Editorial and review policy

Who writes this

Tristan Grech. Every entry on this site is written and source-checked by him, and the responsibility for anything wrong on it is his.

“I am not a doctor or a dietitian. This site exists because every answer to these questions is written by someone selling the thing. Here is exactly how each verdict is reached, and every source I checked it against.”

What “reviewed” means on this site

Entries carry a status, and the word needs defining precisely because elsewhere on the internet it means something stronger.

Reviewed

The sources have been verified by the author. Every mechanism claim was traced to a primary source, and that source was checked to exist and to say what the entry says it says. It does not mean a clinician signed it off.

Draft

Sourcing is incomplete. Draft entries are excluded from search engines and from the sitemap.

How a verdict is reached

Every substance is assessed against five goals separately, because “does this break a fast” has no single answer — it depends on why you are fasting. A substance can be irrelevant to a calorie deficit, disqualifying for autophagy, and forbidden before a blood test, all at once. Two mechanisms decide most verdicts:

• —Energy intake — does it supply calories in a quantity that matters against the goal?
• —Nutrient signalling — does it raise insulin, or supply amino acids that activate mTORC1 and suppress autophagy? This is why a zero-calorie amino acid drink breaks an autophagy fast and a 120-calorie spoon of olive oil does not.

How a depends is decided

Depends is a real answer and is never used to avoid committing. It is used in exactly three situations, and the entry always says which one applies:

1. The product varies. Pure creatine monohydrate and creatine gummies are different answers.
2. The dose decides. A splash of milk and a latte are different answers.
3. The evidence is genuinely mixed. Where trials disagree or the human data does not exist, the entry says so rather than picking the cleaner-sounding side.

What evidence counts

In descending order: systematic reviews and meta-analyses of randomised controlled trials in humans; individual randomised controlled trials in humans; controlled observational studies; mechanistic and animal work. Food composition figures come from USDA FoodData Central rather than from manufacturers.

Animal data is used only to explain a mechanism and is labelled as animal data every time. It never carries a verdict on its own. Manufacturer claims, press releases and secondary reporting count for nothing.

There is a limitation running through nearly all of it, and it is stated on the entries rather than hidden here: almost no study measures what happens to a person who is actually fasting. The trials measure responses to meals. Applying them to a fast is an inference, and where that inference is doing heavy lifting the entry says so.

What would change a verdict

• —A human randomised controlled trial measuring the substance during a fast rather than around a meal.
• —A meta-analysis superseding a single trial the entry currently leans on.
• —A reformulation that changes what is actually in a product.
• —A reader showing that a cited source does not say what the entry claims. This has the same standing as any of the above.

Citations, and how to check them

Every mechanism claim carries a primary source, shown on the page next to the claim with its journal, year and PubMed ID rather than hidden in a footer. There are currently 103 sources behind 41 entries.

Those citations are checked by machine, not only by the person who wrote them. A committed script re-queries the PubMed API for every PMID on the site and fails on any record that has stopped resolving, or whose title, journal or year no longer matches what PubMed returns. Every citation here was verified that way on 15 August 2026 and all of them passed.

To be precise about the limit of that: the script is run before publishing. It is not yet wired to run automatically on every deploy, so it catches drift when someone runs it rather than the moment it happens. A fabricated or mistyped citation looks like diligence while being the opposite of it, so it matters that this is a real check — and equally that we do not describe it as more automatic than it is.

Corrections

Corrections are logged publicly, with the date and what changed, on the entry itself and in the site-wide changelog. Nothing is edited silently. If a verdict changes, the old verdict and the reason it changed stay visible.

If something here is wrong — and on a site of this size, something will be — the report we can act on fastest names the entry, the specific claim, and the evidence against it. Corrections showing that a cited source does not support the claim attached to it are acted on first.

Independence, and how this site makes money

Fasting Index sells no supplements and takes no affiliate revenue from any product it assesses. It is funded by its own guides. This is a permanent constraint rather than a current state of affairs, and it is enforced in the codebase: a build that tries to add a supplement to the catalogue fails.

Brands are named where naming them answers the question — a reader searching for a specific electrolyte sachet deserves a verdict on it — but a brand never receives a link, a recommendation, or a ranking.

The reviewer slot is built and empty

Every entry has a field for a clinical reviewer’s name and the date of review. It renders nothing while it is empty, which is today and every day so far. It exists so that engaging a clinician later upgrades all 41 entries at once instead of requiring the site to be rebuilt around them.

See also the medical disclaimer, the AI disclosure, and the methodology.