I spent the week reading primary literature on supplements that are supposed to make you sharper. I came in expecting to write about creatine, or maybe a lutein trial nobody had heard of.

I came out looking at a $10 bottle of Centrum Silver.

That is not the answer the supplement aisle sells, and it is not the answer I wanted. But it is what the largest randomized cognitive trial ever run on this question found, and it changed how I think about the entire nootropic category.

Do “brain vitamins” actually work?

Brain vitamins, no. A plain multivitamin, modestly yes — and that distinction is the whole article.

Here is what I found when I stopped reading supplement marketing and started reading trial results.

The trial the nootropic aisle never cites

COSMOS (the COcoa Supplement and Multivitamin Outcomes Study) is one of the largest supplement trials ever conducted, and it ran three separate cognitive sub-studies with genuinely rigorous designs. COSMOS-Mind alone randomized 2,262 people with a mean age of 73. The other two sub-studies added more, and the researchers eventually pooled all three.

The supplement arm did not test some boutique “brain formula.” It tested Centrum Silver — the multivitamin at your grocery store.

What the multivitamin actually did

Study	What was measured	Result
COSMOS-Mind (n=2,262)	Global cognition	Significant benefit: z = 0.07 (95% CI 0.02–0.12), P = 0.007
COSMOS-Web	Immediate recall (primary endpoint)	Better at 1 year (P = 0.025) and across 3 years (P = 0.011)
COSMOS-Clinic	Episodic memory, 2-year change	+0.12 SD (0.002–0.23) — significant
COSMOS-Clinic	Global cognition, that study alone	+0.06 SD (−0.003–0.13) — did not reach significance
COSMOS-Clinic	Executive function / attention	+0.04 SD — no significant effect
Pooled meta-analysis	Global cognition	+0.07 SD (0.03–0.11), P = 0.0009
Pooled meta-analysis	Episodic memory	+0.06 SD (0.03–0.10), P = 0.0007

Two things in that table matter more than the headline.

First, the effect is specific to memory. Executive function and attention did not move. If you were hoping a multivitamin fixes your focus, it does not.

Second, look at the global cognition row for COSMOS-Clinic alone: the confidence interval crosses zero. The clean, significant result only appeared when the three studies were pooled. That is a real caveat, and most articles skip past it.

What a 0.07 SD effect means in plain English

The authors did the translation themselves: the effect on global cognition was equivalent to reducing cognitive aging by about two years.

That is not nothing. It is also not a nootropic. A 0.07 SD shift is invisible to you as an individual — you will not feel it, and if you take a multivitamin for a month and swear you think faster, that is expectancy talking.

Where it does matter is population scale and cost. Pennies a day, no meaningful risk, and a two-year difference in cognitive trajectory. That is a good deal. Just don’t mistake it for a smart pill.

Conflict of interest, stated plainly, because it belongs in the article: the multivitamin arm of COSMOS was supported by the manufacturer, and the cocoa arm by Mars Edge. That does not make the data wrong — the design was strong and the results are published in full — but you should weight it accordingly.

The nootropic aisle, by the numbers

Here is every intervention worth knowing about, with the effect size I could actually verify from the primary literature.

Intervention	Best evidence	Effect
Exercise (not a pill)	Umbrella review: 133 reviews, 2,724 RCTs, 258,279 people	General cognition SMD 0.42, memory 0.26, executive 0.24
Creatine 5 g/day	Meta-analysis, 16 RCTs	Memory SMD 0.31 (0.18–0.44); no effect on global cognition or executive function
Multivitamin	COSMOS pooled	Global cognition 0.07 SD; memory 0.06 SD
B vitamins (folate, B12, B6)	Meta-analysis, 95 studies, 46,175 people	MMSE +0.14 points — statistically real, clinically tiny, and only over >12 months
Magnesium bisglycinate	2025 RCT, poor sleepers	d = 0.2 on insomnia severity
Glycine 3 g before bed	Small trials, plus a mechanistic basis (core temperature)	Improved subjective sleep quality. Smaller evidence base than the others here.
Bacopa 300 mg	Meta-analysis, 9 studies	Choice reaction time −10.6 ms. Milliseconds.
Ginkgo	3,069 people aged 75+, median 6.1 years	No dementia prevention (HR 1.12, P = 0.21)
NMN / NR	Systematic review, 2026	Raises blood NAD+. Outcomes “inconclusive.”

