Magnesium Threonate vs. Glycinate: Which One the Evidence Actually Supports

There is a specific, unglamorous failure state that sends people to the magnesium aisle. You slept seven hours. The tracker says the sleep was fine. And yet at nine in the morning you are looking at the first real task of the day through frosted glass — the thinking is available, technically, but it costs more than it should. Underneath it sits a low hum of activation that never quite switched off overnight.

That is not a sleep-duration problem. It is a nervous-system-tone problem: an excitatory system that stayed too loud through the night, so the morning starts from a deficit. And magnesium is the most mechanistically plausible lever anyone has ever pointed at it, because magnesium is literally the ion sitting in the doorway of the brain’s main excitatory receptor, deciding when it opens.

Which is what makes the marketplace’s answer so frustrating. You are told to choose between magnesium L-threonate — the brain one, engineered at MIT, expensive — and magnesium glycinate, the calm one, cheap, on every shelf. The framing is clean, the mechanism story is seductive, and it is roughly half wrong.

Here is the tension this article is built around, planted early so you know a real question is being investigated rather than a foregone conclusion sold: the form with the better brain-delivery mechanism has the weaker independent evidence, and the form with the better independent evidence has almost no brain data at all. Neither of those is a typo. Getting to the bottom of that gap will serve you better than any declared winner.

Magnesium is cleared renally. Anyone with impaired kidney function, or taking diuretics, proton pump inhibitors, or other magnesium-affecting medication, should not supplement without a clinician involved.

Why magnesium should be a nootropic

The gatekeeper at the NMDA receptor

Start with the receptor everything hinges on. The NMDA receptor is the brain’s principal plasticity engine — the gate through which learning physically happens. It is also unusually hard to open, deliberately so. Three conditions must coincide: a co-agonist (glycine or D-serine) must occupy the GluN1 site, glutamate must bind the GluN2 site, and the membrane must depolarize enough to expel a magnesium ion that is otherwise plugging the channel pore.

That last condition is the voltage-dependent Mg²⁺ block. When it lifts, calcium floods in and triggers CaMKII signaling — the cascade that converts electrical activity into structural memory.

So magnesium is not a sedative in any conventional sense. It is a signal-to-noise device. Adequate intracellular magnesium means the receptor stays shut against ambient chatter and opens crisply for genuine learning events. Insufficient magnesium means an over-excitable receptor: more background firing, less contrast, and — the part you actually feel — a nervous system that cannot decide the day is over.

Magnesium’s second job runs the other direction. It acts on GABA receptors, improving the efficiency of the brain’s primary inhibitory system. So the effect is two-handed: more inhibition, less excitation. More brake and less gas, if you like the analogy, though the more accurate version is that magnesium restores the gap between brake and gas, which is what a calm brain actually is.

If magnesium status were the whole story, form would not matter. It matters because of a delivery problem.

The threonate gambit

The blood-brain barrier is very good at excluding ions. Raise serum magnesium and you mostly raise serum magnesium; brain concentrations barely move. This constraint motivated the entire L-threonate program.

The hypothesis is that the L-threonate ligand itself facilitates central delivery, with the leading proposed route being the glucose transport system — GLUT1 and GLUT3 transporters on barrier endothelial cells — permitting active transport rather than passive diffusion. In rodents, oral magnesium L-threonate raised cerebrospinal fluid and brain magnesium more than other magnesium salts at comparable elemental doses, and that elevation tracked with increased synaptic density and better spatial memory. Follow-up mechanistic work found the threonate moiety modulates intraneuronal magnesium and thereby regulates synapse density in its own right — the ligand is not merely a taxi.

There is also a barrier-integrity finding that complicates the tidy narrative in an interesting direction: in a neuromyelitis optica mouse model, threonate pretreatment reduced demyelination, protected tight junction proteins, and blunted astrocyte and microglial activation, with the in-vitro protection running through the TRPM7 channel. So under inflammatory stress, magnesium appears not only to cross the barrier but to help maintain it.

What about the mitochondrial and growth-factor claims that populate supplement copy? The honest answer is short. On BDNF and NGF, no trial isolating either magnesium form’s effect on these neurotrophins was located: data not specified. On astrocyte mitochondria specifically: data not specified. Those bullet points appear on sales pages anyway. They are extrapolation, not evidence.

The glycinate mechanism, and its buried irony

Glycinate’s story is usually told as magnesium plus a calming amino acid, and that is not wrong. After ingestion the chelate dissociates in the small intestine; Mg²⁺ is absorbed via TRPM6 and TRPM7 transporters while glycine enters cells independently. Glycine is inhibitory at its own receptors, it lowers core body temperature — a genuine sleep-onset trigger — and it feeds glutathione synthesis. The chelated structure also resists precipitation in the gut, which is why it is among the least laxative forms.

