How we grade evidence
Three grades, one rule set
Every supplement gets an overall evidence grade, and every specific claim about it gets its own. The tiers are:
- Strong — multiple randomized controlled trials and/or consistent meta-analyses point in the same direction.
- Moderate — the evidence is positive but mixed, or it holds only in a specific population.
- Limited — few studies, small studies, or mostly negative results.
Claims are graded separately because a single grade for a whole substance hides the useful part. The same supplement can sit at "moderate" for one outcome and "limited" for the three other things it is marketed for.
What the rules actually produce
Of 999 supplements, 27 are graded strong — a little over four percent. That number is the simplest test of whether the tiers mean anything. Rules that hand out top grades freely are not rules, and a catalog where most items "work" is a sales list. We publish the distribution because it is the least flattering and most informative thing about the catalog.
References and verification
The catalog cites more than 2,313 references, and every one is a real, PubMed-indexed paper. Before publication, each citation is checked against PubMed's own API: the title and the PMID have to match a record that exists. Anything that fails the check does not get published — it is corrected or removed, never left in as an approximation. Entries are then re-audited periodically, since references break and new trials change the picture. Evidence was last reviewed in October 2026.
Where AI fits, and where it does not
Analyses are compiled with AI assistance inside this pipeline: it reads and drafts, at a volume a person alone could not cover. What it does not do is decide the standard. The tier definitions and the grading rules are set by humans and applied the same way to every entry, and nothing is published until the reference checks pass. An AI-suggested citation that PubMed cannot confirm simply does not survive to the page.
Who writes this, and how it is corrected
SuppScience is compiled and edited by its founder, who is not a physician; no page carries a medical reviewer's name, because none has reviewed it. What the site offers instead is checkability: every grade points to the studies behind it, every reference is verified on PubMed, and the rules are published above so anyone can see whether they were applied. Trials are included when they are peer-reviewed human studies indexed on PubMed; meta-analyses and systematic reviews weigh more than single trials, and animal or in-vitro work never earns a grade on its own.
Errors are corrected in place and logged here, with the date. Reports go to the contact page.
- September 2026 — Erythritol: a sentence implied it is toxic to dogs; xylitol is, erythritol is not. Corrected in all languages.
- September 2026 — HICA: the one trial cited was in soccer players, not wrestlers. Corrected in all languages.
Limits you should know about
A grade describes the state of published research, not what will happen to you. Trials report averages over groups; dose, formulation, duration and baseline health all move the result. "Limited" means the evidence is thin, not that a substance has been disproven — and "strong" says nothing about safety in your particular situation.
This is educational content, not medical advice. Talk to a doctor before starting a supplement, especially if you take medication or are pregnant. How the site pays for itself is set out in the affiliate disclosure, and the rest of the background is on the about page.