How we work
Everything on this site follows the same procedure. It is written down here so you can hold us to it.
The dose calculation
A standardised botanical extract is declared in two parts: how much extract, and what percentage of it is the compound the marketing talks about. Multiply them and you get the amount that actually matters.
300 mg of fenugreek extract standardised to 2% trigonelline delivers 6 mg of trigonelline.
We run this for every standardised ingredient and publish the result. When a label states an active amount itself, we use it as a check on our own arithmetic — if the two disagree, we say so rather than picking the friendlier number.
Evidence grades
Every claim attached to an ingredient gets one of these labels, applied the same way across the whole site:
| Grade | Meaning |
|---|---|
| Human RCT | Randomised, controlled trial in people |
| Human trial | Interventional but not randomised or not placebo-controlled |
| Human observational | Association only — no one was given the compound |
| Animal | Mice, rats, worms |
| In vitro | Cells in a dish |
| Review | Narrative or systematic summary of other work, not new data |
| None | No study we could find supports the claim as stated |
A mechanism demonstrated in cells is not evidence that a capsule works in a person. We do not blur that line, even when a manufacturer does.
What we check before publishing
No page goes live until all ten are satisfied. One failure blocks publication, including when the claim looks obviously right.
- Every factual claim has a named, linked primary source.
- Every cited study was opened and read — design, population, dose, limitations — not just its abstract summary.
- Every dose and standardisation was checked against the manufacturer's published label.
- The extract-times-standardisation arithmetic was redone and reconciles.
- Every number, date and price was verified on the day of writing.
- Every affiliate link was clicked and lands on the right product.
- No product name appears in a page URL, title or meta description.
- Structured data describes only what is visible on the page and true.
- The affiliate disclosure and the "we have not tested this" notice appear before the first link.
- The author is named, the review date is visible, and any third-party image states its origin.
What we will not do
We do not publish ratings we did not collect, review counts we cannot substantiate, or testimonials we cannot link to. We do not present a manufacturer's photograph as our own. We do not repeat a claim a manufacturer makes if the study behind it does not support it — under advertising rules an endorser is responsible for claims they repeat, and beyond that it would simply make this site worthless.
When we quote user reviews, we report the distribution — how many, the average, the recurring complaints as well as the recurring praise — with links to the source. Selecting only the flattering ones is a form of manipulation, and we treat it as one.
Affiliate rules we operate under
We do not build pages targeting product-brand search terms, we do not bid on brand keywords, and we never present ourselves as a manufacturer or an official store. All commercial links are marked sponsored nofollow and pass through our own redirect so we can tell which page produced which click.
Corrections and updates
Pages carry a review date. When we correct something, we change the page and say what changed. Supplement labels and prices change without notice, so anything time-sensitive is dated at the point of the claim.
AI in the workflow
Research and drafting are AI-assisted. Verification is not: every factual claim is checked against a primary source by a person before publication, and that person is named on every page. The tools widen how much literature one operator can read. They do not decide what is true here.