’re flagging it plainly rather than leading with the more flattering 2022 pooled-anxiety number and leaving this one for a footnote.

Where the money comes from — a disclosure we could only partly verify

Several of the most frequently cited trials in this space, including the foundational 2012 study by Chandrasekhar, Kapoor and Anishetty in the Indian Journal of Psychological Medicine (600 mg/day of a high-concentration extract, reducing both stress/anxiety scores and serum cortisol versus placebo), use KSM-66 — a specific patented ashwagandha extract manufactured and commercially promoted by its supplier, Ixoreal Biomed.

Per our Editorial Policy, we disclose funding and commercial ties in the research we cite. In this case, we were not able to independently confirm the funding and conflict-of-interest statement in the primary 2012 paper itself — the full text was not accessible to us during this research pass. What we can say with confidence: KSM-66 is a branded ingredient with an active commercial marketing operation built substantially around this and similar trials, which is a reason for extra scrutiny regardless of the specific funding line in any one paper. We’re naming that limitation rather than either asserting a funding conflict we haven’t verified or omitting the concern entirely.

Safety: the liver injury signal that gets little attention

This is the part of the ashwagandha story that rarely makes it into stress-relief marketing.

The NIH’s LiverTox database — a clinical reference on drug- and supplement-induced liver injury maintained by the National Institute of Diabetes and Digestive and Kidney Diseases — rates ashwagandha at Likelihood Score B: a likely cause of clinically apparent liver injury, based on an accumulating case-report literature rather than controlled-trial data (since trials are generally too small and short to catch a rare event). The typical pattern: liver injury presenting 2 to 12 weeks after starting the supplement, usually cholestatic or mixed in pattern, with jaundice and itching. Most cases resolve within one to five months of stopping the product, but LiverTox documents rare fatal cases and at least one instance requiring emergency liver transplantation, disproportionately in people with pre-existing liver disease. A 2023 case series review found roughly 20+ published cases at that point, still a small fraction of ashwagandha’s overall use, but a real and non-theoretical risk.

NCCIH’s own fact sheet corroborates the broad strokes: ashwagandha “may be safe” short-term (up to about 3 months), with insufficient data on long-term safety, and explicitly lists rare liver injury among the known concerns — alongside drowsiness, stomach upset, diarrhea, and vomiting as more common, milder effects.

Thyroid effects are a separate, documented concern. LiverTox’s bibliography includes a published case of thyrotoxicosis (an overactive thyroid) in a previously healthy woman that developed after starting, and then increasing the dose of, an ashwagandha preparation — resolving on its own after she stopped. NCCIH’s fact sheet independently advises against ashwagandha for people with thyroid disorders, autoimmune conditions, or before surgery, and flags interactions with diabetes medications, blood pressure medications, immunosuppressants, sedatives, and anticonvulsants.

Pregnancy and hormone-sensitive conditions. NCCIH states ashwagandha should be avoided during pregnancy and while breastfeeding. Because it may raise testosterone levels, NCCIH also advises against its use by people with hormone-sensitive prostate cancer.

What we could not check

• We did not verify the funding disclosure of the foundational KSM-66 trial (Chandrasekhar 2012) — flagged plainly above rather than assumed either way.
• We did not independently read the full text of either meta-analysis (Akhgarjand 2022, Albalawi 2025) — the effect sizes and confidence intervals above are drawn from the papers’ own published abstracts, cross-checked across multiple citations of each, not from a full-text read of methods and individual trial data.
• We could not verify a specific dose-response claim. Some secondary sources describing the Akhgarjand 2022 meta-regression cited an anxiety dose-response figure that looked implausibly high compared to any individual trial’s actual dosing; rather than risk repeating a garbled number, we’ve left it out of this piece entirely.
• We did not test any product. We have no independent data on what’s actually in any specific ashwagandha product sold today, including whether it matches its labeled extract type or concentration.
• We did not evaluate long-term safety. Both NCCIH and LiverTox note that data beyond roughly 3 months of continuous use is limited.

Who should talk to someone first

Given the documented, if rare, liver injury signal, anyone with existing liver disease — including cirrhosis or chronic hepatitis — should talk to a doctor before using ashwagandha, and LiverTox explicitly notes it should be avoided in people with cirrhosis or advanced chronic liver disease. The same applies to anyone with a thyroid disorder or autoimmune condition, anyone scheduled for surgery, anyone taking diabetes medication, blood pressure medication, immunosuppressants, sedatives, or anticonvulsants, and anyone who is pregnant, breastfeeding, or has hormone-sensitive prostate cancer.

