Catch errors, hallucinated references, and missing metadata before you submit.
Upload a .bib file or paste directly from Overleaf.
BibTeX Verifier checks every entry against
CrossRef and
Semantic Scholar to find
wrong authors, incorrect years, mismatched venues, missing DOIs, duplicates,
and AI-hallucinated citations that don't exist in any database.
100% in-browser — your data never leaves your machine.
Standard BibTeX & BibLaTeX · Processed entirely in your browser
Drop your .bib file or paste directly from Overleaf. Everything is parsed in your browser — nothing is uploaded to any server.
Each entry is looked up by title against Semantic Scholar and CrossRef. Authors, year, venue, DOI, and other fields are compared to the published record. Entries that don't match any real publication are flagged — perfect for catching AI-hallucinated citations.
Results are color-coded: Verified means everything matches, Auto-Updated shows a diff you can accept or edit inline, and Needs Review flags possible errors. A live preview shows the corrected BibTeX in real time.
Download or copy the corrected .bib file with your changes applied. Ready for LaTeX, Overleaf, or any reference manager.
Flags entries whose titles don't match any known publication — catches fabricated references from AI-generated bibliographies.
Queries both Semantic Scholar and CrossRef to maximize coverage and metadata quality.
See exactly which fields differ — authors, year, venue, DOI, pages — with inline editing and a live BibTeX preview.
Your file never leaves the browser. Only paper titles are sent to public academic APIs for lookup. No server, no tracking.
Fills in missing DOIs, publisher info, volume, pages, and other metadata from the online record automatically.
Fuzzy title matching, LaTeX-aware text processing, and last-name-based author comparison handle real-world formatting differences.
Yes. When a BibTeX entry's title doesn't match any real publication in CrossRef or Semantic Scholar, it's flagged as "Not Found" or "Needs Review." This is how fabricated citations from ChatGPT, Copilot, or other AI tools get caught — the paper simply doesn't exist in any academic database.
Yes. Your .bib file is parsed entirely in your browser using JavaScript. The only network requests made are title lookups to the public CrossRef and Semantic Scholar APIs. No file content is sent to any server.
Not every paper is indexed in CrossRef or Semantic Scholar. ArXiv preprints, workshop papers, theses, and very recent publications may not be found. The tool flags these so you can verify them manually — or remove them if they're fabricated.
To respect the free APIs' rate limits, requests are sent at a controlled pace with adaptive throttling. For a file with 50 entries, expect roughly 2–3 minutes. You can review results and change settings while it's still processing.
Yes! You can copy your entire .bib file from Overleaf and paste it directly into the "Paste BibTeX" tab. After reviewing, download the corrected file and upload it back to Overleaf.
Yes! The full source code is on GitHub under the MIT license. Contributions are welcome.