Compounds › Verified against primary sources

Effect size

How many trials exist is one question; what they found is another — 34 meta-analyses opened

Partially approvedVerified against primary sources

Experimental material. Read before using anything from here.

Nothing on this page is a medical recommendation, prescription or treatment plan. It is an organized translation of protocols circulating in research communities and, where it exists, of what published trials tested. The two are marked differently — and they are not equivalent.

Most of the compounds here have no FDA approval for human use. Several are sold labeled "research use only", which means they have not gone through purity, sterility or dosage controls for human consumption. A community-reported dose is not a validated dose: it is what someone reported having done.

Talk to a licensed health professional before considering any of these compounds. If you already use one and feel anything unexpected, seek care — do not wait for the next routine test.

Specific attention for this compound: Three of the strongest findings on this page have the funder inside the result: the caffeine plus L-theanine meta-analysis has co-authors from Lipton and Unilever; the alpha-GPC trial in healthy people tests a branded product; and the three pieces of evidence for magnesium L-threonate in healthy adults are all for the Magtein ingredient, one signed by employees of the manufacturer. None is invalid because of that, and none is independent.

Summary

The nootropics page counts trials. This one opens up their results. Thirty-four meta-analyses and large trials read on September 4, 2026, with the effect size carried over from the abstract itself. The central finding is that the real divide is not between works and does not work: it is between patient and healthy person. Three popular compounds have a large, negative trial, and one of them was suspended in South Korea. Where an effect exists, it almost always falls between SMD 0.2 and 0.4.

How to read this page

The nootropics page answers how many trials exist. It itself declares the limit of that: the order there is by amount of evidence, not by how well it works. This page is the other half — what the trials found.

Each row gives the effect size as the study itself published it: SMD, Hedge's g, Cohen's d or mean difference, with confidence interval when the abstract provides it. Nothing was calculated, converted or estimated here. Where the study declared GRADE certainty, it is transcribed.

To calibrate the reading: for cognitive outcomes, an SMD of 0.2 is small and 0.5 is moderate. Almost everything that survives a meta-analysis in this area falls between 0.2 and 0.4 — it is not a change you would notice without a standardized test, it is a change that shows up on average, in a group.

The tiers are ordered by whom the evidence applies to, not by magnitude. That is the survey's main finding, and it is explained in the last section.

Tier A — meta-analysis in healthy people

The only group in which aggregated evidence of effect exists in people who are not ill. It is a short list, and two of its items are negative.

Compound	Published effect size	Evidence base	Caveat
Creatine monohydrate	Memory SMD 0.31 (CI 0.18–0.44). GRADE certainty moderate	16 RCTs, 492 people	The highest certainty on this entire page. No effect on executive function or global cognition
Creatine — the cut that matters	Memory SMD 0.88 (CI 0.22–1.55) at ages 66–76 versus 0.03, not significant, at ages 11–31	Separate meta-analysis, 10 RCTs	Doses of 2.2 to 20 g/day and durations of 5 days to 24 weeks did not change the result. What changes is age
Bacopa monnieri	Working memory SMD 2.03 (CI 1.28–2.78) vs. placebo, with SUCRA 100%	Network meta-analysis, 29 RCTs, n=2107	It is the largest effect on the page, and it holds only for doses ≥600 mg/day. No difference in sustained attention, selective attention or processing speed
Ginkgo biloba	Did not separate from placebo on memory; lost to Bacopa at high and low doses	Same network meta-analysis	It is a direct comparison, not an indirect comparison between different studies
Caffeine + L-theanine	Attention switching SMD 0.33 (CI 0.13–0.54); vigilance SMD 0.20 (CI 0.02–0.38), in the 2nd hour. Theanine alone: reaction time SMD −0.35	50 RCTs reviewed, 15 in the meta-analysis	The authors note that the confidence intervals frequently highlight the uncertainty. Two co-authors are from Lipton and Unilever
Nicotine (outside cigarettes)	Positive effect in 6 of 9 domains, size 0.16 to 0.44	41 double-blind placebo-controlled studies	In non-smokers or non-deprived smokers — that is, it is not withdrawal relief. The authors treat the finding as part of the explanation for why dependence persists
Modafinil · methylphenidate · d-amphetamine	Modafinil SMD 0.12 overall. Methylphenidate SMD 0.21. D-amphetamine: no effect	47 studies in all, across three meta-analyses	Literal conclusion of the authors: there is a user perception that these drugs work, and the evidence does not support it. Prescription drugs
Omega-3 (EPA/DHA)	MMSE MD −0.07 (CI −0.25 to 0.10). Nothing	Cochrane review, 3 RCTs of high methodological quality, n=3536	In cognitively healthy older adults. Also no benefit on word learning, digit span or verbal fluency
Omega-3 — the signal that remains	Global cognition g 0.02 (CI −0.12 to 0.15). Only memory holds up: g 0.31 (p=0.003)	25 RCTs in non-demented people; confirmed by a 2025 overview of reviews with 26,881 participants (effect 0.16; CI 0.01–0.32)	No dose-response relationship in either of the two analyses. Still valid as a nutrient — not as a nootropic

