Creatine: reality vs hype
Creatine is the most-studied supplement on the shelf and the internet has decided it is also a nootropic. The evidence says something narrower and more interesting: it works when your brain is under stress, and mostly does nothing when it isn't.
Creatine has had a strange year. It went from a gym powder to a brain supplement, and the jump happened almost entirely on social media rather than in journals. Somewhere along the way it picked up claims about Alzheimer's, memory, focus, and depression, most of which trace back to studies far smaller and far weaker than the claims they're carrying.
So I went through the actual literature. Here is what holds up.
The short version
Creatine produces a small but real cognitive benefit, mostly on memory. The pooled effect size is about 0.3 standard deviations, which is real but modest. It is not a step change in how your brain works.
That benefit is concentrated almost entirely in people whose brains are under some kind of stress: older adults, the sleep-deprived, people at altitude or under low oxygen, possibly people with depression. If you are young, healthy, well-rested and eating meat, the best-powered trial we have found essentially nothing.
The big disease claims failed. Parkinson's, Huntington's and ALS all had large, properly run trials. All negative. Some of them were stopped early for futility.
The Alzheimer's result everyone is quoting does not say what people think it says.
Doses that work for muscle may be too low for the brain. Brain creatine rises slower and less than muscle creatine, and the studies that actually moved the needle used 20 g/day, not 5.
Safety is not the issue. Up to 30 g/day for five years, no detrimental effects in healthy people. The main real-world cost of high doses is your gut.
What creatine actually does up there
The brain runs on ATP and it burns roughly 20% of your body's energy. Phosphocreatine is the buffer that regenerates ATP fast when demand spikes. That is the whole mechanism, and it is genuinely plausible.
The problem is delivery. Creatine crosses the blood-brain barrier through a specific transporter, SLC6A8, which is not abundant and appears to be absent from astrocytic end-feet. The brain also makes its own creatine locally, so it is far less dependent on what you swallow than your muscles are.
The numbers show it. Muscle creatine goes up around 20% with supplementation. Brain creatine goes up 5 to 11%, and sometimes not measurably at all. Dechent's 1999 study got 8.7% after four weeks at 20 g/day. Turner got 9% after a week at 20 g/day. That is the ceiling we're working with.
Most trials never measured brain creatine at all. Which means a null result is ambiguous: it might mean creatine doesn't help, or it might mean the creatine never got there.
That ambiguity runs through the entire literature and a 2026 methodological review says so explicitly.
What the meta-analyses actually say
This is where it gets messy, and where most popular coverage stops reading.
Prokopidis et al. (2023) is the paper everyone cites for "creatine improves memory." It reported a pooled effect of +0.29, with a large effect in adults aged 66 to 76 and essentially nothing in the young. Then a letter to the editor pointed out the analysis double-counted participants through inadequate statistics. The authors re-ran it properly. The overall memory effect was no longer significant — except in older adults.
Xu et al. (2024) reported benefits for memory, attention and processing speed. EFSA reviewed it in 2024, found it had pooled non-independent cognitive tests from the same studies and inflated its sample sizes, and concluded no conclusions could be drawn. EFSA rejected the cognitive health claim outright.
Meanwhile, the largest cognition RCT we have — Sandkühler et al. in BMC Medicine, 123 participants, cross-over, double-blind, six weeks — found nothing much. Borderline on backward digit span, nothing on Raven's matrices, nothing across eight exploratory tests. Vegetarians did not benefit more than omnivores. Side effects were four times more common on creatine.
Moriarty (2023) tried the obvious fix, raising the dose to 10 or 20 g/day for six weeks in healthy young adults. No effect on processing speed, episodic memory or executive function, and no change in prefrontal activation on fNIRS. So "just take more" is not the answer either, at least not for that population.
The one place it clearly works
Sleep deprivation. This is the most consistent finding in the whole field.
McMorris (2006) gave 20 g/day for a week and then kept people awake for 24 hours. Creatine improved mood and blunted the decline on prefrontal-heavy tasks. A follow-up found the same at 36 hours.
Then Gordji-Nejad (2024, Scientific Reports) did something new: a single high dose, 0.35 g/kg, in sleep-deprived subjects, with 31P-MRS to actually watch the brain chemistry. It raised phosphocreatine and ATP, prevented the drop in brain pH, and improved cognition and processing speed for a window of roughly three to seven and a half hours. That is the first demonstration that one dose can transiently change brain energetics.
A 2026 follow-up at a lower dose (0.2 g/kg, 29 participants) found improvements up to 12%, confirming dose dependence, with larger effects in women and vegetarians.
Both studies are small. Neither has been independently replicated. But this is the most mechanistically coherent result in the field: the buffer helps most when the system is running dry.
Turner (2015) found the same pattern under hypoxia. Creatine raised brain creatine 9%, increased corticomotor excitability, and preserved attention that otherwise collapsed at 10% oxygen.
Vegetarians
The hypothesis is elegant: vegetarians eat almost no creatine, so they should have more room to gain. Rae (2003) found large improvements in working memory and reasoning in 45 vegetarians. Benton and Donohoe (2011) found memory improvements in vegetarians but not omnivores.
Then Sandkühler (2023) ran half its cohort as vegetarians and found no difference. It also cites work showing no baseline brain creatine difference between vegetarians and omnivores, even though vegetarians clearly have lower muscle and serum creatine.
