FREE TOOL · NO SIGN-UP
Free citation checker
Paste a reference list. Every entry is checked against Crossref and OpenAlex: references that do not exist, wrong years or DOIs, and retracted papers are flagged.
Free: 5 runs a day per IP address, shared across the free tools. No account. Up to 10 references per run, one per line.
Input
Results
Press Run to see live results here. See a saved sample result below
How the citation checker works
- 1
Paste your references
Put each reference on its own line, in the style you already use. The first 10 lines are checked. A DOI, PMID or arXiv ID in the line is used when it is there.
- 2
Each one is looked up
The tool reads the title, authors, year, journal and pages from each line with fixed rules, then looks the paper up in Crossref and OpenAlex: by DOI first, otherwise by title. No language model is involved.
- 3
Read the verdicts
Every reference gets one status: Verified, Needs correction (with the cited value and the value on record), Not found or Uncertain. A paper that has been retracted also carries a Retracted label and the notice.
Sample result from the citation checker
Saved from a real run of this tool on 13 October 2026, using the example input that is pre-filled above. The lookup took 1.9 seconds in that run. Live results can differ as the databases change.
Example input
LeCun Y, Bengio Y, Hinton G. Deep learning. Nature. 2015;521:436-444. Jumper J, et al. Highly accurate protein structure prediction with AlphaFold. Nature. 2019;596:583-589. Wakefield AJ, et al. Ileal-lymphoid-nodular hyperplasia, non-specific colitis, and pervasive developmental disorder in children. Lancet. 1998;351:637-641. Zhang Q, Okafor P. Quantum lattice memory in songbirds. Cell. 2022;185:1-12.
The fourth reference is one we made up for the example, and the year in the second was changed on purpose. The other details are as the tool returned them.
Saved result
4 references checked: 2 verified, 1 need corrections, 1 not found, 0 uncertain; 1 retracted.
Jumper J, et al. Highly accurate protein structure prediction with AlphaFold. Nature. 2019;596:583-589.
Highly accurate protein structure prediction with AlphaFoldyear: 2019 → 2021
Wakefield AJ, et al. Ileal-lymphoid-nodular hyperplasia, non-specific colitis, and pervasive developmental disorder in children. Lancet. 1998;351:637-641.
Ileal-lymphoid-nodular hyperplasia, non-specific colitis, and pervasive developmental disorder in childrenRetracted on 2010-02-06 (notice: https://doi.org/10.1016/s0140-6736(10)60175-4). Cite it only to discuss the retraction.
Zhang Q, Okafor P. Quantum lattice memory in songbirds. Cell. 2022;185:1-12.
No record with this title in Crossref or OpenAlex.
How fake references from AI chatbots are detected
A chatbot can write a reference that looks right: real-sounding authors, a real journal, a plausible title, sometimes a DOI. This tool does not judge how the text reads. It looks for the record, and an invented paper has none.
It catches three shapes of invented reference:
- A title with no record in Crossref or OpenAlex comes back as Not found.
- A real title cited with a first author who is not among the paper’s authors comes back as Uncertain, with the real authors listed.
- A DOI that resolves to a different paper is reported as such, and the tool then searches by title instead.
What “Not found” does and does not mean
Not found means that neither Crossref nor OpenAlex has a record with this title. For a recent journal article, that is a strong sign the reference is wrong or invented, and it is worth searching for it by hand before you keep it.
It is not proof. Books and book chapters are not always indexed in these two databases, so a reference that looks like a book is marked Uncertain instead, with a note to check a library catalogue. Theses, reports and other works that have no record in either database will also come back unmatched. A title that is close to a real one but not the same is marked Uncertain and shown next to the closest record.
Verified has limits too. It means the paper exists and the details you cited match the record. It does not mean the paper says what your sentence claims.
Which databases are checked
Two: Crossref, the registry behind most journal DOIs, and OpenAlex, an open catalogue of scholarly works. PubMed is not queried. If a reference carries a PMID or an arXiv ID, that identifier is resolved through OpenAlex.
The Retracted label comes from the same two sources: Crossref metadata, which includes Retraction Watch data, and the retraction flag in OpenAlex. Google Scholar, Scopus and Web of Science are not checked.
What “Needs correction” compares
When a paper is found, six fields from your line are compared with the record: first author, year, journal, volume, first page and DOI. Each mismatch is listed as the value you cited and the value on record, as in the sample above, where 2019 should be 2021.
Some differences are left alone on purpose. A book cited with another edition’s year gets a note, not a correction. Conference names are cited in too many forms to compare, and a preprint has no journal or pages to check.
FAQ
Related tools
- Free retraction checker
Check a list of DOIs for retractions, corrections and expressions of concern.
- Free journal checker
Look up a journal’s open-access status, DOAJ listing and APC.
- Reference verification for AI agents
The same check as a tool your agent calls while it writes.
Data sources behind this tool: Crossref data source, OpenAlex data source. All ScholarMuse web tools
The same check, inside Claude or ChatGPT
Connect ScholarMuse once and your agent can check up to 50 references per call while it writes, without you pasting lists back and forth. Calls from an agent need an account and use credits.