How we fact-check a claim

The verdict scale

✓True

The best available data supports the claim as stated, and the size of the effect is roughly what the claim says.

≈Partly true

The core of the claim holds, but it exaggerates the size, generalises too far or leaves out an important condition.

?No evidence

No reliable data supports the claim, but the data cannot properly rule it out either.

✕False

Reliable data contradicts the claim, for example the observed number is within the normal range we would expect anyway.

…Can't say yet

Data exists, but it is contradictory or too weak for a verdict. We say so instead of guessing.

Every verdict comes with one sentence that explains it, a chart and the date it was checked. When new data changes a verdict, the old verdict, the new verdict and the reason stay visible in the page's history.

Evidence levels

We rate how sure the evidence is on four levels, borrowed from GRADE, the system used to rate health evidence: High, Moderate, Low and Very low. Several large registry studies pointing the same way is high. A single small study or a viral video is very low.

Base rates: how many would happen anyway

Many claims sound convincing because the events are real. The question is whether there are more of them than we would expect anyway. We calculate the expected number from official rates for the same group, period and place, show the normal range of random variation, and compare it with what was observed.

Sources

We prefer official statistics, registries and peer-reviewed research, then transparent surveys, then other sources clearly labelled. Every figure links to its original source with the date it was checked. Our own calculations are marked as ours, with the formula and the data to download.

What people say

For each claim we read a defined sample of forum and comment threads, sort what people say into a few versions of the claim, and have a second reviewer check the sorting. We never name private individuals.

Health claims

Health pages are only published after a named medical or research reviewer has checked them. The reviewer's name and qualifications appear on the page.

How we use AI

We use AI tools to help find sources, draft text and check calculations. A person checks every number against its original source before a page is published, and a second check recalculates our own figures.

Independence

FactsOut has no advertisers, sponsors or affiliate links. If that changes, it will be stated on every affected page.

Corrections

Found an error? Tell us. We fix it, note the change on the page and keep the history.