But look again at the NMDA gate. Glycine is the obligatory co-agonist at the GluN1 site — the molecule that must be present for the receptor to open at all. So magnesium glycinate is, at the level of molecular bookkeeping, delivering one ligand that plugs the NMDA channel and another that is a required condition for opening it.

Whether that self-cancellation matters at physiological doses is genuinely unresolved: data not specified. It is worth raising not to score a point, but because it is exactly the kind of mechanistic wrinkle that vanishes from marketing copy — and because it should make you suspicious of anyone telling you the mechanism story here is simple.

The myth worth busting

The myth: threonate is the magnesium that reaches your brain, everything else stops at the neck.

The correction: the CNS-targeting claim exists only for L-threonate, and that part is true. But the evidence for it is animal and in vitro. No human neuroimaging trial has confirmed that oral magnesium L-threonate raises brain magnesium in people. And on plain oral absorption, threonate is merely comparable to other well-absorbed organic salts; glycinate and citrate absorb well too. Threonate’s differentiator is a hypothesis about where the magnesium goes after absorption, not a demonstrated advantage in getting absorbed at all.

There is a second, purely arithmetic problem the marketing never leads with. L-threonate is only about 7–8% elemental magnesium by weight. A 2,000 mg dose delivers roughly 144–145 mg of elemental magnesium — below the RDA, and well under what glycinate delivers at typical doses.

So the premium form ships less of the actual mineral and bets that the small amount it does ship lands somewhere better. That is a real bet. It also has a scoreboard.

What people report

Everything in this section is anecdotal — community reports from sleep and nootropics forums. It is signal about what to expect and what to watch for, not proof of anything.

Nothing much happens in the first thirty minutes, from either form. Nobody credible describes an acute hit, and that alone should reframe your expectations: magnesium is a repletion intervention, not a stimulant.

By the first week the reported signal is not the one people expect. Users converge on magnesium improving sleep maintenance rather than sleep onset — staying asleep rather than falling asleep faster. A frequently described secondary effect is reduced sleep anxiety: the three a.m. loop of noticing you are awake and being upset about it. A subset report more vivid dreams in the first week or two, usually framed positively, usually settling.

By the first month the reports diverge by form. Glycinate users converge on stable nightly calm. Threonate’s following is smaller and explicitly cognitive — a distant second for sleep in these threads, and a first choice only among people optimizing for memory and focus.

The doses actually in use: for glycinate, 200–400 mg elemental at night, with 240–300 mg recurring in the positive reports. For threonate, the studied compound dose of 1.5–2 g.

The characteristic glycinate complaint at the top of the range is a floppy, over-relaxed feeling the next morning — which, if morning clarity is the goal, is an own goal. The fix is unglamorous and correct: lower the dose. The community also names three usual reasons magnesium “doesn’t work”: using poorly absorbed oxide, quitting after three nights instead of waiting two-plus weeks, or having a sleep disorder that no mineral can fix.

Hold onto that last cluster. The gap between what this section promises and what the trials deliver is the real subject of this article.

The gap: a beautiful mechanism with a mediocre scoreboard

The mechanism section above is strong. It has a receptor, a transporter, a plausible route across the barrier, rodent synaptic-density data, and a coherent story about why an over-excited nervous system produces a foggy morning. On paper this should be among the most reliable nootropics in existence.

Now the scoreboard.

The most rigorous recent human trial of threonate — 100 adults, 2 g/day for six weeks — is a study worth praising for its honesty and then reading very carefully. It hit its primary endpoint: NIH Total Cognition Composite improved more than placebo (p = 0.043), with the cleanest individual signals on working memory (p = 0.033) and reaction time (p = 0.031, a 6.3% improvement). It reported an eye-catching 7.5-year reduction in estimated cognitive age. And, directly relevant to the complaint that brought you here, all cognitive testing was administered in the morning between 8 and 11 a.m., caffeine-free. This is the closest thing that exists to a trial of your actual problem.

Now the same trial’s null results, which the authors preserved and which most coverage of this compound quietly omits:

• Fluid intelligence: p = 0.953. Not a trend. Nothing.
• NIH Fluid Composite: p = 0.277. Null.
• Crystallised Composite: p = 0.111. Null.
• Self-reported sleep disturbance: p = 0.316. Null.
• Restorative sleep: p = 0.439. Null.
• General wellbeing: p = 0.436. Null.
• Not a single Oura Ring sleep parameter showed a significant between-group difference.

So the objective sleep story collapses entirely, and the cognitive story survives on a primary composite that cleared significance by a hair, driven by two subtests out of many. And the study was funded by Threotech Inc., the Magtein patent licensor, which also provided the intellectual property and participated in study design.

That is not a smear; it is a structural fact you have to price in. Every published clinical trial of this compound for sleep has been funded by entities holding the patent, and independent replication does not exist. Examine.com’s assessment is blunt on the point: the human research is highly limited, industry-funded, and insufficient to confirm efficacy.

