Every compound covered on this site carries one of four evidence labels. The labels describe the human evidence — how much of it exists, how consistent it is, how large the effect was, and in whom it was found. They do not describe how promising a mechanism sounds, how long a plant has been used, how good the marketing is, or how much we like the ingredient.
The same four labels are used on every page, worded identically each time, so that a rating on one article means exactly what it means on another.
The four evidence tiers
| Label | Meaning |
|---|---|
| STRONG | Multiple RCTs and/or meta-analyses showing a consistent, clinically meaningful cognitive effect |
| MODERATE | Some RCTs with positive but limited, mixed, or population-specific effects |
| WEAK | Mostly preclinical, small studies, or inconsistent human data |
| INSUFFICIENT/NEGATIVE | Well-studied and largely failed to show benefit, OR essentially unstudied in humans |
In the body of an article the label appears as a badge: Strong Moderate Weak Insufficient / Negative
Note what the bottom tier lumps together. “Studied properly and failed” and “nobody has ever tested this in humans” are different situations, and we say which one applies in the text. What they have in common is that neither gives a reader a reason to buy anything.
Nothing on this site is rated STRONG
Not one ingredient we have reviewed earns STRONG for improving cognition in healthy adults. The honest top of this field is MODERATE, and MODERATE usually means a small effect, in a specific population, in trials that were often small and sometimes funded by the company selling the ingredient.
The one large, long-duration randomized trial in this space with a positive, replicated cognitive outcome is COSMOS — and what worked there was a plain daily multivitamin-mineral in adults over 65, not an exotic botanical. Cocoa flavanols, tested in the same trial, did nothing for cognition. That result is the tier list in miniature, and we walk through it in a daily multivitamin vs a nootropic stack. The full ranking is in our tier list of what actually works.
Population-attached: the rule that does the most work
“Population-attached” means every effect claim on this site names the group of people the effect was observed in. Not “improves memory” — but “improved delayed recall in adults over 65 with mild cognitive impairment,” or “reduced reaction time in healthy young adults roughly an hour after a single dose.”
“Improves memory” is always false as written, because there is no population called everyone. Nobody has ever run a trial on everyone. Every result comes from a defined sample — a particular age range, a particular baseline, a particular dose, a particular test — and a claim that strips those details out is not a summary of the evidence, it is an advertisement.
This is not pedantry, because in this category the same compound routinely gives opposite answers in different populations:
- Ginkgo biloba: a 2026 Cochrane review found a modest six-month benefit in dementia, little or no effect in mild cognitive impairment, and two enormous prevention trials that failed outright. One ingredient, three verdicts.
- Vitamin B12: correcting a deficiency matters enormously. Supplementing someone who is already replete does nothing.
- L-tyrosine: measurable when catecholamines are being depleted by cold, noise, sleep loss or heavy multitasking. Useless at rest.
So we never generalize from a deficient population to a replete one, from an impaired population to a healthy one, or from adults over 65 to a 30-year-old. Those are different questions with different answers.
What we report instead
- Who — age, health status, baseline deficiency, sample size.
- What dose, in milligrams, and in what form.
- For how long — a 12-week effect is not an acute effect, and vice versa.
- How big, in numbers rather than adjectives.
Effect sizes get stated plainly. Creatine’s memory effect sits at a standardized mean difference of roughly 0.29 to 0.31, concentrated in older adults, vegetarians and people who are stressed or sleep-deprived. That is a small effect, and we call it small rather than “significant” — statistical significance and practical size are different things, and the word “significant” is the most reliable place for a supplement page to hide.
What moves a rating up or down
Up: independent replication by a group with no commercial stake; larger trials; longer duration; preregistered protocols and prespecified primary outcomes; and results in a population that is actually being sold the product.
Down: failure to replicate; effects that appear only in one country’s trials; industry sponsorship of every positive study; sample sizes under 100; and outcomes swapped after the fact, where a trial misses its primary endpoint and reports a subgroup instead. A single dramatic result does not raise a tier. Two failed large trials can lower one.
What is not evidence here
Mechanism is not effect. A compound that raises nerve growth factor in cell culture, or NAD+ in blood, has demonstrated a mechanism — nothing more. We label preclinical work as preclinical and say whether the human trial that would test it has been run.
Anecdote is not data. Community reports can be genuinely informative about side effects, and they sit in a labelled box, never mixed into a paragraph about trial results.
And vendor blogs, affiliate review sites, brand-hosted “study” pages and press-release newswires are never cited as evidence that something works. They are cited, occasionally, as evidence about marketing.
A flag, not a tier: do not recommend
Separately from the four tiers, some ingredients carry a Do not recommend flag. This is a regulatory and safety judgment, not an efficacy rating.
Vinpocetine, the racetams, noopept and phenibut are flagged because the FDA’s position is that they are not lawful US dietary ingredients, and it has issued warning letters on that basis — the detail is in our guide to the unapproved drugs sold as nootropics. We do not recommend them and do not link to products containing them. Huperzine A is a separate case: plant-derived and widely sold, but pharmacologically a reversible acetylcholinesterase inhibitor, the same class as prescribed dementia drugs. We discuss it; we never recommend it.
Seeing the tiers applied
The clearest worked example of a WEAK rating is lion’s mane: seven small, short trials with inconsistent results, one of which found worse recall in young adults, sitting underneath a nerve-growth-factor story that comes from rodents and cell culture. The marketing and the evidence are pointing in different directions, and the tier follows the evidence.
If you think a rating here is wrong, tell us what we missed. Ratings change when the evidence does — see the editorial policy for how we handle corrections and how often pages are re-checked.
These statements have not been evaluated by the Food and Drug Administration. Nothing here is intended to diagnose, treat, cure or prevent any disease.
This article is information, not medical advice. Talk to a doctor or pharmacist about your own situation, especially if you take prescription medication.
