Questions we get
Answers for the person who has to sign off on a claim, and for the writer who has to find the source.
Every factual claim in a piece of content: whether the citation behind it leads to a document that states the claim, and which source would be stronger to cite. One verdict per claim, plus the document itself.
A model answers from memory and will produce a confident citation that does not exist. We retrieve the actual document, then check the claim against it. If we cannot get the document, we say so instead of guessing.
Nobody can promise that. What we can do is make sure the citations on your page point at the authority that published the number and that the claim is written so it can be quoted and attributed cleanly.
Yes. Most teams start there, because the library is where untraceable statistics accumulate. You get a claim-level view of which pages are worth revisiting first.
The corpus that governs the claim — PubMed and agency publications for health, filings and regulator documents for finance, official statistics for economic figures. Pages that have gone offline resolve through web archives.
No. It gives your reviewers the evidence before they start, so their time goes to judgement rather than to finding sources. The sign-off stays with them.
A draft document or a published URL. Claims come back with their place in the text, so edits land exactly where the claim sits.
Minutes for a single page. Chains that have already been walked for another claim resolve immediately, so checks get faster as the library grows.