Read that table honestly and two things fall out. Exercise beats every pill on the list by a wide margin, and it is free. And creatine has a larger memory effect than the multivitamin — more than four times larger, on the one domain it works on.

So I won’t pretend the multivitamin crushed everything. What I will say is that the $80 nootropic stack loses to both, and both cost about $10–30 a month.

Creatine is the best actual nootropic, and almost nobody markets it that way

Creatine’s brain story is better than its reputation. The 2024 meta-analysis found a real memory benefit (SMD 0.31), and a single high dose (0.35 g/kg) partially reversed sleep-deprivation cognitive decline in a 2024 trial, with a lower dose still improving performance by up to 12% in the follow-up.

There is a genuine disagreement in the literature about who benefits, and I’d rather flag it than pick the flattering version. One meta-analysis reports the benefit is larger in people aged 18–60 and in women. A separate meta-analysis on memory reports the opposite — that older adults aged 66–76 gained most, with almost no effect in younger participants. Both are real analyses of real trials, and they contradict each other on the population question. The safest summary is that the effect exists, is moderate, and is most reliable in older adults, vegetarians, and the sleep-deprived.

The other honest caveat: a 2026 critical review asked whether we have “put the cart before the horse,” noting that most studies never confirmed that supplementing actually raises brain creatine. The memory effect looks real and modest; the mechanism is less settled than the marketing implies.

It is still the only thing I found that appears on both the “feel better” list and the “think better” list.

The $6 bottle that beat both — and it starts with a blood test

The single highest-yield finding of the entire week was not a nootropic at all.

A randomized trial in non-anemic menstruating women with unexplained fatigue — normal hemoglobin, low ferritin — found that iron supplementation dropped fatigue scores 47.7%, versus 28.8% on placebo (difference −18.9%, 95% CI −34.5 to −3.2; P = 0.02).

And the part that makes it credible: iron did not improve quality of life (p=0.2), depression (p=0.97), or anxiety (p=0.5). It moved fatigue and blood markers and nothing else. A trial that only reported the win would be less believable; this one is specific, which is what a real effect looks like.

The cutoff that mattered was a ferritin under 50 µg/L. That is well above the threshold most labs flag as abnormal, which means a large number of people are told they are “normal” while sitting squarely in the range where correcting it produced a 19-point swing in how tired they felt. That is a prescription-drug-sized effect, from a bottle that costs a few dollars.

If you are tired and your labs are “fine,” get ferritin checked specifically, and look at the number rather than the flag. Same logic for vitamin D and B12. Correcting a measured deficiency is the only place in this entire category where the effect size is genuinely large — and it is the one thing no supplement company can sell you, because you need a test to know whether you should buy it.

Where I have to correct the popular cocoa claim

Cocoa flavanols are the darling of wellness writing, and the coverage is misleading.

In COSMOS-Clinic, cocoa extract had no significant effect on 2-year global cognition (−0.01 SD, 95% CI −0.08 to 0.05), episodic memory, or executive function. In COSMOS-Mind, cocoa extract also showed no effect on global cognition (mean z = 0.03, 95% CI −0.02 to 0.08, P = 0.28). The positive flavanol result that gets circulated came from a subgroup analysis — older adults with poorer diet quality and lower habitual flavanol intake. That subgroup finding is legitimate science, but it is not the headline people quote.

The accurate summary: null overall, possible benefit if your diet is already bad. Which is an argument for eating better, not for buying capsules.

What I actually take

Full disclosure, because that’s the point of the format. This is the current rotation.

