This page explains how a NutriScrolls page is made, where the facts come from and how we check them. It also explains what the labels on our pages mean.
How a page is made
Most of our articles are drafted with the help of Claude, an AI by Anthropic. Claude searches PubMed and writes a draft from published studies. Gerritjan Boeve, the editor, then edits the page and decides what goes live.
Every page drafted this way says so in the author line at the top. We think you should know when AI was involved.
The numbers in the nutrient tables are not written by AI. They come from the databases and reference tables listed below, and each one links to its source.
How we check facts
- Every PubMed ID is checked against PubMed before a page goes live.
- The numbers on each study card are checked against the study’s abstract.
- When a study was paid for by the food industry, we say so on the study card.
- Every health benefit gets an evidence label, explained below.
Where our facts come from
- Food composition. Most nutrient values come from USDA FoodData Central. Some values come from Frida, the Danish food composition database of the Technical University of Denmark (DTU). Other named databases cover compounds that USDA does not measure.
- Reference intakes and daily values. These come from the U.S. National Academies of Sciences, Engineering, and Medicine (NASEM), the European Food Safety Authority (EFSA), the U.S. Food and Drug Administration (FDA) and EU labelling rules.
- Health effects. These come from peer-reviewed studies found through PubMed. We prefer systematic reviews, meta-analyses and randomized trials over single small studies.
The How we use data page lists every source, with its version.
Evidence labels
Every health benefit on NutriScrolls carries an evidence label, such as “Evidence: Strong”. It shows how much weight the claim can carry. These are the five labels:
- Strong. Several randomized trials, pooled in a meta-analysis, point the same way, or regulators (FDA/EFSA) have approved the claim. You can rely on this result.
- Moderate. Good randomized trials support it, but there are fewer of them, or their results vary. Likely true; the exact size of the effect is less certain.
- Limited. Small, short or single trials, or mixed results. Promising, but it needs confirmation.
- Observational. A large population study found a link. People who eat more of this food may differ in other ways, so it does not prove cause and effect.
- Lab only. Studies in cells or animals. Useful to explain why something might work, but not proof that it does in people.
How we rate evidence lists what we look at when we choose a label.
What “Reviewed by” means
Some pages show “Reviewed by Gerritjan Boeve” and a date at the top. It means Gerritjan has read the page carefully and checked its claims and numbers against the cited sources.
It is not a medical review. Gerritjan is a health enthusiast, not a doctor or dietitian. The review checks that a page matches its sources, not whether its advice suits you.
Pages without this line have not had this review yet. They have still passed the PubMed and study card checks described above.
Images
The food and nutrient images on NutriScrolls are illustrations made with AI image tools. They show a food in general, not a real portion. We never use them to show data.
Corrections
If you find a mistake, please email [email protected] with the page and what you found. We check every report against the source and correct confirmed errors. The contact page says what to include.
Independence
NutriScrolls has no ads, no sponsors and no affiliate links. No company pays to appear on the site, and no company sees a page before it goes live.