Skip to content
Multilo Docs

Claim Check

On-device citation verification: it reads every citation, fetches the source, and tells you whether the claim is actually supported.

Claim Check is citation verification. It reads every citation in your draft, fetches the cited source, and tells you whether the source actually supports the claim you attached to it. It’s the agent that turns “sounds right” into “checked.”

What it does

For each citation, Claim Check compares the claim in your text against the real content of the source and flags three cases: the source supports the claim, the source is weak or unrelated, or a stronger source exists in your library. It catches the failure mode that matters most in academic work — a confident sentence backed by a citation that doesn’t say what you claimed.

How to run it

  1. Run it on your draft

    Start Claim Check from the document. It walks through the citations in your text.
  2. Review the findings

    For each citation you see whether the source supports the claim, with the relevant evidence surfaced so you can judge for yourself.
  3. Fix in place

    Accept a suggested replacement or adjust the claim — the change is applied directly to your document, citations and all.

Accepting a better source

When a citation is weak, Claim Check can propose a stronger source from your library. Accept it and the citation is swapped cleanly — the in-text marker and the reference update together, so your bibliography stays consistent. (In the rich document view, citation edits are applied through the editor itself, not by hand-editing text, so nothing drifts out of sync.)

Run it before you submit

Make Claim Check the last pass on any draft — your own or one written by the Full Draft Writer. It’s the cheapest insurance against an unsupported citation slipping through.

On-device and honest

Claim Check is designed to be faithful: it reports what the source actually says and admits when it can’t verify a claim rather than rubber-stamping it. This honesty-first behavior is part of Multilo’s research-integrity approach — the AI’s job is to be trustworthy, not just confident.