How we rate
How we decide what works
Supplement marketing promises a lot. The research behind it is real, but it is written for scientists. We read the research and tell you, plainly, what it found.
Why we do this
Between the label and the lab, there is a gap. On one side is marketing, built to sell. On the other is clinical research, careful but written for experts. Most people are left with what a friend said or a video claimed.
We sit in that gap. We read the research and tell you, in plain words, what it actually found, including when the answer is no.
Evidence over claims
A rating comes from studies, never from a label, a brand or a headline.
Plain words
If we cannot say what a study found in one simple sentence, we have not understood it yet.
Honest about doubt
When we are not sure, we say so, and we say why.
The same rules for every rating
Much of the reading is done with the help of AI. What keeps it honest is the rules. They are the same for every study, and this page explains how they work.
How it works
Every rating goes through the same five steps, in the same order.
- 1Find the sourcesPublished studies and medical guidelines, from the places researchers use.
- 2Read every studyThe same seven questions, every time.
- 3Grade each studyHow good is it, and what did it find?
- 4Combine them into one ratingThe best study on each side, weighed against each other.
- 5Try to prove ourselves wrongWe look as hard for what failed as for what worked.
Step 1 of 5
Where the studies come from
We use studies published in medical journals, and guidelines from medical bodies. We find them in PubMed, Cochrane and Crossref, the same places researchers search.
Before we trust a source, we check it is real. The link has to lead to the study we think it is. That catches more than you would expect.
Not every source counts the same. Before we lean on one, we look at three things.
Who paid for it
Many supplement studies are paid for by the company that sells the product. That does not make them wrong, but it matters. A use cannot be rated Strong if a seller paid for the main study, or if we do not know who paid. We have not found that out for every study yet. That is our gap, and it is why only a few uses are rated Strong today.
Who wrote the guideline
Some medical bodies are fully independent. Others have ties to the industry. Only the most independent can set one of our ratings. We hold 18 of these, from bodies like NICE in the UK and the American College of Rheumatology.
What kind of study it is
A review that pools many trials usually counts for more than one trial. Studies done in animals or cells do not count, and nor do single studies that only watched what people already took. Someone has to have been given the supplement.
Step 2 of 5
We read every study
We hold over 1,000 published studies. For each one, we ask the same questions.
- Does the link to the study work, and is it the study we think it is?
- What kind of study is it, really?
- How many people took part?
- What did it find?
- How good is it?
- What exactly were people given — which form, and how much?
- Can we say what it found in one plain sentence?
So far almost 900 have passed and over 100 have not, because the link was broken or the paper was not what it claimed to be. We have not checked a few yet. A use needs at least one study that passed before it gets any rating at all.
We also check every study against the public record of retracted papers. A study that was retracted or withdrawn does not count, even if it once looked strong. We run that check again over time, because papers can be retracted after we read them.
Step 3 of 5
We grade each study
Each study gets two answers of its own, before we put anything together.
How good is it?
How big it was, how carefully it was run, and whether its results held together. Where the authors or an independent review have already graded it, we use their grade. Most have not, so we grade those ourselves.
What did it find?
It found a benefit, it found none, or it could not tell. A careful study that found nothing is still a good study. Its answer just happens to be no.
Step 4 of 5
We combine them into one rating
We take the best study that found a benefit and the best that found none, and weigh one against the other. Then a few more things decide where the rating lands.
Whether anyone found the same thing
One result is a start. The same result twice is much stronger. A use cannot be rated Strong on one study alone.
Whether they tested the same thing
Magnesium citrate and magnesium oxide are not the same. A study of one does not prove the other works. When a study tested a different form, we say so on the page, and the use cannot be rated Strong. See how the forms compare.
Whether a medical body has already ruled
When one of the independent bodies above has looked at the question, its verdict sets our rating. It searched more widely than we can.
Who was studied
A study of people who were short of a vitamin may not apply to someone who is not. This does not change the rating. We say so on the page, so you can judge for yourself.
What each rating means
The result is one of six ratings. It is worked out from the studies, not picked by us. A new use stays Not verified until we have checked it.
Good studies found it works. Someone else found the same thing. Nothing good says otherwise, and we know who paid for the main study.
For example: Vitamin B2 for headaches/migraines →The studies disagree. Some found a benefit and some did not.
For example: Biotin for brittle nails →Studies point to a benefit, but they are small or weak. A promising start, not proof.
For example: Milk Thistle for acne →It was properly tested, and it did not work. That is a real answer, not a gap.
For example: Glucosamine Sulfate for joint pain →We cannot stand behind this one yet. The page tells you why.
For example: Magnesium Citrate for irregular digestion →Step 5 of 5
We try to prove ourselves wrong
A search can go wrong in either direction. Look only for studies that agree with you, and everything seems to work. Look only for problems, and nothing does. So we look both ways, every time.
We learned that the hard way. Our first review of the ratings only looked for problems. It lowered over 200 of them and raised none. That told us it was looking in one direction, so we changed how we search.
We check the study tested what we name
A study can be real and still be about something slightly different. We go back and ask whether it tested this supplement, in this form, for this problem.
We check one result against its neighbours
When a study shows one use does not work, we ask whether it says something about the uses next to it too.
We settle disagreements on facts
When two readings of a study disagree, we do not pick the more convincing one. We go back to something we can check: what people were given, and who they were.
When we are not sure yet
Of the uses on this site, 90 say Not verified. That is a lot, and we show them on purpose. Hiding them would make us look more certain than we are. A page that says we looked and found nothing useful is worth more to you than no page.
Each one tells you why. The most common reasons are:
- The studies measured something else.
- The study tested a different form from the one on the label.
- We searched and found no study in people.
- We have found studies, but have not finished reading them yet.
What we will not do
Take money from supplement brands
No brand pays us, and none has a say in what we write.
Pick a rating
Every rating is worked out from the studies. If one looks wrong, the fix goes into the studies underneath it.
Soften a result to sound nicer
A No Benefit stays a No Benefit. Being clear about what does not work is a big part of why you can trust what does.