Reference checking: a one-page checklist

1. Does the work exist?

1. Resolve the DOI when the citation has one: paste it after https://doi.org/. "DOI Not Found" on an otherwise polished citation is the fabrication pattern. A DOI that opens a different paper means the citation was assembled from parts.
2. Search the exact title in quotation marks in the library catalogue, then in CrossRef or OpenAlex. Quotation marks matter: without them the search helpfully finds something similar, which is the failure you are trying to detect.
3. Check the author's own publication list or ORCID record. Invented citations often attach a real, prolific author to a paper they never wrote.
4. Check the container. Confirm the journal exists, that the volume matches the year, and that the page range falls inside that issue.

2. Match the record against what you wrote

A reference can be real and still wrong. Compare year, author order, spelling of names, volume, issue, and page range against the record you just opened. Most amber results trace to a typo picked up from a reference manager.

3. Has it been retracted?

1. Open the publisher's page. A retracted article usually shows "RETRACTED" in the title and a banner linking the notice, though not every publisher does this.
2. Query https://api.crossref.org/works/<DOI> and read the updated-by field: an entry of type retraction gives the notice DOI and its date. CrossRef has distributed Retraction Watch data since 2023.
3. Note the near neighbours. An expression of concern means the journal has published doubts without withdrawing the paper. A correction can be trivial or can change the result you relied on.

4. Know where the registries stop

CrossRef and OpenAlex index journal articles, books, and much conference material. They index almost none of the following, so absence there is not evidence a work is fake:

• Consultancy and think-tank reports (some agencies register DOIs, many do not)
• Newspapers, magazines, blogs, and press releases
• Government documents, standards, and datasets
• Dissertations and most grey literature
• Non-English scholarship, unevenly

For anything in this list, go to the publisher or issuing body directly, or ask a subject librarian, whose licensed databases index material the open registries leave out.

5. When a reference will not resolve

Remove it, and check whether the claim it supported survives without it. If the claim rests on that source alone, the claim goes too. Citing a retracted work is legitimate only to discuss the retraction, and then the text should say so.