Citation Verify
Confirms that the references a document cites actually exist and are correctly identified, by looking each one up in the major scholarly databases and matching the cited details against the record. Also flags any cited paper that is retracted.
Technical description
Verifies a manuscript's references for existence and identity. When a bibliography is present, each reference is parsed and checked against CrossRef, OpenAlex, PubMed, and Semantic Scholar, and the cited title, authors, and year are matched against the record found. This separates a reference that merely exists from one that is correctly identified, so a citation whose identifier resolves to a different paper, or whose details do not match, is caught rather than passed. Located references are additionally checked against the research-integrity database for retractions. When no bibliography is found, it falls back to verifying identifiers and author-year citations in the text.
How it works
Layer 3 (external lookup): Separates and parses the bibliography into individual references. Searches CrossRef, OpenAlex, PubMed, and Semantic Scholar. Accepts a record only when its title, authors, and year match the cited reference closely enough, distinguishing found-and-matched from found-but-mismatched from not-found. Labels each problem reference by failure type (not found, identifier points to a different paper, details partly wrong, or matches nothing closely). Checks located references against the research-integrity database for retractions. Falls back to in-text identifier and author-year verification when no bibliography is present.
Why this matters
A fabricated citation is invisible on the page: it has plausible authors, a real-sounding journal, and a well-formed identifier, and only a lookup reveals the problem. The most deceptive case is the reference that resolves to the wrong paper: a valid identifier pointing to a real but unrelated work, which a plain existence check waves through. Matching the cited details against the record catches these. The same lookup also surfaces references to papers that have since been retracted, which should not be cited as valid.
Score thresholds
- 0-1
- Every checked reference was found and matched, with no retractions
- 2-3
- A minority could not be confirmed, did not match the record, or were retracted
- 4-5
- A large share could not be found or were misidentified
Limitations
Depends on external services, so it is slower than the Layer 1 checks and can be affected by downtime or rate limits; it verifies only the first several references per document. A genuine reference that is very new, in a niche venue, or indexed differently can fail to match and be reported as unconfirmed, so an unmatched reference is a prompt to check by hand rather than proof of fabrication. The retraction check reflects what is recorded at the time of lookup.
References
- Ansari MS. (2026). Compound deception in elite peer review: a failure mode taxonomy of 100 fabricated citations at NeurIPS 2025. arXiv preprint arXiv:2602.05930
- Abbonato D. (2026). CheckIfExist: detecting citation hallucinations in the era of AI-generated content. arXiv preprint arXiv:2602.15871
- Li M, Lin Z, Ma S. (2026). Source or it didn't happen: a multi-agent framework for citation hallucination detection. arXiv preprint arXiv:2605.08583
- Russinovich M. (2025). RefChecker: a tool that validates academic paper references. open-source software
- The Center for Scientific Integrity. (2018). The Retraction Watch Database. research-integrity database