The most quoted number in the creatine and memory conversation was walked back by its own authors in 2023, and I have yet to see a product page mention it. That tells you a lot about how this topic gets sold versus how it reads in the journals.
Here’s the short answer. Human trials have produced a real but narrow set of results on memory tests: one striking 2003 study in vegetarians that a larger 2023 replication only partly reproduced, a signal in older adults that rests on two small trials, and clear short term benefits during sleep deprivation that came from doses far larger than anyone takes daily. When the trials in healthy adults are pooled and analyzed correctly, there is no statistically significant improvement in memory overall. The European food safety regulator read the same papers in 2024 and declined to authorize a cognitive claim. None of that makes creatine and memory a dead subject. It makes it an early one, and the gap between early and clinically proven is where most of the marketing lives.
What Memory Meant Inside These Trials
It helps to know what these studies were measuring, because memory on a supplement label and memory in a trial are different things. Nearly every trial in this field tested working memory, which is the ability to hold a small amount of information in mind and do something with it for a few seconds. The classic task is the backward digit span: a researcher reads a string of numbers, you repeat it in reverse order, and your score is the longest string you get right.
Other trials used forward digit span, spatial recall tasks where you reproduce a sequence of locations, word list recall, and occasionally a long term memory test where you learn material and reproduce it later in the same session. The 2003 study that started the whole conversation also used Raven’s Advanced Progressive Matrices, a pattern completion test of abstract reasoning rather than memory, even though it often gets described as a memory result.
What no trial has measured is the thing most people searching for creatine and memory care about: whether you’ll misplace your keys less often, remember names at a party, or retain what you studied last week. The outcomes are laboratory scores on tasks lasting minutes, usually on the same day.
The 2003 Trial Everything Else Gets Compared To
The study by Caroline Rae and colleagues at the University of Sydney, published in 2003, gave 45 young adult vegetarians 5 grams of creatine monohydrate a day for six weeks in a double blind, placebo controlled crossover design, meaning each person took both creatine and placebo in separate phases without knowing which was which. Creatine improved backward digit span and Raven’s matrices scores, with p values below 0.0001 on both.
Two things made this paper influential. The dose was the ordinary 5 gram daily serving, so it looked directly relevant to everyday use, and the effects were large, which is the kind of result that gets cited for twenty years.
The choice of vegetarians was deliberate. Creatine comes from meat and fish as well as from the body’s own production, so vegetarians eat far less of it, and the researchers reasoned that people with lower stores might have more to gain. That hypothesis is sensible, and as we’ll see, not well supported by later work. An effect this large in a 45 person nutrition trial is also a reason for caution rather than celebration, because unusually big effects in small samples have a poor record of surviving replication.
The Replication That Should Be on Every Label and Isn’t
In 2023, a team led by Julia Fabienne Sandkühler at the University of Bonn published the largest trial of creatine and cognition to date in BMC Medicine. It was a preregistered, double blind, placebo controlled crossover study, 5 grams a day for six weeks, in 123 people, about half of them vegetarian. It used the same two primary tests as the 2003 study plus eight exploratory cognitive tests. Preregistered means the researchers committed to their analysis plan before seeing the data, which rules out quietly picking the flattering result.
The result was a shrug with a small hopeful note. The creatine effect on backward digit span bordered on significance at p = 0.064 and did not reach it, and the effect on Raven’s matrices was nowhere close at p = 0.327. A Bayesian analysis, which weighs how much the data shift the odds for or against an effect, supported a small benefit. None of the eight exploratory tests improved. Vegetarians did not benefit more than omnivores. And side effects were reported about four times as often on creatine as on placebo.
My reading: a small effect on one working memory task is possible, an effect on reasoning is not supported, and the dramatic 2003 result did not reproduce at its original size. That’s not a failure of creatine. It’s what replication normally looks like.