What	Why
GAT SPORT Men’s Multi Plus Test — 2 tablets with a meal	My multivitamin. See the long explanation below, because the reason is not the one on the front of the bottle.
Nutricost Creatine Monohydrate — 5 g daily	The only item that landed on both the memory list and the physical-performance list. Mandarin orange, mixes fine in water.
Horbäach Magnesium Glycinate powder — before bed	Effect size is only 0.2, so some of this is placebo. It’s cheap, it’s pleasant, and I’ll take a pleasant placebo with a small real effect over nothing.
Nutricost Glycine powder — 3 g before bed	Stacked with the magnesium. Glycine has a genuine physiological hook — it lowers core body temperature, which is one of the signals your body uses to initiate sleep — but the human trials behind it are small. Cheap enough that I don’t mind the uncertainty.
Nutricost L-Theanine 200 mg — with morning coffee	Not in the COSMOS data, but caffeine plus theanine is the best-supported “clean focus” combination that exists. Theanine alone does less than people claim.
Carlyle Melatonin 300 mcg — only for schedule shifts	Deliberately the 300 microgram dose, not the 5 or 10 mg on most shelves. This moves my clock; it doesn’t sedate me. Nightly use isn’t what the evidence supports.

I take a “testosterone” multivitamin, and it has nothing to do with testosterone

Yes, my multivitamin says Multi Plus Test on the label, and yes, it contains tribulus fruit powder. Let’s deal with that first, because it’s the obvious objection.

Tribulus doesn’t work. It’s one of the more thoroughly failed ingredients in the supplement world — controlled trials show no meaningful testosterone increase. It’s in there as label decoration, a few milligrams of fruit powder riding along so the word “Test” can go on the front. I’m not defending it and I’m not taking it for that. If the tribulus in my multi were doing anything, that would be a problem, not a feature.

The actual reason I take this one is vitamin A.

The gene thing, explained properly

Most multivitamins deliver vitamin A as beta-carotene. Beta-carotene is a provitamin — your body has to convert it into retinol before it can use it. The enzyme that does that conversion is BCMO1 (β-carotene 15,15′-monooxygenase).

And BCMO1 activity varies enormously between people because of common genetic variants. A Journal of Nutrition study found that several single-nucleotide polymorphisms near the BCMO1 gene significantly influence how efficiently someone converts plant-based provitamin A into usable vitamin A — and that the frequency of the affected alleles ranges from 43% to 84% depending on ethnic origin. A separate study in 693 Filipino children and adolescents found combined loss-of-function variants affecting roughly 7.6% of the group, with one variant inversely associated with actual vitamin A status.

I’m one of the people with the low-conversion variant. Note the precision, because it matters: this isn’t an absorption problem. I absorb beta-carotene fine. I just don’t turn it into retinol efficiently. So a multivitamin with 100% beta-carotene — which is most of them — gives me the carotenoid and not much of the vitamin.

Retinyl palmitate is preformed retinol. It’s already past the step I’m bad at, so it’s absorbed and used directly, no BCMO1 required. It’s also the form that actually has a toxicity ceiling, which is how you know it’s the biologically active version — there’s no tolerable upper limit for beta-carotene, because your body rations the conversion.

Here’s what the four multis I’ve run actually use, since this is the difference nobody puts on the front of the bottle:

Multivitamin	Vitamin A form	Conversion needed?
GAT SPORT Men’s Multi Plus Test	Retinyl Palmitate — 1,500 mcg RAE (≈5,000 IU)	No — 100% preformed
NutraBio Multisport for Men	Beta carotene	Yes — all of it
Optimum Nutrition Opti-Men	Beta carotene, mixed carotenoids	Yes — all of it
NAKED Men’s Multi	Retinyl palmitate and beta carotene	Partly — the best of the other three

I’ve taken all four. They’re all decent multis, and if you convert beta-carotene normally, any of them will do the same job. NutraBio and Opti-Men are perfectly good products — they just don’t solve my problem, because every microgram of their vitamin A lands on the exact step I’m bad at.

NAKED is the one I’d point someone to if they want preformed retinol without the testosterone branding. It has retinyl palmitate plus beta-carotene, so you get some direct retinol and some plant-sourced carotenoids. That’s arguably the better-formulated choice, and it’s the one I’d recommend to most people reading this.