So the brain seems to defend its creatine levels regardless of diet. Plausible hypothesis, inconsistent evidence.
Older adults
This is the subgroup that survives every correction. Prokopidis found a large effect in the 66 to 76 range, and after the statistical fix it was the only subgroup that stayed significant. McMorris (2007) found improved short and long-term memory in 76-year-olds after a single week.
Marshall et al. (Nutrition Reviews, 2026) reviewed it and landed on the honest position: limited evidence suggests a benefit, high-quality trials are still needed.
If there is one group with a real case for taking creatine for cognition, it is this one.
Where it failed
Parkinson's: NET-PD LS-1, 1741 patients, 10 g/day for five or more years, published in JAMA in 2015. No difference in progression. Stopped for futility.
Huntington's: CREST-E, 553 patients, up to 40 g/day for up to four years. Halted at the first interim analysis. Functional decline numerically favoured placebo. More GI adverse events and more withdrawals in the creatine arm.
ALS: three separate RCTs, no benefit on survival, strength or function.
These are not underpowered pilots. These are the trials the field asked for, and they came back negative. Anyone selling creatine as neuroprotection is ignoring them.
The Alzheimer's story is not what you read
You may have seen that creatine "slowed Alzheimer's decline by about 30% versus placebo." That number does not exist in the literature.
The study is CABA, published in 2025. It had 20 patients, ran for eight weeks, and had no placebo arm at all. It was a single-arm feasibility pilot. It successfully raised brain creatine by 11% at 20 g/day, which is a genuinely useful finding.
The cognitive numbers are within-group, unadjusted paired t-tests. Some moved: working memory, fluid cognition, oral reading recognition. Several did not move at all: episodic memory, processing speed, executive function. MMSE actually went down slightly.
The authors themselves write that the design prevents any conclusion about efficacy, and that the improvements may be artifact — test-retest effects, placebo, and so on. A proper placebo-controlled trial is planned. It has not reported.
A single-arm study cannot produce a "versus placebo" number. There was no placebo.
That claim was invented downstream of the paper and then repeated until it sounded like a finding.
Depression is the interesting outlier
Lyoo et al. (American Journal of Psychiatry, 2012) added 5 g/day of creatine to escitalopram in 52 women with major depression. Separation from placebo by week two. Remission at eight weeks was 52% versus 26%. Effect size around 1.1, which is large for psychiatry.
Kondo showed dose-dependent increases in frontal-lobe phosphocreatine at 2, 4 and 10 g/day, tracking mood improvement in SSRI-resistant adolescent girls. So there's a mechanism and a dose-response, which is more than most of this field can claim.
The catch: it's almost entirely female cohorts, and there's a small negative trial. It needs larger mixed-sex replication. But of all the clinical claims, this is the one I'd bet on.
Dosing
For muscle, 3 to 5 g/day works fine, with or without a loading phase.
For the brain, that's probably not enough. Every study that measurably raised brain creatine used around 20 g/day, or a single acute dose in the 0.2 to 0.35 g/kg range. Reviews in this space have converged on "more is likely better" for brain bioenergetics.
But note the tension: Moriarty gave healthy young adults 10 to 20 g/day for six weeks and got nothing cognitively. So higher doses raise brain creatine more reliably; they don't reliably raise cognition. Optimal brain dosing remains an open question, and anyone giving you a confident number is ahead of the data.
Safety
This part is boring, which is a compliment.
The ISSN position stand is unambiguous: no compelling evidence that creatine monohydrate up to 30 g/day for five years has detrimental effects in healthy people or clinical populations. No renal harm in healthy individuals, including five-year follow-up in athletes.
Two practical notes. Creatine raises serum creatinine, which looks like reduced kidney function on a blood panel — tell your doctor before you get tested. And GI symptoms are the real cost of high doses; CREST-E at 40 g/day saw meaningfully more dropouts for that reason. Split the dose.
Also: only monohydrate has this evidence base. HCl, ethyl ester and buffered forms are marketing, not data.
What I'd actually do
If you're young, healthy, sleeping well and eating meat, and you're taking creatine specifically for your brain — you're probably wasting the money. Take it for muscle and bone, where the evidence is genuinely strong, and treat cognition as a free option that may not pay out.
If you're over 65, vegetarian, chronically under-slept, or otherwise running your brain in deficit, the case is much better. 3 to 5 g/day is cheap, safe and well-tolerated. Higher if you're specifically chasing brain effects, accepting that the incremental benefit is unproven.
If you have a hard cognitive demand after a bad night, a single high dose has two promising studies behind it and zero validation for routine use.
And if you're taking it hoping to hold off Parkinson's, Huntington's or ALS: that question has been asked properly and the answer was no.
The pattern here
This isn't a story about creatine being useless. It's a story about how a supplement with modest, conditional, subgroup-specific effects gets flattened into a universal claim, and then the flattened version travels faster than the correction.
Two of the three most-cited pro-creatine meta-analyses have formal statistical objections against them. One author group re-ran their own numbers and lost their headline. A regulator looked at the whole pile and declined to approve the claim. A twenty-person uncontrolled pilot became a "30% versus placebo" headline.
The underlying science is fine. It's the distance between the papers and the posts that's the problem.