Two other trials in young adults found nothing. A 2008 study by Rawson and colleagues gave young adults 0.03 grams per kilogram per day for six weeks, about 2 grams a day, and found no change on a broad neuropsychological battery, though a dose that low may simply not move brain creatine. More telling is a 2023 dose response trial from Moriarty and colleagues, which randomized 30 healthy young adults to 10 grams a day, 20 grams a day, or placebo for six weeks and found no effect on processing speed, episodic memory, or attention at either dose. Two of that paper’s authors are among the most prominent researchers arguing for creatine’s brain benefits, which makes the null result harder to wave away.
The pattern in rested young adults is consistent. Daily creatine does little or nothing on memory tests, whatever the dose.
The Meta Analysis That Got Corrected
If you’ve read that a meta analysis found creatine improves memory, it almost certainly traces back to a 2023 review in Nutrition Reviews by Prokopidis and colleagues. It pooled eight randomized controlled trials and reported that creatine improved memory with a standardized mean difference of 0.29. A standardized mean difference is a way of putting different tests on one scale, where roughly 0.2 counts as small, 0.5 as medium and 0.8 as large. The headline effect in older adults aged 66 to 76 was 0.88, which is large.
Then two Brazilian researchers, Igor Eckert and Eric Pascher, wrote a letter to the journal pointing out a statistical problem called double counting. Several trials reported multiple memory tests from the same participants, and the meta analysis had entered each test as if it came from a separate group of people. That inflates the apparent number of observations beyond the real number of participants, making the results look more precise and significant than they are.
The authors agreed and published a corrected reanalysis. After averaging the memory outcomes within each study, the overall effect fell to 0.19 with a confidence interval running from minus 0.07 to 0.46 and a p value of 0.15. In plain terms, the pooled trials no longer showed that creatine improves memory. Neither doses of 5 grams a day or less nor higher doses produced a significant effect. The one finding that survived was the older adult subgroup, at 0.80 with a p value of 0.004. Young adults aged 11 to 31 came out at 0.02, effectively zero.
The uncorrected 0.29 is still the number quoted in most articles, which is a handy test of whether a writer read past the abstract.
A second pooled analysis, published in Frontiers in Nutrition in 2024 by Xu and colleagues, took 16 trials with 492 participants, including people with medical conditions, and reported a memory effect of 0.31. This is the 2024 meta analysis now cited as fresh confirmation. In April 2026, a commentary by Tom Citherlet at the University of Lausanne documented the same double counting error, noting that single trials had contributed four to seven memory subtests each as if they were independent, and that EFSA had already flagged the problem and decided no conclusions could be drawn from the pooled data. As of this writing the 2024 analysis has not been rerun.
So the two positive meta analyses that carry this field’s reputation share one flaw, and the one that was fixed came back null overall. That rarely gets mentioned.
Older Adults: A Real Signal Standing on Two Small Trials
The corrected older adult finding is the most interesting thing in the literature, and it deserves both attention and a clear view of how thin its foundation is. The 66 to 76 age range in that subgroup corresponds to exactly two studies.
The first is a 2007 trial by McMorris and colleagues in 32 people averaging about 76 years old. Fifteen of them took placebo for a week and then 5 grams of creatine four times a day, 20 grams daily, for a second week, while 17 took placebo for both weeks. Creatine improved forward number recall, forward and backward spatial recall, and a long term memory task, but not backward number recall or random number generation. It was double blind, it was small, and the one week loading dose is four times what most people take.
The second is a 2013 trial by Alves and colleagues in older women averaging 66, randomized in groups of 14 to placebo, creatine, placebo with strength training, or creatine with strength training for 24 weeks. Creatine was given as 20 grams a day for five days and then 5 grams a day. The cognitive battery covered memory, selective attention and inhibitory control, and the authors reported no effect of creatine on cognition at either 12 or 24 weeks. That is the longest trial in older adults, at the everyday dose, and it found nothing.
Two 2026 reviews confirm how little else exists. A systematic review in Nutrition Reviews on creatine and cognition in aging found only those two intervention studies, with the remaining evidence coming from cross sectional surveys of dietary creatine intake, which can’t establish cause. A scoping review in Ageing Research Reviews published in August 2026 screened more than 8,000 records looking for trials on creatine and cognition in older adults and identified no randomized controlled trials at all, only a single 20 person pilot with no randomized control group.
