Does creatine improve memory?

Creatine has a strong reputation in muscle research and a plausible role in brain energy. Neither fact settles whether adding more makes a healthy person’s memory better.

The clean answer is narrower than the current search results suggest. A small everyday benefit remains possible. But the most prominent pooled result lost its overall statistical significance after a counting error was corrected, the largest individual healthy-adult trial found at most a small signal, and a European scientific panel did not accept a general cause-and-effect claim.

Two newer experiments add a different clue: under severe sleep loss, a single study exposure helped preserve selected cognitive performance. That is interesting evidence about acute stress. It is not proof of a routine memory upgrade.

The positive meta-analysis counted tests as if they were people

In 2023, a systematic review identified ten randomized placebo-controlled trials of memory in healthy people and pooled eight of them. Its headline result was positive: a standardized mean difference of 0.29, with a 95% confidence interval from 0.04 to 0.53.

The subgroup split looked even more compelling. Adults aged 66 to 76 had an estimated effect of 0.88; younger participants had an estimate of 0.03, essentially no average difference.

Then two researchers identified a unit-of-analysis error. Several trials had reported multiple memory tests from the same participants. The meta-analysis entered those correlated outcomes as if each were an independent observation. That does not create new randomized people. It makes the evidence look more precise than it is.

The authors accepted the problem and re-analyzed the studies using composite outcomes. The overall estimate fell to 0.19, with a 95% confidence interval from −0.07 to 0.46, and was no longer statistically significant. The older-adult subgroup remained (0.80, 95% CI 0.25 to 1.34), but it rested on few small trials and remains a subgroup rather than a general result.

A second meta-analysis published in 2024 reported a positive memory effect from 16 trials and 492 participants. An April 2026 commentary found the same basic problem: individual trials contributed four to seven correlated memory subtests as separate effect sizes. Until that analysis is corrected with a method that respects the unique participants, its narrow confidence interval should not be read as settled precision.

The largest healthy-adult trial found, at most, a small effect

A preregistered 2023 crossover trial tested 123 participants for six weeks under creatine and placebo. About half were vegetarian and half omnivorous. The two primary tasks measured working memory and reasoning; eight more cognitive tests were exploratory.

The working-memory result leaned toward creatine but did not cross the study’s conventional significance threshold (p=0.064 in the abstract). Its estimated effect size was small (d=0.17). The reasoning result was also small and non-significant (d=0.09, p=0.327). No improvement appeared across the exploratory battery, and vegetarians did not benefit more than omnivores.

Bayesian analyses gave weak-to-moderate support for a small benefit. They also gave strong evidence against the much larger effect from an earlier small study. That is a useful correction: the realistic question is not whether a dramatic nootropic effect has been proved. It is whether a small, context-bound effect can be confirmed by larger trials designed to detect it.

EFSA did not accept a general cognitive claim

In 2024, the European Food Safety Authority evaluated an industry application for a health claim that daily creatine consumption improves cognitive function. The panel considered 21 identified human intervention studies plus two later studies.

Two short studies had found acute working-memory effects under one high short-term protocol. The effect did not recur across lower protocols or continuous consumption in the other studies the panel weighed. A response- inhibition finding was isolated among ten intervention studies in healthy people, and the three disease-population studies did not support the claim.

EFSA concluded that a cause-and-effect relationship had not been established between creatine supplementation and improvement in one or more cognitive domains.

That conclusion has a precise meaning. It does not prove that no person under any condition can benefit. It says the available intervention record did not support the broad claim presented for authorization.

Sleep deprivation may be a different question

The most interesting recent results come from a deliberately abnormal state. In 2024, 15 healthy participants completed a randomized crossover experiment during 21 hours of sleep deprivation. Researchers measured cognition and brain energy-related metabolites after a single 0.35 g/kg study exposure or placebo.

Compared with placebo, the creatine condition performed better on a word-memory task and on processing-time measures for word memory, language, logic and numeric tasks. Brain spectroscopy also registered metabolic changes. The sample was 15, and the analysis pooled repeated post-exposure measurements in a way that can overstate nominal precision.

In 2026, the same group reported a lower-exposure follow-up in 29 participants using the same sleep-loss framework. It pooled three post-exposure observations. Of the all-participant creatine-versus-placebo outcomes, logic (p=0.005) was the only one to clear the paper’s stated Bonferroni threshold of 0.0063; numeric and language results were p=0.04, and a psychomotor-vigilance dispersion result was p=0.02. The paper exempted that pooled comparison from the correction. It found no word-memory benefit.

These are not ordinary daily-life trials. Both used small crossover samples, acute study exposures and severe sleep loss. The follow-up came from the same team and was retrospectively registered because the authors initially did not classify it as a clinical trial. It is not independent confirmation, and most of its pooled outcomes were nominal rather than correction-resistant. The protocols belong to the evidence record, not to consumer instructions.

Older adults remain the live question

The corrected memory synthesis retained its older-adult subgroup. A review published online in 2025 and assigned to a 2026 issue found five positive reports among six older-adult studies. That ratio sounds stronger than the design mix. Four studies were cross-sectional dietary-recall associations, and all four were positive. Only two tested supplementation; just one was explicitly randomized, and that RCT was null. The review called the intervention results conflicting and the evidence insufficient for causal inference.

Age is therefore a hypothesis to test, not a demonstrated moderator. Small subgroups can look large by chance; observational dietary associations can also reflect health, diet, physical activity and social differences rather than a creatine effect.

A stronger trial would prespecify one primary memory outcome, measure baseline and follow-up brain creatine, recruit enough older participants and prevent multiple subtests from masquerading as independent samples.

An Alzheimer pilot cannot answer the healthy-memory question

A 2025 pilot followed 20 people with probable Alzheimer disease through eight weeks of creatine monohydrate. Nineteen met the compliance target, total brain creatine rose 11%, and several cognitive scores improved from baseline.

The study had no placebo group. It was designed primarily to test feasibility, not efficacy. Without a comparator, the observed score changes cannot be separated from practice effects, regression to the mean, concurrent care or time.

The result justifies a randomized trial in that clinical population. It does not show that creatine treats Alzheimer disease, and it cannot be imported into a claim about everyday memory in healthy readers.

Primary sources

• Prokopidis et al., original memory meta-analysis
• Eckert and Pascher, double-counting letter
• Prokopidis et al., corrected author reply
• Citherlet, April 2026 methodological commentary
• Xu et al., 2024 cognitive meta-analysis
• Sandkühler et al., 123-person randomized crossover trial
• EFSA 2024 scientific opinion
• Gordji-Nejad et al., 2024 sleep-deprivation crossover
• 2026 same-team lower-exposure sleep-deprivation follow-up
• 2025-online/2026-issue older-adult systematic review
• Smith et al., 2025 Alzheimer disease pilot

Independent editorial summary. The authors are not affiliated with LifeScore.