If new, unexplained fatigue, itching, dark urine, or yellowing of the skin or eyes appears while taking ashwagandha, LiverTox’s own case reports suggest that’s a reason to stop the product and seek medical attention rather than wait it out.

Nothing here is medical advice, and no one on our team is a clinician. That’s stated plainly on our About page.

Our rating, and why

For stress as a broad outcome: Moderate. Under our evidence scale, this reflects a real, if heterogeneous, body of randomized trial evidence — 12 trials and over 1,000 participants in the largest pooled analysis, with a consistent direction of effect across most individual studies, even though the trials vary considerably in quality and the pooled statistics show high heterogeneity.

For anxiety treated as a distinct, separate claim: Limited. NCCIH’s own assessment — “unclear” — is more conservative than the pooled statistical significance in the broader meta-analyses would suggest on its own, and we’re deferring to that more cautious read rather than the more flattering summary statistic, consistent with this site’s standing rule against rating inflation.

On the cortisol-versus-felt-stress question specifically: we’re treating the 2025 finding (cortisol down, perceived stress unchanged) as an open, unresolved tension rather than folding it into either rating — it’s exactly the kind of result that should make a reader skeptical of any product claim that leans only on “lowers cortisol” language, since lowering cortisol did not, in that analysis, reliably make people feel less stressed.

If the research develops further — particularly longer trials, or work that resolves the cortisol/perceived-stress gap — this rating will change, and we’ll say so on this page when it does.

---

Sources

1. National Center for Complementary and Integrative Health (NCCIH). Ashwagandha: Usefulness and Safety. Last updated March 2023. https://www.nccih.nih.gov/health/ashwagandha
2. LiverTox: Clinical and Research Information on Drug-Induced Liver Injury. Ashwagandha. National Institute of Diabetes and Digestive and Kidney Diseases, NIH. Last updated December 3, 2024. https://www.ncbi.nlm.nih.gov/books/NBK548536/
3. Akhgarjand C, Asoudeh F, Bagheri A, et al. Does Ashwagandha supplementation have a beneficial effect on the management of anxiety and stress? A systematic review and meta-analysis of randomized controlled trials. Phytotherapy Research. 2022;36(11):4115-4124. https://onlinelibrary.wiley.com/doi/10.1002/ptr.7598
4. Albalawi AA, et al. Dual impact of Ashwagandha: Significant cortisol reduction but no effects on perceived stress — A systematic review and meta-analysis. 2025. https://journals.sagepub.com/doi/abs/10.1177/02601060251363647
5. Chandrasekhar K, Kapoor J, Anishetty S. A prospective, randomized double-blind, placebo-controlled study of safety and efficacy of a high-concentration full-spectrum extract of ashwagandha root in reducing stress and anxiety in adults. Indian Journal of Psychological Medicine. 2012;34(3):255-262.
6. Björnsson HK, Björnsson ES, Avula B, et al. Ashwagandha as a cause for liver injury. Liver International. 2020;40:2035-2036. (Cited via LiverTox, source 2.)
7. Philips CA, Valsan A, Theruvath AH, et al. Ashwagandha-induced liver injury — A case series from India and literature review. Hepatology Communications. 2023;7:e0270. (Cited via LiverTox, source 2.)
8. van der Hooft CS, Hoekstra A, Winter A, de Smet PA, Stricker BH. [Thyrotoxicosis following the use of ashwagandha]. Ned Tijdschr Geneeskd. 2005;149:2637-2638. (Cited via LiverTox, source 2; original in Dutch.)

Evidence rating: Limited · Fact-checked: July 29, 2026

Magnesium is one of the most heavily marketed sleep supplements in the United States, and the case for it sounds convincing: it’s involved in hundreds of processes in the body, a lot of people don’t get enough of it from food, and it has a calming reputation.

We went looking for the trials behind that pitch. There are fewer than you’d expect, they’re smaller than you’d expect, and the most-marketed form of magnesium is not the one that was studied.

Here is what we found, including the parts that don’t support the marketing.

