Paste your reference list and this reference checker looks each citation up in Crossref and PubMed to confirm a matching record exists. It flags references that cannot be located, mismatched years and authors, and papers carrying retraction or correction notices. Works with any citation style, numbered lists, and BibTeX.
Any citation style works. Numbered prefixes like "1." or "[1]" are stripped automatically. Each reference is parsed for its digital object identifier (DOI), title, first author, and year, then checked against Crossref and PubMed directly from your browser; your list is never sent to our server.
The strongest identity check for a journal article is its digital object identifier (DOI). When a reference includes one, the checker queries the Crossref registry directly: a registered DOI returns the authoritative record, and the tool then compares the cited title, first author, and year against that record so a wrong or recycled identifier does not pass as verified. Used this way it doubles as a DOI checker: it tells you not only whether the identifier resolves, but whether it resolves to the paper you actually cited. Crossref also attaches update notices to its records, sourced in part from the Retraction Watch database, so retractions, expressions of concern, and corrections surface in the same pass.
References without an identifier go through a bibliographic search instead. The checker extracts the probable title, first author surname, and year from the raw citation text, queries Crossref's bibliographic search, and scores the top candidates on title word overlap, year agreement, and author match. Strong matches are reported as verified with the matched record linked; borderline matches are reported as possible matches for you to review. Anything Crossref cannot place is then searched in PubMed by exact title through the National Library of Medicine's E-utilities, which catches biomedical articles with sparse Crossref metadata. Because matching happens on metadata rather than punctuation, the tool works as a citation checker for any referencing style; if you need to build or reformat the list itself, the citation generator handles formatting, and the reference deduplication tool cleans up duplicates before you check.
Next step
End-to-end systematic review, meta-analysis, or biostatistics support, handled by a PhD methodologist.
Our promise: Free rework on search, screening, or synthesis if reviewers push back.
Will your review include a meta-analysis? We run the systematic review and the pooled analysis together, in one project.
Quote my review + meta-analysisTimeline
Most projects deliver in under 2 weeks. We confirm an exact date in your quote.
If reviewers push back
If reviewers question the search, screening, or synthesis, we rework the section free.
Confidentiality
NDA available on request before any project discussion. Your data, study design, and manuscript stay private either way.
Want a PhD methodologist to handle the whole project?
Get a complete systematic review or meta-analysis handled end-to-end. Free rework on search, screening, or synthesis if reviewers push back. Pay only after you approve your quote.
Large language models produce references the same way they produce every other sentence: by predicting plausible text. A fabricated citation therefore looks exactly like a real one, with a genuine journal name, authors who publish in that field, a sensible year, and an identifier formatted like a valid DOI. Peer-reviewed audits of chatbot-generated bibliographies have repeatedly found that a substantial share of the references they produce either do not exist or point to a different paper than claimed, and academic librarians now regularly field requests for articles that were never published. The failure is invisible to proofreading because nothing about the citation looks wrong on the page.
The consequences land on the author, not the tool. Journals and universities treat citing nonexistent sources as a serious integrity problem regardless of whether an assistant produced them, and several widely reported legal and academic cases have turned on exactly this. The practical response is not to avoid writing assistance but to verify every reference against an authoritative index before submission, which takes seconds per entry here instead of a manual search each. The same check protects against older, fully human failure modes too: citations copied from another paper's reference list, transposed digits in an identifier, and papers that were quietly retracted after you first read them.
Every reference receives an explicit verdict, and each verdict is deliberately conservative about what it claims:
The checker covers what Crossref and PubMed index, which is most of the scholarly journal literature but not all of the citable world. Books, book chapters, dissertations, conference papers without registered identifiers, standards, government and organizational reports, web pages, and older or regional journals may all be perfectly real and still show as could not be located. Treat that verdict as a to-verify list, not an error list. The reverse limitation also applies: a verified match confirms the record exists and the metadata agrees, and the retraction pass is recall-oriented rather than exhaustive, since Crossref update notices can lag or be missing for some publishers. No automated pass replaces reading the source, and the final decision on every reference stays with you.
Title matching is also heuristic by nature. Citations with heavily abbreviated journal-style titles, very short titles, or non-Latin scripts match less reliably and are more likely to land in the possible-match band. For those, the matched record link is the fastest way to settle it. If the reference list came out of a screening or review workflow, the systematic review toolkit includes the documentation templates for recording these verification decisions, and the PRISMA flow diagram generator accounts for records you exclude along the way.
Look each reference up in an authoritative bibliographic index rather than trusting the formatting. If the reference has a digital object identifier (DOI), resolve it at doi.org or query the Crossref API; the record it returns should match the cited title, authors, and year. If there is no DOI, search the exact title in Crossref, PubMed, or Google Scholar. A reference whose identifier resolves to a different paper, or whose title returns no record anywhere, needs to be verified against the original source before you rely on it. This tool automates those lookups for a whole reference list at once.
Verify a reference in three steps: first resolve its digital object identifier, if present, and confirm the returned record matches the citation; second, search the title in a bibliographic database such as Crossref or PubMed and compare authors and year; third, open the source itself and confirm it actually supports the claim it is cited for. The first two steps are mechanical and can be automated, which is what this checker does. The third step always needs a human reader, because a real paper can still be cited for something it never says.
Correctness goes beyond existence: the authors, year, journal, volume, and pages in your citation must match the published record. Paste your list into the checker and review the metadata mismatch verdicts, which show the cited year or author alongside what the matched Crossref or PubMed record says. Mismatches most often come from citing the online-first year instead of the print year, misspelled author names, or copying a citation from another paper's reference list without checking the original.
Paste your full reference list, one entry per line or as BibTeX, and run the check. Each entry is parsed for its digital object identifier, title, first author, and year, then matched against Crossref and PubMed. Verified entries link to the matched record so you can confirm it is the work you meant. Anything marked as could not be located should be checked by hand, especially books, chapters, dissertations, and reports, which these indexes cover less completely than journal articles.
A fabricated citation usually looks perfectly formatted, which is exactly why formatting is a poor test. The reliable signals are behavioral: the digital object identifier does not resolve, or resolves to an unrelated paper; the exact title returns nothing in Crossref, PubMed, or Google Scholar; the author has no other work in that field; or the journal, volume, and page combination does not exist. No single failed lookup proves fabrication, since real books and gray literature can also be missing from indexes, so treat every failed lookup as a prompt to find the original source.
Once every reference is verified, the citation and reference list builder formats them consistently in Vancouver, APA, AMA, Harvard, MLA, or Chicago. If your list was merged from several database exports, run the duplicate citation finder first so you only verify each work once. And when the references feed a review manuscript, the literature review outline generator structures the synthesis those citations support.
Reviewed by
Dr. Sarah Mitchell holds a PhD in Biostatistics from Johns Hopkins Bloomberg School of Public Health and has over 15 years of experience in systematic review methodology and meta-analysis. She has authored or co-authored 40+ peer-reviewed publications in journals including the Journal of Clinical Epidemiology, BMC Medical Research Methodology, and Research Synthesis Methods. A former Cochrane Review Group statistician and current editorial board member of Systematic Reviews, Dr. Mitchell has supervised 200+ evidence synthesis projects across clinical medicine, public health, and social sciences. She reviews all Research Gold tools to ensure statistical accuracy and compliance with Cochrane Handbook and PRISMA 2020 standards.
Our PhD methodologists rerun the statistics, add the analyses reviewers requested, fix the reporting, and draft the technical point-by-point responses. You submit the revision with every box ticked.
Our promise: Free re-run and re-write if reviewers question the analysis or reporting.