How CiteProve works
In short
CiteProve verifies a citation in four steps: it extracts any identifier (DOI, PMID, arXiv ID, ISBN, or legal reporter), searches Crossref, OpenAlex, PubMed and CourtListener in parallel, scores each candidate field by field (title, authors, year, venue, volume, pages), and returns a careful verdict — verified, partial, not found, or can't check — with every source consulted shown.
1. We split and clean your bibliography
Paste anything — a numbered reference list, a hanging-indent bibliography, or a single citation. We segment it into individual references and normalize the typography (smart quotes, ligatures, dashes) that often breaks automated matching.
2. Identifiers get checked first
If a reference carries a DOI, PMID, arXiv ID, ISBN, or a US legal reporter citation, we look it up directly in the authoritative registry. This is the fastest and most reliable path — but we still validate the rest of the citation against what the identifier resolves to. A real DOI attached to the wrong paper is one of the most common fabrication patterns we catch.
3. Search across authoritative databases
Every reference is searched in parallel against Crossref (160M+ scholarly works), OpenAlex, PubMed (for biomedical citations), and CourtListener (for US case law). Each database returns its best candidates.
4. Field-level validation
Candidates are scored field by field: title, author surnames, year (±1 tolerance), journal or venue, volume, and pages. The verdict depends on what actually matches — not just whether something similar exists. When fields differ, you see exactly which ones, with a one-click corrected citation.
5. Careful verdicts
We flag, we never accuse. Verified means a high-confidence match. Partial means we found the source but some fields differ. Not found means we could not locate it in any of the databases we consulted — with the list of databases shown, so you can judge the coverage yourself. Every verdict links to the sources it was checked against.