FOR ACADEMIC RESEARCH

The manuscript arrived polished. Verify its citations before your name goes on the review.

Editors, reviewers, committees, and research offices act on documents other people wrote: manuscripts, theses, grant reports, literature reviews. AI now drafts and summarizes them at every stage, and a fabricated reference or a citation that does not say what the text claims reads exactly like a real one.

The errors hide in the apparatus the work stands on: a DOI that does not resolve, a reference whose metadata does not match the record, a cited paper that does not support the claim attributed to it, a statistical assertion the numbers in the paper contradict. Asking another AI to check the document is not a fix. In our pre-registered study of twenty company profiles, three frontier checkers disagreed by as much as 9.6 to 1 on the same document, and the harshest one returned zero errors on documents containing sixteen verifiable ones.

The AI is the witness. The math is the judge. GauntletScore verifies each claim against primary sources, then computes the trust score with deterministic math. The agents gather the evidence; no model sits in the verdict. The same evidence always produces the same score.

What GauntletScore checks
Citations and references against CrossRef and PubMed, so every cited work resolves to a real publication.
DOIs and reference metadata against the registry record, so the authors, journal, and year in the reference list match the publication they point to.
Statistical claims with review of the reported figures, so the numbers the argument depends on are internally consistent.
The argument's causal structure because a paper is a chain of reasoning, and the causal pass tests whether the chain holds.

GauntletScore runs a dedicated analysis of cause-and-effect claims, testing temporal order, proportionality, and logical structure. A conclusion whose causal chain fails the test counts heavily against the document. Sound reasoning counts only weakly for it, because internal consistency is not external proof.

What the first run looks like

Upload the document at gauntletscore.com. Minutes later you are reading the Gauntlet Report: a trust score with a 95% interval, every flagged claim with its verdict and the primary source behind it, and a cryptographically signed, tamper-evident certificate. The fabricated reference is caught before the review, the acceptance, or the citation carries it forward. In a live production run, the engine examined nineteen claims in one fluent, credible document and returned two debunked against the court record, each with the source that contradicts it. Every voting agent independently recommended against proceeding.

LIMITS

It does not judge the novelty or significance of the work and it does not verify what no public source can confirm, including unpublished data. A claim it cannot check is reported as unverifiable, not as false.

Where it fits
Editorial screening.

Run the submission's checkable apparatus before it goes out for review.

Review support.

Verify the references and statistical claims behind the argument you are being asked to endorse.

Pre-submission check.

Run your own manuscript before it carries your name to a journal.

There is a manuscript in your queue right now whose references nobody has resolved. Run it. Either they hold, or one does not, and you found it before it published.

Analyze a Document
Three free credits. No card. No call.

Validation study pre-registered on the Open Science Framework, March 2026; in progress, manuscript in preparation.