The short version

• The best-quality summary of the evidence pooled three trials in 151 older adults. People fell asleep about 17 minutes faster than on placebo.
• In those same trials, total sleep time did not improve to a statistically significant degree. Falling asleep faster and sleeping longer are not the same claim.
• The researchers who did that review rated the evidence as low to very low certainty and wrote that the literature is not good enough for doctors to make well-informed recommendations from.
• The trials used magnesium oxide and citrate. The form sold hardest for sleep — glycinate — was not among them.
• The one recent trial of glycinate also delivered 1.5 grams of glycine, which has its own sleep research. That trial cannot tell you which of the two ingredients did anything.

Our evidence rating for magnesium and sleep is Limited. That is not a verdict that it doesn’t work. It’s a statement that the research is too thin and too flawed to support confident claims in either direction.

What the strongest evidence actually shows

The most rigorous summary available is a 2021 systematic review and meta-analysis published in BMC Complementary Medicine and Therapies, which searched MEDLINE, EMBASE, clinicaltrials.gov and two grey-literature databases, and assessed study quality using the standard RoB 2.0 and GRADE frameworks.

After all that searching, it found three randomized controlled trials, together enrolling 151 older adults across three countries.

Pooling them:

Outcome	Result	Statistically significant?
Time to fall asleep	17.36 minutes faster than placebo (95% CI: 7.44 to 27.27 minutes)	Yes (p = 0.0006)
Total sleep time	16.06 minutes longer (95% CI: −5.99 to 38.12)	No (p = 0.15)

That second row is the one that tends to disappear from marketing copy. The confidence interval crosses zero, which means the data are compatible with magnesium adding half an hour of sleep — and equally compatible with it costing you six minutes.

Why the reviewers weren’t impressed

The authors rated all three trials at moderate-to-high risk of bias, and the evidence overall at low to very low certainty. Some specifics worth knowing:

• None of the three trials reported how they randomized participants, how they concealed allocation, or how they blinded people — despite all three describing themselves as double-blind.
• Two were judged high risk of bias outright. One of those reported an effect on a single sub-score of a seven-part sleep questionnaire, using a cutoff that didn’t appear in its own methods section.
• Trial length ran from 20 days to 8 weeks. Nobody has tested what happens over a year.

Their conclusion, in their words, is that the quality of the literature is substandard for physicians to make well-informed recommendations about magnesium for insomnia in older adults.

They did note the other side: magnesium is cheap and widely available, which changes how much certainty you might reasonably demand before trying it. That’s a fair point, and it’s a different kind of argument than an evidence claim.

These trials were not about most people buying magnesium for sleep

The participants were older adults, mostly 55 and over, with insomnia diagnosed by a validated questionnaire, a clinician, or a sleep diary.

That matters twice. Older adults have lower dietary magnesium intake, absorb less of it, and excrete more of it — so they are more likely than a younger person to actually be short of magnesium. And people with diagnosed insomnia have more room to improve than people who merely sleep badly some nights.

If you are 32 and sleep poorly because of your phone, these trials were not run on you.

The glycinate question

Walk into any supplement aisle and the form marketed for sleep is magnesium glycinate — or bisglycinate, the same thing. The pitch is that it’s gentler on the stomach and more calming.

The trials in the review above used magnesium oxide and magnesium citrate. Not glycinate. So the “17 minutes” figure does not belong to the product most people are being sold.

There is one relevant recent trial. Published in Nature and Science of Sleep in 2025, it randomized 155 adults aged 18–65 with self-reported poor sleep to either magnesium bisglycinate or placebo for four weeks.

The result was positive and small. Insomnia Severity Index scores fell by 3.9 points on magnesium versus 2.3 on placebo — a difference that reached statistical significance at p = 0.049, which is about as close to the conventional threshold as a result can land while still counting. The authors reported the effect size as Cohen’s d = 0.2, which is small by the usual convention.

The confound is in the capsule. Each daily dose delivered 250 mg of magnesium and 1,523 mg of glycine. Glycine is not filler — it has its own body of sleep research, and doses in that range are what those studies used. This trial had no arm testing glycine alone, so nothing in it can tell you whether the benefit came from the magnesium, the glycine, or the combination.

We’re not saying that to dismiss the trial. It was competently run, and its funding was institutional rather than industry — the Institute of Food and One Health at Leibniz University Hannover, with the supplement supplied by manufacturer Biogena. One author disclosed directing a contract research organization that takes funding from nutraceutical companies. We mention that because our Editorial Policy says we will, not because it invalidates anything.