I take the GAT one because it’s 100% preformed at a dose I’m happy with, and it’s the cheapest per serving of the options that actually work for me. That’s the whole reason. The tribulus is along for the ride.

One real caution on preformed vitamin A

Because this is the form with a ceiling, it’s worth doing the math. The tolerable upper intake level for preformed vitamin A in adults is 3,000 mcg RAE per day. At 1,500 mcg, this multi is half of it. That’s fine on its own — but it means you should not stack other retinol sources on top: cod liver oil, additional “vitamin A” capsules, or a second multi. A multivitamin plus a separate vitamin A supplement plus cod liver oil is how people get into trouble, and it’s the same cumulative-dosing trap I mentioned earlier for B6, zinc, and iodine.

What I would buy, in order

• A basic multivitamin. The trial used Centrum Silver. A generic with a USP Verified or NSF seal is fine — the seal is what tells you it contains what the label says. This is the single most defensible purchase in the category. Unless you’re a poor beta-carotene converter — then check the vitamin A form first, because most multis get it wrong for you.
• Creatine monohydrate, 5 g/day. Plain monohydrate, no “advanced” version.
• Vitamin D, B12, or iron — only with a lab result. Iron especially. Do not take iron on a guess. If you do supplement D without a test, 5,000 IU is a reasonable daily dose for most adults in winter; the 50,000 IU versions are a prescription-strength weekly dose, not a daily vitamin.
• Magnesium glycinate, if you like it. Cheap, low-risk, small real effect.
• Omega-3, 2–3 g EPA/DHA — the one I’ve rebought more than anything else. It does nothing for cognition specifically, but it’s the item on this page I’d defend longest for general health.

Who this is for, and who should skip it

Reasonable if: you are 60+, your diet is genuinely poor, or your intake of fruits and vegetables is low. This is the population that showed the benefit.

Skip it if: you are young, eat reasonably well, and are already taking a multivitamin. You are paying for an effect that is invisible at your age and scale.

Be careful if: you take warfarin (many multivitamins contain vitamin K), have kidney disease (watch potassium and the vitamin A forms), or are stacking multiple products. That last one is the real hazard. Nobody adds up the total vitamin A, B6, zinc, and iodine across five bottles — and that is where people get hurt, not from any single sensible dose.

What I would skip, and why

• “Brain health” formulas. Usually a standard multivitamin plus caffeine, theanine, and a markup. The evidence is for the plain version.
• Anything bought for testosterone support. Tribulus fails in controlled trials, and the category is the one the FDA has repeatedly caught containing undeclared anabolic steroids. I’ll take a multivitamin that happens to contain tribulus as filler, but I would never buy a product because it claims to raise testosterone.
• Nootropic blends with proprietary blends. If the doses are hidden, you cannot evaluate it, and hiding them is a choice.
• Greens powders. A 10–12 g scoop of dried grass. The ORAC numbers they quote come from a USDA database withdrawn in 2012, which stated those values have no established link to human health.
• Ashwagandha. A 2026 meta-analysis reports an effect size of −5.88 for stress. No psychological or psychiatric intervention in medicine produces an effect that large — not CBT, not SSRIs. An effect that size is proof of bias in the pooled trials, not of power. Add a 2026 scoping review documenting 25 patients with liver injury across 13 publications — including one acute liver failure requiring transplant and three deaths among patients with pre-existing cirrhosis — and “it’s natural so it’s safe” stops working.
• NMN and NR. They raise NAD+ on a blood test and have inconclusive clinical outcomes.
• Ginkgo. Failed the largest, best-designed trial of dementia prevention.
• Melatonin as a nightly sleep aid. A 2025 analysis of 110 products found melatonin content ranging from 0% to 667% of the label claim, with servings from 0.042 mg to 50 mg. Melatonin shifts your body clock — that is its real job — but it does not add total sleep time. And by 2020 it had become the substance most frequently ingested by children reported to US poison control centers, which is worth knowing before you leave gummies on a nightstand.

Do brain vitamins actually work? (FAQ)

Which multivitamins are good for brain health?

