Plenty of health websites repeat the same three sentences about a study, often with the same mistake copied from site to site. The rule here is different. A page cites the paper itself, not someone else’s summary of it, and links it so you can read it too.
The four evidence grades
The site grades health claims on a four-step scale. A grade belongs to a page’s main claim, not to every sentence on it, and it sits in the badge near the top. Right now 103 of the site’s 372 pages carry one. A page without a badge hasn’t been graded, so read its health statements with that in mind.
- Strong human evidence. Several well-run randomized controlled trials, or a meta-analysis of them whose trials agree without large unexplained differences.
- Moderate human evidence. One human trial with a clear result on its main outcome, or a few small ones that point the same way, or a meta-analysis whose trials differ widely. Promising, not settled.
- Preliminary evidence. Early or inconclusive: lab or animal research, a human study that was small or uncontrolled, or human data that found no clear effect on its main outcome.
- Traditional or untested. No clinical trial of the claim yet: either centuries of traditional use, or a claim nobody has tested. Traditional doesn't mean wrong. It means unproven.
Examples from this site: sea buckthorn oil for dry-eye symptoms rests on one trial of 100 people, and blood lipids rest on two meta-analyses of 11 and 15 trials that found lower lipids but whose trials differ a lot, against a 2026 meta-analysis of 12 trials that found no significant pooled effect. Most of the exciting claims about sea buckthorn sit at Preliminary, and the pages say so when they do.
What counts as a source
We cite primary research: randomized trials, systematic reviews, and meta-analyses. Review papers help map a topic, but they don’t carry claims on their own. Brand-funded studies get used when they’re the only human data that exists, and the pages that use them say who paid.
Here’s the whole shelf. The site’s pages cite 97 distinct sources. Of those, 72 are matched to their PubMed record and 16 to their DOI record at Crossref. The other 9 are regulator documents and papers with no PubMed or DOI record, and each of those was matched against the document itself. Across the site, 241 of 372 pages list sources. Every benefit, safety, dosage, or research page lists the sources it relies on, and no page lists a source that none of its sentences use.
We publish the results we didn’t want
The 2022 meta-analysis we cite most also found that sea buckthorn did nothing significant for blood sugar, blood pressure, or BMI. A lot of sites selling sea buckthorn quietly skip that part. You’ll find it on our cholesterol, weight loss, and dosage pages, because a site that hides null results is marketing, not information.
Mistakes
There will be some. When one turns up, the page gets corrected under the corrections policy. If you spot one first, email hello@seabuckthornworld.com with the study or the source.
The source registry
No page cites a study the registry does not hold. The registry records the citation, the record type, and the retraction status of the sources it holds, and its records come from PubMed and Crossref. A page draws its citations from the registry, never from another website’s summary of a summary. That one habit stops most of the errors that travel between health blogs.
How we handle brand-funded studies
Sometimes the only human data comes from the people selling the ingredient. We use it, because weak evidence beats no evidence, and we label it on the page. The skin research is the clearest example. The best-known trial used a branded oil and had no placebo arm for the oral group, so we print its caveats next to the results, as the skin studies page shows in public.
Claims we refuse to make
Some claims fail our bar no matter how often the industry repeats them. Sea buckthorn doesn’t cure or treat any disease. It isn’t a cancer treatment. It did not lower blood sugar, blood pressure or BMI in the meta-analysis that tested them. It isn’t clinically proven to boost immunity, because the one cold trial came back null. And concentrated oil supplements aren’t proven safe in pregnancy, because no trial has studied them there. If a page on this site ever says otherwise, it’s a bug, so report it.
We check whether the paper is still standing
A published study isn’t permanently true. Journals retract papers, and when that happens the study doesn’t become weak evidence. It stops being evidence at all. Most sites never notice, because they cited the paper once and never looked again.
So each cited PubMed record and DOI is re-checked for a retraction notice every month. Sources with no PubMed or DOI record aren’t covered by that check. This isn’t hypothetical. One of the trials this site used to cite, a 2017 study of seed oil in adults with high blood pressure, was retracted by Clinical Nutrition in 2026. It was the one oil trial that had reported lower blood pressure, and it’s gone.
We didn’t quietly delete it. The citation stays in the source lists, labelled as retracted, because the retraction is now the useful fact. The pages that leaned on it were updated, and our answer on blood pressure changed to match the meta-analysis that found no effect. You can read the result on the blood pressure page.
When evidence conflicts
Two studies disagree sometimes, and the resolution is boring. Better design wins. A meta-analysis of randomized trials outranks a single small trial, and a placebo-controlled trial outranks an open-label one. The how to read studies guide teaches the same ladder. When the conflict is real and unresolved, the page says so, and the research hub keeps both sides on the shelf.
Frequently asked questions
Why do your pages sometimes contradict a product's marketing?
Because marketing cites the best-case result and stops. We cite the whole picture, including the trials that found nothing. The 2022 meta-analysis behind the cholesterol claims also found no effect on blood sugar, blood pressure, or BMI, and a label will never tell you that part. Publishing both halves is the entire point of this site.
Illustration.