Methodology

We never reproduce an article. We link to the original and quote at most 300 characters of the sentence under discussion. The rest of the row is our own work: what the documents say.

A sentence only becomes a case if it states something a document can confirm or contradict. Adjectives, predictions and opinions are not cases.

Contradicted

A document says the opposite of the published sentence. The document is shown.

Unproven

Nothing we found supports the sentence, and nothing contradicts it either.

The outlet was right

The document confirms what was published. We publish these with the same prominence.

Opinion

Not checkable against a document. It stays in the case for context, without a verdict.

Unlabelled

Still being checked. It is shown neutral, never red.

Before a case is published we contact the outlet and the person named, and we open a window of 72 hours.

If they reply, we publish their text unedited, in full, next to the sentence it answers. We do not shorten it, we do not rebut it in the same row.

If they do not reply, the case says «no reply». That is a fact about what happened, not a verdict about them. A reply sent later is still published, with its date.

A published reply is sealed with a SHA-256 fingerprint of its text and its timestamp. From then on the database refuses to change it. Anyone can recompute the fingerprint from the text on screen.

When we get something wrong we correct it, say what changed, and leave the old version described in the corrections log. A site that judges other people's accuracy has no standing unless its own errors are as visible.

The person who runs this site can send a case and can reply to one as an ordinary user, like anyone else. He cannot approve, label or publish a case in which he is involved. That decision goes to someone else.