Methodology v1.2 · 7 September 2026
How the Truth Index is scored.
These are the rules in full. They are public because a score you cannot argue with is worth nothing. If a rule below is wrong, write to us and say which one.
This page is free to read and stays free. Nothing in these rules changes for a brand that asks.
What we score
We score claims, not ingredients. A claim is one checkable statement a brand makes about a product on its own page. Each claim is read in the brand's exact wording and checked on its own.
“Fragrance-free”
The published ingredient list, read entry by entry against the claim
“100% natural”
The published ingredient list, and a document for every entry called natural
“5% rosehip extract”
A specification, batch test or certificate stating the figure for this formula
“A potent blend of ginger and turmeric”
Where that botanical sits on the brand's own ingredient list, or a sheet naming the percentage
“Vegan”
The published ingredient list, read against the animal-derived INCI names
“Never tested on animals”
A Leaping Bunny or Cruelty Free International certificate, or a policy naming the markets
“Certified organic”
The certificate itself, in date, naming this product or ingredient
“Wildcrafted in Bulgaria”
An origin declaration or supplier document naming the place and the grower
“Clinically proven to reduce redness”
The study: how many people, how long, what was measured, and by whom
Not claims, and never scored: price, packaging, scent, texture, instructions, and anything a reader could only agree or disagree with. A product page with no checkable claims is not scored at all.
What counts as a document
A document has to name the product or the ingredient in the claim, and it has to speak to the specific statement. A general page about a plant documents nothing.
Marketing copy, a brand's own uncited summary, an influencer post and a supplier's catalogue photograph are not documents. Every document we use is cited with its date and its source.
How the first pass is made
The first reading of a product page is automated and takes about half a minute. It is deterministic: the same page read twice gives the same claims, the same verdicts and the same number. Nothing is inferred and nothing is generated.
The reading log is printed on every score page, so you can see what the machine actually read. A page we cannot fetch is not scored: we say so and offer to score text you paste instead, and the page records that the text came from a visitor. Every score can be disputed by a human, and a document sent to us is read by one.
The three verdicts
A document we have read supports the claim as written. We cite it, with its date.
Nothing we can find supports the claim. It is a statement about our search, not about the product.
A document we have read disagrees with the claim. We quote the line that disagrees.
There is no fourth verdict, and there is no partial credit. A claim we cannot check is Undocumented, and we say what would document it.
How the number is built
Every claim carries a weight. A claim about the product's hero botanical — a potency claim, or an active concentration — carries 2. Every other claim carries 1. The botanical the product is named after is weighted most heavily on purpose: it is the reason she is paying.
Truth Index = 100 × earned weight ÷ total weight, rounded, and never below 0.
A product with 11 claims, 3 documented and 1 contradicted, therefore scores low and reads as it should: mostly unchecked, once contradicted. The counts are always published beside the number, so the number can be recalculated by hand.
What we never do
How to argue with a score
Send the document. That is the whole procedure, and it is the same for a brand and for a shopper.
We reply naming which claim, which document, and exactly what would change the verdict. If the document arrives and it speaks to the claim, the page changes and carries the date it changed. If it does not, the page stays as it is and says that we reviewed it.
Verification is free. There is no queue to pay for, and a brand that sends nothing is not punished for it: its claims are recorded as Undocumented, which is a description, not an accusation.
Write to usChangelog
Three kinds of claim added — Potency, Vegan and Cruelty-free — and the 1% line rule that settles a potency claim from the brand's own ingredient list. Free-from now also reads “free from”, “without” and “no added”. Copy printed inside a product photograph is read; descriptions of photographs are not. Weights, verdicts and the formula are unchanged, but pages read before today are re-read, because more claims are now found on them.
The automated first pass is described above and the published ingredient list is named as a document, for and against a claim. Weights, verdicts and the formula are unchanged, so no score moved.
First published. Weights, the three verdicts and the document list as written above.
When a rule changes, the version number changes, this list gains a row, and every affected page is re-read and re-dated. Old scores are never silently rewritten.