One new trial is worth watching. In August 2026, Richard Kreider’s group at Texas A&M published a 12 week randomized, double blind study in which 64 sedentary adults aged 45 to 65 took 10 grams of creatine a day or placebo, with or without an exercise and diet program. The abstract reports favorable changes in selected markers of cognitive function and memory. Read the methods and you’ll find that the individual cognitive tasks were prespecified as exploratory rather than confirmatory endpoints, meaning the study was not powered to prove a memory effect, and the authors say so. The participants also averaged 54, so this is a middle aged trial reported as an older adult one. It’s a useful signal for a bigger study, not a result.
My view is that older adults are the most plausible place for creatine to matter for memory and the least adequately tested. Both statements are true, and the second should temper how the first gets used.
Vegetarians, Women, and the Baseline Theory
The tidy story is that creatine helps the people who start low, which would explain why vegetarians and older adults show up in the positive results. The trials only partly cooperate.
Benton and Donohoe’s 2011 study gave 128 young women, vegetarian and omnivore, 20 grams of creatine or placebo a day for five days. Memory improved in the vegetarians but not in the meat eaters, creatine reduced the variability of reaction times in both groups, and there was no effect on verbal fluency or vigilance. That is the best single piece of evidence for the baseline theory, and once again it used a loading dose for less than a week.
Against it stands the 2023 Bonn trial, where vegetarians on 5 grams a day for six weeks gained no more than omnivores. And the Prokopidis review’s own discussion points to something the theory needs to explain: a 2013 imaging study found no difference in brain creatine between vegetarians and omnivores, even though their muscle creatine differs. The brain seems to defend its creatine supply in a way muscle doesn’t, which undercuts the premise that vegetarians walk around with depleted brains waiting to be topped up.
Women are the newest chapter. The 2024 Frontiers meta analysis reported more benefit in females and in people aged 18 to 60, but subgroup results from an analysis with a double counting problem are the weakest evidence in this article, and I wouldn’t build anything on them. The 2026 sleep deprivation study discussed next also reported that women benefited more on some tasks. That is a reasonable hypothesis, not a finding.
Sleep Deprivation Is a Different Experiment
The strongest and most consistent creatine results on cognition come from people who have been kept awake, and it matters why they don’t transfer to a scoop in your morning coffee.
The early work came from two small 2006 studies by McMorris, which loaded young adults with 20 grams a day for a week and then kept them awake for 24 or 36 hours. The creatine groups showed smaller declines on random number and movement generation tasks, choice reaction time, and mood. Then in 2024, a group at Forschungszentrum Jülich published a study in Scientific Reports that changed the conversation. Fifteen healthy adults averaging 23 years old took a single dose of 0.35 grams per kilogram of body weight or placebo at the start of 21 hours without sleep, while brain scans tracked phosphocreatine and ATP, the molecules that store and deliver cellular energy. Creatine shifted those energy markers in the opposite direction to sleep loss, prevented the drop in brain pH that sleep loss causes, and improved processing speed and short term memory tasks, most visibly around four hours after the dose.
For someone weighing 70 kilograms, about 154 pounds, that dose is roughly 24 grams of creatine swallowed at once. The same group’s 2026 follow up in Nutrients tested 0.2 grams per kilogram, still about 14 grams, in 29 people and found smaller benefits on logic, language processing speed and vigilance, but this time not on short term memory, which the authors say appears to depend on dose.
The accurate sentence is this: a single large dose of creatine, roughly three to five times a normal serving, reduced the cognitive damage of one night without sleep in small lab studies. It is not evidence that 5 grams a day improves memory in a rested person, and the two shouldn’t be blended, which they constantly are.