Meanwhile, look at what happened when threonate was tested by people without that stake. In a 2023 trial in women after breast cancer surgery, 1.2 g/day (91.8 mg elemental) for 12 weeks produced no cognitive improvement versus placebo. That trial is legitimately compromised — high risk of bias, roughly 29% missing outcome data in the magnesium arm, and a low dose. But it is the one non-vendor cognitive test available, and it came back empty.

The comparison, and an honest complication

The strongest trial in this space is not a threonate trial at all. It is a randomized, double-blind, placebo-controlled study of magnesium bisglycinate — 155 adults aged 18 to 65 with self-reported poor sleep, 250 mg elemental daily, published in Nature and Science of Sleep in 2025 (Schuster et al., PMID 40918053). It found a significantly greater reduction in Insomnia Severity Index than placebo at week four: −3.9 versus −2.3, p = 0.049, with a candidly reported effect size of Cohen’s d = 0.2. Small. Real, but small.

And exploratory analysis found the improvements were notably greater in participants reporting lower baseline dietary magnesium intake — which is the single most important sentence in this article, and we will come back to it.

Here is where honesty requires a correction to the story I would otherwise like to tell. It is tempting to frame this as the independent trial against the vendor trials, and that framing is too clean. The study was led from the Institute of Food and One Health at Leibniz University Hannover, and three of its four authors declare no conflicts of interest. But the third author is Adrian Lopresti — the same researcher who led the threonate trial above — and he discloses that he is managing director of a contract research organization that receives funding from nutraceutical companies, and has taken speaking honoraria from them.

So the contrast is real but it is a difference of degree, not a clean division into paid and unpaid science. An academically led trial with one industry-connected co-author is a stronger evidentiary position than a trial funded and co-designed by the patent holder. It is not the same thing as disinterested replication, and anyone who tells you otherwise — including an earlier draft of this article — is simplifying.

Four explanations for the gap

One: the dose is arithmetically self-limiting. Threonate’s brain-targeting advantage is purchased at the cost of shipping only about 144 mg of elemental magnesium. You cannot both under-deliver the mineral and expect a large systemic effect. The entire bet rides on the routing being that good.

Two: the barrier claim is unverified in humans. The GLUT-mediated transport story rests on rodents and cell models. No human imaging study has confirmed that oral threonate raises brain magnesium in people. The premise of the premium is, at this moment, an inference.

Three: statistical significance is not practical significance. A p of 0.043 on a composite, and a d of 0.2 on insomnia severity, are the signatures of small real effects — not of a compound that fixes your mornings. The 7.5-year cognitive rejuvenation headline is a rescaling of a 2.24-point raw-score difference. It is not a lie. It is a very good piece of framing.

Four, and most important: the responder subgroup is being averaged into invisibility. Both trials point the same direction. The glycinate study found bigger effects at low baseline dietary magnesium. The threonate study found bigger effects in participants with worse baseline sleep. Magnesium works on deficits. Recruit people who are already replete and already sleeping adequately, and you are measuring a repletion intervention in people who need no repletion — then reporting the resulting null as a verdict on the compound.

That is the gap. Not “magnesium doesn’t work,” but something more precise: magnesium is a correction, and the trials keep testing it on people with little to correct. Which tells you a great deal about whether you should take it.

The protocol, if you choose to run it

This describes how to run a self-experiment properly. It is not a claim that efficacy is settled.

Pharmacokinetics, honestly. Half-life of the threonate ligand: data not specified. Formal timing pharmacokinetics for threonate: not well characterized — in trials it has generally been taken with food. Tmax for either form: data not specified. Anyone offering you a confident absorption curve for these compounds is inventing it.

Absorption saboteurs, which are real and worth respecting. Phosphoric acid in soft drinks binds magnesium and precipitates it. Calcium competes for intestinal absorption, so separate them. High sodium, high protein, caffeine and alcohol all increase renal magnesium wasting. Diuretics, proton pump inhibitors, insulin and bisphosphonates are magnesium-depleting drugs. If you take any of these, you may be a responder for reasons that have nothing to do with the brand on the bottle.

Doses used in the actual studies:

No tolerance or receptor-downregulation signal appears in the literature for either form: data not specified. Magnesium is a mineral being restored to physiological range, not a receptor agonist being pushed past it. Continuous use is what the trials tested.

The verdict: an asymmetric bet, with one condition

The bet is worth taking — but only if you are plausibly the responder the trials keep accidentally excluding. The signal is unmistakable once you stop reading trials for their headline and start reading them for their subgroups. The older adults, the poor sleepers, the low-dietary-magnesium participants got real effects. The well-fed young adults sleeping seven hours got p = 0.953.