Tier B — the evidence exists, but it is all in patients

Here is most of what the market sells as nootropics. The literature is real, sometimes robust, and it is entirely in dementia, stroke, traumatic brain injury or schizophrenia. None of these has a meta-analysis in healthy people.

Extrapolating from the study population to someone who wants to perform better at work is the field's most common error — and it is silent, because the number cited is technically correct.

Compound	Published effect size	Population	Caveat
Huperzine A	Among the 5 best on MMSE and the best on activities of daily living. In schizophrenia: memory quotient WMD 10.59 (CI 5.65–15.53) and IQ WMD 3.97–5.66	Network meta-analysis of 194 RCTs and 21 drugs, in vascular dementia; and 12 RCTs (n=1117) in schizophrenia	The 12 schizophrenia trials were all conducted in China. It is a genuine acetylcholinesterase inhibitor — do not combine with donepezil and the like
L-oxiracetam	+8.97 points on the LOTCA vs. placebo (CI 5.69–12.26), Cohen's d 0.48; and superior to regular oxiracetam	Multicenter double-blind phase 3 RCT, 51 hospitals in China, n=590, in mild to moderate traumatic brain injury	It is the best-designed positive trial on this page. Nothing in it speaks to healthy people
Citicoline	SMD of 0.56 (CI 0.37–0.75) to 1.57 (CI 0.77–2.37), depending on the sensitivity analysis	7 studies in mild cognitive impairment, Alzheimer's or post-stroke dementia; 6 in the meta-analysis	The authors themselves rate the quality of the studies as poor, with significant risk of bias in favor of the intervention. The large effect and the caveat come in the same article
Alpha-GPC	With donepezil: cognition MD 1.72 (CI 0.20–3.25), function MD 0.79, behavior MD −7.61. Alone: MD 3.50 (CI 0.36–6.63)	Meta-analysis of 7 RCTs + 1 cohort, in cognitive dysfunction linked to cerebrovascular injury; replicated in a Korean RCT with n=119	The documented gain is as an adjuvant, not as a standalone routine item
Oxiracetam	No difference. MMSE p=0.49; CDR-SB p=0.38	Multicenter double-blind RCT, n=500, 36 weeks, prevention of post-stroke cognitive decline	The authors write that the finding supports the regulatory decision to suspend use in South Korea. This site's nootropics page lists 22 trials for it — this is the outcome of the largest of them
Cerebrolysin	Failed to demonstrate superiority on mRS and Barthel	Meta-analysis of 7 RCTs, n=1779, acute ischemic stroke	Safe, and without effect. The authors' conclusion: routine use is not supported by the available evidence
Piracetam	Appears among those that “do not appear to be effective”	Systematic review of 44 studies and 22 strategies in dementia with Lewy bodies	In the vascular dementia network it ranks well only on safety profile — which is a different sentence from efficacy, and tends to be cited as if it were the same
Acetyl-L-carnitine	No effect on reaction time, vigilance, immediate memory or delayed recall	Cochrane review devoted to people without cognitive impairment: only 2 eligible RCTs in the entire world literature	Very low quality evidence; the authors state that no conclusion could be drawn. It is the portrait of the field in one line

Tier C — healthy people, but a thin base with the manufacturer inside

Two compounds have a trial in healthy adults with a positive result. In both, the design is small or the sponsor is the interested party — and in both cases this is declared in the article itself.