But if a product’s marketing points at this trial to sell you magnesium glycinate, the honest reading is that the trial tested a magnesium-and-glycine combination, and that’s what it found something for.

What to know before you buy

Doses in the trials were higher than the official ceiling. The trials of older adults used 320 to 729 mg of elemental magnesium daily. The National Institutes of Health sets the Tolerable Upper Intake Level for supplemental magnesium at 350 mg per day for adults. That upper limit is about supplements specifically, not food. It’s worth understanding that several of these trials operated above it.

The label number you want is elemental magnesium. A capsule containing 893 mg of magnesium bisglycinate provides 125 mg of magnesium — the rest is the glycine it’s bound to. US labels are required to declare elemental magnesium, but the front of the bottle isn’t always the panel doing that.

Not all forms absorb equally. According to the NIH, magnesium in aspartate, citrate, lactate and chloride forms is absorbed more completely than magnesium oxide and magnesium sulfate. Oxide is common in cheap products because it packs a lot of elemental magnesium per pill — it just doesn’t all arrive.

Loose stools are the expected side effect, not a rare one. In one of the three trials in the review, every participant reported them. Magnesium is an active ingredient in laxatives for exactly this reason.

The FDA’s own position is unenthusiastic. The agency permits a qualified health claim about magnesium and blood pressure while requiring the accompanying statement that the evidence is “inconsistent and not conclusive,” and it caps supplement doses carrying that claim at 350 mg.

Who should talk to someone first

Magnesium interacts with real medications and is cleared by the kidneys, so this is not a universal “worth a try.”

Talk to a doctor or pharmacist before taking supplemental magnesium if you have kidney disease or reduced kidney function — impaired clearance is the main route to magnesium building up to dangerous levels. Also check first if you take antibiotics, bisphosphonates, diuretics, or proton pump inhibitors, all of which have documented interactions with magnesium in either direction.

If your sleep problem has lasted months, is wrecking your days, or comes with loud snoring and daytime sleepiness, the useful step is a clinician rather than a supplement. Insomnia and sleep apnea have treatments with far better evidence behind them than anything on this page.

Nothing here is medical advice, and no one on our team is a clinician. That’s stated plainly on our About page.

What we could not check

• We did not test any product. We have not had magnesium supplements analyzed for purity or label accuracy, and we can’t tell you what’s actually in any bottle.
• We could not resolve the form question. No trial has directly compared glycinate against citrate or oxide for sleep, so claims that one form works better for sleep are currently untested rather than proven.
• We could not separate magnesium from glycine in the only glycinate trial, because the trial wasn’t designed to allow it.
• We have no long-term data. The longest trial ran eight weeks.
• We did not evaluate whether you personally are low in magnesium. Blood tests for magnesium status are unreliable, and that’s a question for a clinician, not an article.

Our rating, and why

Limited. Under our evidence scale, this rating means sparse human data — the evidence rests on a small number of short trials rather than a well-replicated body of research.

That’s the case here. Three small trials in one narrow population, all at moderate-to-high risk of bias, showing an effect on one sleep measure and not another — plus one newer trial in a different population, testing a different form, with an unresolvable confound and a p-value sitting on the line. None of it adds up to the trial volume or consistency the “Moderate” or “Strong” tiers describe.

If the research improves, this rating will change, and we’ll say so on this page when it does.

Sources

1. Mah J, Pitre T. Oral magnesium supplementation for insomnia in older adults: a Systematic Review & Meta-Analysis. BMC Complementary Medicine and Therapies. 2021;21:125. Read the study
2. Schuster J, Cycelskij I, Lopresti A, Hahn A. Magnesium Bisglycinate Supplementation in Healthy Adults Reporting Poor Sleep: A Randomized, Placebo-Controlled Trial. Nature and Science of Sleep. 2025;17:2027–2040. doi:10.2147/NSS.S524348. Read the study
3. National Institutes of Health, Office of Dietary Supplements. Magnesium — Fact Sheet for Health Professionals. Read the fact sheet
4. U.S. Food and Drug Administration. FDA Announces Qualified Health Claim for Magnesium and Reduced Risk of High Blood Pressure. 2022. Read the announcement