What Regulators Said After Reading the Same Papers
In 2024 the German creatine manufacturer Alzchem asked the European Food Safety Authority to approve the claim that daily creatine supplementation can contribute to improved cognitive function. EFSA’s expert panel reviewed 23 human intervention studies and concluded that a cause and effect relationship had not been established for any cognitive domain, memory included. According to the panel’s reasoning as reported in the trade press, the acute working memory effects it did see occurred only at 20 grams a day, not at lower doses and not with 5 grams a day over six weeks, and only one of ten studies on response inhibition found anything. In May 2026 the European Commission made it official with Commission Regulation 2026/1118, which refuses the claim outright, so no food or supplement sold in the EU can make it.
That ruling does not bind a single product on a US shelf, and it’s worth saying why. In the United States, supplements don’t need premarket approval, and a company can make a structure/function claim like “supports memory” with a standard disclaimer and a notification to the FDA rather than a review. The legal check is the Federal Trade Commission, whose 2022 Health Products Compliance Guidance requires competent and reliable scientific evidence, which it says generally means randomized controlled human trials, and which warns that after the fact subgroup analyses usually don’t count. So a US label can legally say “supports memory” while the same words are banned in Europe. The science didn’t change at the border. The burden of proof did.
A 2025 perspective in the Journal of Nutrition, written by Igor Eckert, one of the two researchers who caught the double counting, argues that public and commercial enthusiasm for creatine as a cognitive enhancer has run well ahead of an evidence base that remains limited. Having read the trials, I think that is correct, and I say it as someone who works in this industry.
The Doses in the Papers Versus the Scoop in Your Kitchen
Lining up the doses shows the mismatch. Five grams a day for six weeks, the everyday protocol, produced the 2003 result and the mostly null 2023 replication. Twenty grams a day for five to seven days, a loading dose, produced the vegetarian, older adult and early sleep deprivation findings, and it’s the only protocol EFSA accepted as showing acute working memory effects. Single doses of roughly 14 to 24 grams produced the recent sleep deprivation results. Ten or 20 grams a day for six weeks produced nothing in rested young adults, and about 2 grams a day produced nothing at all.
Part of the explanation is transport. Muscle soaks up supplemental creatine readily, but the brain makes some of its own and takes up creatine from the blood slowly, across a barrier that admits it in limited amounts. A 2022 review in Nutrients summarizes the imaging studies, which show that supplementation raises brain creatine by considerably less than muscle. An industry scientist quoted after the EFSA decision put the typical rise at around 10 percent, ranging from about 3 to 15 percent between individuals. That variability may be part of why small trials disagree, and it is a fact about creatine as an ingredient, not something any particular product has solved.
On safety, the 2017 International Society of Sports Nutrition position stand concludes that doses up to 30 grams a day for five years have been well tolerated in healthy people and that there is no compelling evidence of kidney harm in healthy adults. That review was commissioned by the Council for Responsible Nutrition, a supplement trade group, with a grant to its lead author, which doesn’t make it wrong but does make it worth reading alongside independent sources. There is no formal US upper limit for creatine. The 2023 Bonn trial recorded side effects significantly more often on creatine than on placebo. If you have kidney disease, take medications, or are considering the large doses used in the sleep studies, that is a conversation for your doctor, not for a blog.
What Would Change the Picture
I’d move my own assessment of creatine and memory a long way with a single well run trial: a few hundred adults over 65, preregistered, at least 12 weeks, memory as the primary outcome, and brain creatine measured by imaging so we can finally see whether the people whose brain creatine rises are the people whose scores improve. Nothing like it exists yet.
I’d also want the 2024 Frontiers meta analysis rerun with the method its critics recommended, the Jülich single dose findings tested at everyday doses, and the Texas A&M cognitive data reported as a confirmatory analysis. Until then, the fair description of the evidence is that creatine has shown real effects on some memory tests in some people under some conditions, mostly at high doses and mostly briefly, and that the one claim that got the full regulatory treatment didn’t pass. If you decide to take creatine as an ingredient anyway, you’ll be doing so on the strength of its muscle evidence and a reasonable hypothesis about your brain, and there is nothing wrong with that as long as you know which one you’re paying for.