Reference Checker
Paste your bibliography and find out which citations are real. Every reference is looked up against Crossref and Retraction Watch, so you see which ones are retracted, which sit in a hijacked journal, and which cannot be found at all. That last group matters more than it used to: about 1 in 277 papers published this year carries a reference that does not exist, and an AI writing assistant is usually why.
We don’t train AI on your data. Bibliography text uses short-lived result caching.
Paste any format
DOIs (one per line), BibTeX, RIS, or plain-text references · auto-detected.
We resolve every entry
Crossref polite pool fetches metadata; Retraction Watch fast-path catches retractions.
Per-reference cards
A verdict per reference (real / retracted / hijacked / not found anywhere) with the evidence behind it.
Why reference integrity matters
The Retraction Watch database tracks over 50,000 retracted papers. Studies sampling published manuscripts find that roughly 1 in 200 references in active citation flow is a paper that has been retracted, and a paper continues to be cited an average of ~10 more times after retraction. Reviewers and journal staff increasingly screen reference lists against these databases at submission; a retracted citation discovered after publication can trigger a correction or retraction of your own paper.
This runs the same checks reviewers do, in the pre-submission window, so a fixable citation never becomes a downstream correction notice.
Limitations
- A DOI gives the strongest answer. References with a DOI can be confirmed or ruled out against the registry. For a reference without one we search Crossref by title, author, and year, and accept a match only when it is unambiguous. A reference we still cannot find is reported as unchecked, never as invented: books, theses, conference papers, and older or non-English work are frequently real but simply not indexed.
- Crossref-registered DOIs. Some publishers use non-Crossref registrars (e.g., DataCite for some preprints, mEDRA, KISTI). DOIs from those registrars surface as “DOI did not resolve” · not as a clean pass and not as a flag.
- Retraction propagation lag. Retraction notices typically appear in Crossref within 24-72 hours of publication. Very fresh retractions may not be caught yet; we mirror Retraction Watch’s curated list nightly to fast-path known cases regardless of Crossref ingest time.
- Hijacked-journal name match is conservative. ISSN match is reliable; pure name match runs against the Retraction Watch Hijacked Journal Checker tracker. We do not flag based on title-similarity heuristics, only exact matches.
- 200-reference cap. V1 processes the first 200 references per check. For full-bibliography sweeps on book-length manuscripts, run multiple checks or use the manuscript-level readiness scan.
Manuscript-level read
Want every claim verified, not just every reference?
This tool checks the references themselves. The full Manusights Readiness Scan reads your entire manuscript and verifies whether each citation actually supports the claim it’s attached to · alongside methods, statistics, journal fit, and reviewer-flag patterns. Free preview, $39 only if you want the full report.
Calibrated from early work with 35+ CNS-experienced reviewers and senior scientists. Their reviewer language shaped how Manusights flags unsupported claims, weak evidence, journal-fit risk, and likely reviewer objections. Logos indicate researcher affiliation, publication context, or calibration-network context, not endorsement.
Want the data sources behind every flag? Read the full methodology · Crossref polite-pool details, Retraction Watch and Hijacked Journal Checker data sources, update cadence, and known limitations. Or read About + credits for attribution to the original data providers.
How common is the problem? We checked 12,402 recent papers across 54 major journals: 7.1% cite a retracted, EoC, or corrected reference (1.4% retracted-or-EoC). Method + named examples in the data study. We also measured the sharper timing question: citations to work that was already retracted when the citing paper appeared.
Checking citation claims, not just references? Run Manusights Verify · paste a sentence with a citation, get a verdict on whether the cited paper actually supports the claim.