"Creatine benefits" is not one question. It is at least six, and they have wildly different amounts of evidence behind them. Sorting them by weight of evidence produces a ranking that almost no marketing page reproduces, and one entry on it has already been rejected by a food regulator. This is research journalism and contains no usage guidance, per our editorial standards.
The census, so the ranking is not an opinion
A crude but reproducible measure of how much a question has been studied is how many meta-analyses exist about it. Counted on PubMed on 2026-09-16, restricting to the meta-analysis publication type and requiring "creatine" in the title or abstract, the totals are:
| Endpoint (search terms crossed with creatine) | Meta-analyses |
|---|---|
| Strength / resistance training | 86 |
| Endurance / aerobic | 23 |
| Sarcopenia / older adults | 17 |
| Cognition / memory | 15 |
| Depression | 12 |
| Bone | 4 |
| All creatine meta-analyses, any topic | 457 |
Anyone can re-run those counts. The shape is the point: the muscle question has been asked roughly twenty times more often than the bone question, and a claim's confidence should scale with how hard people have tried to falsify it.
Grade A — muscle phosphocreatine, and short high-intensity work
This is the endpoint creatine was discovered on and it is not in dispute. Supplementation raises intramuscular creatine and phosphocreatine, and the pathway phosphocreatine serves — rapid ATP resynthesis over seconds — is the one where performance improvements show up. The mechanism, the ceiling it runs into and why response varies so much between people are covered on what creatine is; why the extra phosphocreatine shows up as muscle only when there is training to spend it on is read from tracer and biopsy studies on creatine for muscle growth. This is the one sense in which creatine gives "energy" that a regulator accepts; how it differs from feeling less tired is on does creatine give you energy.
The safety literature is unusually large for a supplement: a 2025 analysis assembled side-effect reporting from 685 human clinical trials covering 12,839 supplemented participants and found no significant excess over placebo (Kreider et al., J Int Soc Sports Nutr 2025, PMID 40198156). We read that literature against the adverse-event reports on creatine's side effects. People who start with low stores, vegetarians most clearly, show the largest rises in muscle creatine; those trials are on creatine for vegetarians.
Grade B — body composition, where two measuring methods disagree
A 2024 meta-analysis of 12 randomised trials in adults under 50 found creatine added to resistance training produced +1.14 kg lean body mass (95% CI 0.69 to 1.59), -0.88 percentage points body fat and -0.73 kg fat mass against training alone (Desai et al., J Strength Cond Res 2024, PMID 39074168; doi:10.1519/JSC.0000000000004862).
A 2023 meta-analysis asked the same question of a narrower evidence base — only trials that imaged muscle directly with MRI, CT or ultrasound. Across 44 outcomes from 10 studies the pooled standardised effect was 0.11, with a 95% credible interval of -0.02 to 0.25 (Burke et al., Nutrients 2023, PMID 37432300). The authors describe it as "a very small effect", and the interval includes zero.
Both are competent meta-analyses of overlapping trials. They differ in what they count as muscle: absorptiometry sorts the body into fat, bone and everything else, and water is in "everything else". Imaging measures the tissue. That gap is the whole argument about how much of creatine's weight effect is muscle, and it is the subject of does creatine make you gain weight.
Grade C — older adults, endurance, mood
Seventeen meta-analyses address sarcopenia or older adults; the signal there is real but the trials are small and the outcome measures vary, and it depends on training: the trials, including two two-year bone studies, are read one by one on creatine for older adults. Endurance has 23, and the mechanism argues against a large effect — phosphocreatine serves the energy system that runs out in seconds, not the one that carries a 10 km run. Depression has 12, in a literature where the intervention is usually adjunctive and the trials are short.
None of these is a settled claim. All of them are better supported than the next one.
Grade D — cognition, where a regulator has already ruled
In November 2024 the EFSA Panel on Nutrition, Novel Foods and Food Allergens published its assessment of a health claim for creatine and improvement in cognitive function. The application came from Alzchem Trostberg GmbH, a creatine manufacturer, submitted through the competent authority of Austria. The panel's conclusion is one sentence:
a cause-and-effect relationship has not been established between creatine supplementation and an improvement in cognitive function in one or more of its domains
(EFSA NDA Panel, EFSA Journal 2024;22(11):e9100, PMID 39564533; doi:10.2903/j.efsa.2024.9100) The applicant's trademarked raw material, and the EU claim application it did win, are documented on Creapure explained.
The reasoning is the useful part, because it is about consistency rather than about any single trial. The applicant put forward 21 human intervention studies, plus two more identified later. Weighing them, the panel recorded that:
- the acute working-memory effect appeared in two studies at 20 g/day for 5 to 7 days, and was not seen at lower amounts of 2.2 to 14 g/day, nor with continuous consumption of 5 g/day for six weeks following a five-day loading phase;
- the effect on response inhibition at 20 g/day for seven days was an isolated finding among 10 intervention studies in healthy individuals, with no effects on other cognitive domains;
- the three intervention studies in diseased individuals did not support an effect;
- the available evidence for a mechanism was weak.
That last point has an independent measurement behind it. When a 2017 study measured phosphocreatine in muscle and brain in the same participants on the same protocol, muscle rose 10.3% to 27.6% and brain moved -0.7% to +3.9%, unchanged in every group (Solis et al., J Appl Physiol 2017, PMID 28572496).
What about the meta-analyses that found an effect?
They exist, and they are more interesting read closely than quoted.