Compound	Published effect size	Evidence base	Caveat
Alpha-GPC in healthy people	Stroop total d 0.61 with 630 mg and d 0.48 with 315 mg; completion time d 0.56	Double-blind crossover RCT, n=20 trained men, single dose	No difference on Flanker, N-Back, physical performance or growth hormone. n=20, single dose, single author, and the product tested is branded
Magnesium L-threonate	Three positive trials: sleep and mood (n=80, 21 days); memory (n=109, 30 days); total cognition on the NIH Toolbox p=0.043 and reaction p=0.031 (n=100, 6 weeks)	Three RCTs in healthy adults, all with the branded Magtein ingredient	One is signed by employees of AIDP, the ingredient's manufacturer. Another tested a combined formula with phosphatidylserine and vitamins C and D, which prevents attributing the effect to magnesium. The third was conducted by a CRO. There is no independent meta-analysis

Three large, negative trials

This is worth isolating, because it changes the kind of argument that applies. For most of the compounds on this site, the problem is absence of evidence — nobody tested properly. For these three, it is not: the evidence exists, it is large, and it is contrary.

Oxiracetam, n=500, 36 weeks: no difference on either of the two co-primary outcomes, and the authors link the result to the suspension of use in South Korea. Cerebrolysin, n=1779 in 7 RCTs: failed on both efficacy outcomes. Piracetam: listed by name among the ineffective in a systematic review.

All three are still sold, and all three appear on nootropic lists with high trial counts. The count is right. The outcome is what is not.

Whoever pays for the study shows up in the result

Caffeine + L-theanine: the meta-analysis has two co-authors with declared affiliation to Lipton Teas and Infusions and to Unilever.
Alpha-GPC in healthy people: the RCT tests a branded product identified in the article itself, with n=20 and a single dose.
Magnesium L-threonate: the only three pieces of evidence in healthy adults are for the same patented ingredient; one of them has employees of the manufacturer among the authors, and another tests a combined formula that does not isolate the compound.
Pycnogenol, already recorded on the nootropics page, has the same characteristic: a good share of the trials is funded by the brand owner.
None of these studies is invalid because of that. But none is independent evidence — and the difference between it works and the manufacturer showed it works is the only thing separating this field from marketing.

What this survey shows

The real divide is not between works and does not work — it is between patient and healthy person. Citicoline, huperzine A, L-oxiracetam and alpha-GPC have serious literature, and it is all clinical.
Where there is an effect, it is small. SMD of 0.2 to 0.4 in tier A. The exception is Bacopa on working memory, and it moved neither attention nor processing speed.
The gain concentrates in those with a deficit. Creatine: 0.88 in the elderly, 0.03 in the young, with the same dose and the same duration.
The Cochrane review of acetyl-L-carnitine is the portrait of the field. A review devoted precisely to healthy people found two eligible trials in the entire world.
User perception did not match the evidence — the literal conclusion of the stimulants meta-analysis, precisely the block where the sensation of effect is greatest.

What was left out

I read abstracts, not full articles. Every effect size on this page was carried over from the structured abstract of the study itself, which is where a meta-analysis publishes its numbers.
I redid no calculation. I did not convert SMD to Hedge's g, did not recalculate confidence intervals and did not pool studies on my own. Where units differ between rows, they differ because the studies published them that way.
I did not check the regulatory status item by item. The suspension of oxiracetam in South Korea is a claim by the trial's authors, not independent verification at a Korean regulatory source.
The WADA list, that one I did check — at the official source, on September 4, 2026. The 2026 Prohibited List applies, in force since January 1. From section S6.A, non-specified stimulants, prohibited in-competition only, listed by name are: Modafinil, Adrafinil, Bromantan, Fonturacetam [4-phenylpiracetam (carphedon)], Hydrafinil (fluorenol), Fladrafinil, Flmodafinil and Lisdexamfetamine. Meldonium is in another section and is more restricted: S4.4.3, prohibited at all times, not only in-competition — which confirms what this site's meldonium page already said.
Compounds without a meta-analysis were not included. Noopept, Semax, Selank, lion's mane and the minor racetams have no aggregated evidence from which to draw an effect size — they are on the nootropics page, counted, which is what can be done with them today.
I did not search literature outside PubMed. Registries of ongoing trials, grey literature and regional databases were left out.

References

This page did not come from the secondary source. Each effect size was carried over from the structured abstract of the study itself, read on PubMed on September 4, 2026. According to PubMed, the works below are what exists. Section S6.A was checked at WADA's official source on the same date.

← Back to all compounds