So before buying anything, ask the only question that predicts your response: do you have a deficit to correct? Heavy caffeine or alcohol intake, a diuretic or PPI prescription, low intake of greens, nuts and legumes, chronic stress, poor sleep. Those are the qualifiers. If none of them describe you and you sleep fine, the expected effect on your mornings is somewhere between small and nothing — and this is the paragraph that saves you the monthly cost of a supplement you did not need.

If you do qualify, here is how the evidence says to run it.

Start with glycinate, 200–300 mg elemental, sixty minutes before bed, for four uninterrupted weeks. It is cheaper, it delivers more of the actual mineral, and it has the better-quality trial behind it. Repletion, not routing, is the mechanism to bet on first, because it is the mechanism the evidence best supports.

Track three markers rather than impressions:

1. Resting heart rate and HRV during sleep. This is the quiet finding of the threonate trial: the objective physiological effects survived even when the subjective sleep scores did not. Autonomic tone is the cleanest magnesium readout you can measure at home.
2. Sleep-related impairment, not sleep disturbance. Note which endpoint kept clearing significance and which kept failing. The trials suggest the daytime feel improves before the night-time architecture does. Rate your nine a.m. clarity daily, one to ten.
3. A reaction-time task. Sixty seconds of any aim trainer, same time each morning, caffeine-free. It is the one objective cognitive metric that moved.

Escalate to threonate only on a specific trigger: four weeks of glycinate delivered somatic calm and better nights, but the morning fog is unmoved. That is precisely the phenotype threonate is theorized to address, and at that point you are running a real experiment with a real prior rather than paying a three-to-fivefold premium for a delivery mechanism never confirmed in a human being. Then use the studied dose — 2 g, about 145 mg elemental — for six weeks and re-measure the same three markers.

Stop or reduce if you get loose stools or abdominal pain, headaches, or that floppy over-relaxed morning that means you overshot. And if you have any degree of renal impairment, do not start without a clinician. Magnesium clearance is renal, and that is the one place where the downside stops being close to zero.

The honest summary: you are not choosing between a brain magnesium and a body magnesium. You are choosing between a cheap, reasonably evidenced repletion and an expensive, elegantly argued routing hypothesis. Buy the repletion first. Make the routing earn it.

About this article

Written by Leah Elish. Covers diet, micronutrients, the gut-brain axis, and longevity eating patterns. Not a dietitian or clinician — traces nutritional claims back to the studies that supposedly support them.

Medical review: None. NeuriFuel does not currently have a licensed clinician on the editorial team, and this article has not been medically reviewed. We state this rather than implying an authority we do not have. See our About page for our full methodology.

A note on this article’s own corrections. An earlier draft described the bisglycinate trial as independent and university-funded. Checking the record showed that its third author discloses industry ties, and that we could not verify the funding source. The claim has been narrowed accordingly. We mention it because the argument in this piece is partly about who funds evidence, and it would be poor practice to apply that scrutiny only to other people’s studies.

Known limitations: exact p-values for two of the cited trials were not retrievable from their abstracts and are marked as data not specified rather than estimated.

Corrections: Found an error? Write to hello@neurifuel.com with a source and we will fix it and log the correction.

Last updated: 22 July 2026

References

1. Slutsky I, et al. (2010). Neuron. DOI 10.1016/j.neuron.2009.12.026; PMID 20152124
2. Sun Q, Weinger JG, Mao F, Liu G. (2016). Neuropharmacology. PMID 27178134
3. Liu G, et al. (2016). MMFS-01. J Alzheimers Dis. PMID 26519439
4. Zhang C, et al. (2022). Nutrients. PMID 36558392
5. Surman C, et al. (2021). J Diet Suppl. PMID 32162987
6. Hausenblas HA, et al. (2024). Sleep Med X. DOI 10.1016/j.sleepx.2024.100121
7. Lopresti AL, Smith SJ. (2026). Front Nutr. PMID 41601871 — funded by Threotech Inc.
8. (2023) Magnesium L-threonate after breast cancer surgery. Breast Cancer (Dove Med Press). DOI 10.2147/BCTT.S413435
9. Boyle NB, Lawton C, Dye L. (2017). Nutrients. DOI 10.3390/nu9050429
10. Noah L, et al. (2021). Stress Health. DOI 10.1002/smi.3051
11. Schuster J, Cycelskij I, Lopresti A, Hahn A. (2025). Nature and Science of Sleep 17:2027–2040. PMID 40918053; DOI 10.2147/NSS.S524348
12. Magnesium glycinate + SSRI in MDD. ClinicalTrials.gov NCT04880460
13. (2024) Magnesium L-threonate, demyelination and blood–brain barrier integrity in NMOSD mice. Brain Research.
14. Examine.com — magnesium L-threonate research summary
15. Alzheimer’s Drug Discovery Foundation, Cognitive Vitality — magnesium researcher review