Prokopidis et al. 2023 pooled eight randomised trials of memory in healthy people and reported an overall standardised mean difference of 0.29 (95% CI 0.04 to 0.53, I² = 66%). Its own subgroup analysis splits that average in half:
| Subgroup | SMD | 95% CI | p |
|---|---|---|---|
| Adults 66–76 years | 0.88 | 0.22 to 1.55 | 0.009 |
| Participants 11–31 years | 0.03 | −0.14 to 0.20 | 0.72 |
(Nutr Rev 2023;81(4):416–27, PMID 35984306). The headline number is an average over two groups that did not behave alike, and in the younger group the effect is indistinguishable from nothing. That paper also carries a published comment in the same journal.
Xu et al. 2024 pooled 16 randomised trials in 492 participants and found significant effects on memory (SMD 0.31), attention time and processing speed — and no significant improvement in overall cognitive function or executive function. Its own GRADE assessment rates the memory evidence moderate certainty and everything else low. It was corrected by a corrigendum in February 2025 and answered by a published commentary in April 2026 (PMID 39070254; corrigendum PMID 40034739; commentary PMID 42039906).
A body of evidence in which the leading meta-analysis needs correcting, draws a formal commentary, and finds nothing on the global measure is not a body of evidence that supports a confident consumer claim. EFSA said so in a document anyone can download.
The condition-specific claims that attach to this grade — dementia, ADHD, brain fog — have their own, much thinner evidence, and are traced claim by claim in creatine and the brain.
Grade E — bone
Four meta-analyses. That is the whole answer for now.
One thing worth knowing about who wrote this literature
The favourable review and position-stand literature on creatine is written substantially by a small group of researchers who disclose commercial ties, and those disclosures are printed in the papers. The chair of the 2025 safety analysis discloses chairing the scientific advisory board of AlzChem; a co-author of the 2023 imaging meta-analysis discloses serving on the scientific advisory board of Alzchem; a co-author of the 2025 safety analysis discloses membership of Alzchem Group AG's "Creatine for Health" advisory board.
Alzchem Trostberg GmbH is also the company that applied to EFSA for the cognition health claim that the panel refused.
None of that makes any individual study wrong, and the underlying trial data are public and checkable — which is the point of listing primary studies rather than reviews throughout this page. It does mean a reader encountering a confident summary of "the benefits of creatine" is often reading a document produced inside a commercial relationship, and is entitled to know it.
Benefit and risk are separate ledgers, and the risk side now has a page of its own built from the largest posted safety dataset rather than from symptom lists.
The honest summary
- Strong: muscle phosphocreatine loading, and performance in short repeated high-intensity efforts.
- Moderate, method-dependent: lean mass and body composition, where absorptiometry and direct imaging give different answers.
- Unsettled: older adults, endurance, mood.
- Not established, per a regulator: cognitive function. EFSA, November 2024.
The same grades, set beside the measured downsides in one table, are on creatine pros and cons.
- Barely studied: bone.
Two endpoints sit below Grade D because their literatures are barely large enough to grade at all: three small trials have measured what creatine does to sleep itself, and four have measured what it does to testosterone — between them fewer participants than a single one of the strength trials above.
For the other creatine questions this site has traced to their sources, see what the loading studies measured, what the dose figures in circulation actually rest on, creatine and the kidneys and creatine and hair loss. Where a benefit claim meets a clinical condition — diabetes, blood pressure, cancer — the evidence is counted separately and it is thinner.
Grading by endpoint has a counterpart worth reading next to it: how little of this evidence base is checkable against a pre-registered plan. Forty-one registered creatine trials list a muscle-mass outcome and not one has posted results, which is a different weakness from a weak result.
This page reports what has been published and when. It contains no recommendation, and decisions about supplementation belong with a clinician.
Sources
- EFSA Panel on Nutrition, Novel Foods and Food Allergens (NDA). Creatine and improvement in cognitive function: evaluation of a health claim pursuant to Article 13(5) of Regulation (EC) No 1924/2006. EFSA Journal 2024;22(11):e9100 — PMID 39564533 · doi:10.2903/j.efsa.2024.9100
- Desai I, Wewege MA, Jones MD, et al. The effect of creatine supplementation on resistance training-based changes to body composition. J Strength Cond Res 2024;38(10):1813–21 — PMID 39074168 · doi:10.1519/JSC.0000000000004862
- Burke R, Piñero A, Coleman M, et al. The effects of creatine supplementation combined with resistance training on regional measures of muscle hypertrophy. Nutrients 2023;15(9):2116 — PMID 37432300 · doi:10.3390/nu15092116
- Prokopidis K, Giannos P, Triantafyllidis KK, et al. Effects of creatine supplementation on memory in healthy individuals. Nutr Rev 2023;81(4):416–27 — PMID 35984306
- Xu C, Bi S, Zhang W, Luo L. The effects of creatine supplementation on cognitive function in adults. Front Nutr 2024;11:1424972 — PMID 39070254 · doi:10.3389/fnut.2024.1424972 · corrigendum PMID 40034739 · commentary PMID 42039906
- Solis MY, Artioli GG, Otaduy MCG, et al. Effect of age, diet, and tissue type on PCr response to creatine supplementation. J Appl Physiol 2017;123(2):407–14 — PMID 28572496
- Kreider RB, Gonzalez DE, Hines K, et al. Safety of creatine supplementation. J Int Soc Sports Nutr 2025;22(sup1):2488937 — PMID 40198156 · doi:10.1080/15502783.2025.2488937
All sources retrieved 2026-09-16; PubMed counts run the same date. Digital object identifiers verified by content negotiation.