The evidence supports a standard multivitamin-mineral — the trial used Centrum Silver. It does not support specialty “brain” formulas, which cost more and have not been tested at scale. Look for a USP Verified or NSF seal so you know the label is accurate. If you convert beta-carotene poorly, also check whether the vitamin A is retinyl palmitate rather than beta carotene.

What is the #1 brain supplement?

There isn’t one. Creatine has the largest memory effect in the literature (SMD 0.31), a multivitamin has the broadest trial evidence, and exercise beats both (SMD 0.42) while costing nothing. Anyone selling you a single “number one” is selling.

Is a multivitamin good for your brain?

Modestly. Pooled COSMOS data show about 0.06–0.07 SD improvement in global cognition and episodic memory, which the authors equated to roughly two years less cognitive aging. Real, small, and specific to memory — not to focus or attention.

Do brain supplements actually work?

Not the ones marketed as brain supplements. The boring multivitamin has better evidence than almost everything with “nootropic” on the label, and the one intervention that clearly outperforms the entire category is exercise.

Bottom line

I went looking for a pill that makes you smarter and found a multivitamin, a creatine tub, and a ferritin test. Combined, those three cost less than a month of one boutique nootropic stack, and they have better evidence than everything on the shelf next to them.

The one genuinely useful thing I learned from my own shopping habits is the vitamin A form thing. Nobody markets it, no supplement brand leads with it, and for a chunk of the population it’s the difference between a multivitamin that works and one that quietly does nothing.

No magic pill. Just a few small, real effects — and one boring blood test that produces a bigger swing than any of them.

Not medical advice. I’m a guy who reads trials, not your doctor. Check ferritin before taking iron, and talk to a physician if you take warfarin, have kidney disease, or take any prescription medication. Some links above are affiliate links — if you buy through them I may earn a commission at no extra cost to you.

Sources

• Multivitamin and cognition meta-analysis, COSMOS-Clinic — Am J Clin Nutr 2024
• Multivitamin improves memory in older adults (COSMOS-Web) — Am J Clin Nutr 2023
• Cocoa extract and multivitamin, cognitive function (COSMOS-Mind) — Alzheimer’s & Dementia 2023
• Cocoa extract in the COSMOS clinic subcohort — Am J Clin Nutr 2024
• Dietary flavanols and hippocampal memory (subgroup) — PNAS 2023
• Iron supplementation for fatigue in non-anemic women — CMAJ 2012
• Creatine and cognitive function, meta-analysis — Front Nutr 2024
• Creatine and memory in healthy individuals — Nutr Rev 2023
• Single-dose creatine during sleep deprivation — Sci Rep 2024
• Single-dose creatine, lower dose follow-up — Nutrients 2026
• Exercise for cognition, umbrella review — Br J Sports Med 2025
• Magnesium bisglycinate and sleep quality — Nat Sci Sleep 2025
• Glycine and sleep quality — J Pharmacol Sci 2012
• Sleep-promoting and hypothermic effects of glycine (mechanism) — Neuropsychopharmacology 2015
• B vitamins and prevention of cognitive decline, meta-analysis — Nutr Rev 2022
• Bacopa monnieri, meta-analysis — J Ethnopharmacol 2014
• Ginkgo biloba for prevention of dementia — JAMA 2008
• Ashwagandha meta-analysis — Complement Ther Med 2026
• Ashwagandha-associated liver injury, scoping review — Cureus 2026
• Melatonin content in dietary supplements — Drug Test Anal 2025
• Pediatric melatonin ingestions — MMWR 2022
• NAD+ supplementation systematic review — Ageing Res Rev 2026
• BCMO1 SNPs and provitamin A conversion efficiency — J Nutr 2012
• Genotype effects on β-carotene conversion to vitamin A — Food Nutr Bull 2022
• FDA — Caution: bodybuilding products can be risky (undeclared steroids and steroid-like substances) — FDA
• Vitamin A form comparison — GAT SPORT (retinyl palmitate 1,500 mcg RAE), NutraBio (beta carotene), Opti-Men (beta carotene), NAKED (retinyl palmitate + beta carotene) — product labels