We re-checked the evidence behind 1,200+ supplement claims

What we did

Every finding on Winnowleaf is one supplement, one outcome and one grade, such as a supplement and its effect on sleep. Each one cites studies. For each finding we opened every study it cites on PubMed, read the abstract, and asked:

• Is the study real, and does its PubMed record match what we cite?
• Is it a study of this supplement, in the form people take, rather than a different compound, a cream or an injection?
• Did it measure this outcome, in people the finding applies to?
• Does the finding say what the studies found: which way the effect goes, how big it is, and how strong the evidence is?

Studies that failed were taken off the finding. The result, size and grade were then set again from the human studies that were left. Grades now rest on human studies only, so a finding backed only by animal or lab work shows no grade. The full checklist is in What "Reviewed" means.

What changed

• 1,155 of the 1,208 findings changed in a way you can see on the page: a study taken off or added, or a new result, size, grade or count of people studied.
• 160 grades went down, and 417 findings lost their grade because no human study supported them.
• We took 2,732 study links off supplement findings they did not support, and added 3,920 that do.
• 54 findings now say "Worsens" where the studies found a harm and the page used to say something else. 14 lost a "Worsens" label their studies did not support.

What you will see on the site

• "Reviewed" and a date beside each re-checked finding, on pages in our new layout, which most supplement and condition pages now use. A finding we are still checking shows "Under review" and no grade.
• Harms stated plainly. When the studies found a harm, the finding says so. A few protocol pages that listed a supplement where trials found harm now say "Not recommended" and give no amount.
• Picks only from checked findings. The picks on the home page and the lead finding at the top of a page never feature a finding that is under review, and our tools, such as the myth checker, answer only from re-checked findings.
• A public corrections log. The corrections log lists every fix, newest first: what the page said, what it says now, and why.

What is still to do

Most findings on the site have not been through this check yet. They carry no "Reviewed" label until they have, and the check goes on. Each new batch of fixes appears in the corrections log the day it goes live. If you find a finding that looks wrong, our editorial policy says how to tell us.

The evidence itself stays where it was: on each supplement and condition page, and in our guides.

How we counted

These numbers are worked out from our data each time the site is built, and count only changes dated on or before 8 November 2026:

• A claim is a finding on a supplement page that showed a grade from A to D when the check began. Peptide research pages were checked too (823 findings) and are left out of every other number here.
• Re-checked means the finding went through every step above and carries a review stamp in our database.
• Changed counts the same kinds of edit the corrections log lists. A study taken off and later added back counts as neither.
• The "Worsens" counts compare the label a finding showed before the check with the label it shows now, across all 2,918 re-checked supplement findings, graded or not.

The evidence check writes one change file for each batch of fixes, and the site builds both the corrections log and these numbers from those files.