Why references go wrong#
A reference can fail in several ways, and only some of them are obvious:
- It does not exist. Generative AI tools can produce references that look plausible (real-sounding authors, real journal names, a well-formed DOI) but match no published work.
- The details are wrong. A wrong year, misspelled author, or wrong volume often comes from copying a reference from another paper instead of from the source.
- It exists but has been retracted or corrected. Retracted papers continue to be cited long after the retraction.
- It does not support the claim. The paper is real, but it does not say what the citing sentence says it does, or says something narrower.
The last problem is common even among professionals. Jergas and Baethge's 2015 systematic review and meta-analysis of quotation accuracy in medical journal articles (PeerJ, 3, e1364) found a pooled quotation error rate of 25.4%. That is a reason to check claims, not just bibliographic details.
Step 1: Resolve the DOI#
A DOI (Digital Object Identifier) is a persistent identifier that begins with "10." followed by a prefix, a slash and a suffix, for example 10.7717/peerj.1364. Most journal articles published in the last two decades have one.
- Paste the DOI after https://doi.org/ in your browser, or into the search box at doi.org.
- If it resolves, you will land on the publisher's page for the item. If you get a "DOI not found" error, check for typos (a missing character or a stray full stop at the end is common), then treat the reference as unverified.
- Compare the landing page with your reference: title, all authors in order, year, journal, volume, issue and pages.
Step 2: Look it up in Crossref, PubMed or a catalogue#
If there is no DOI, or you want a second check, search for the title in a bibliographic database.
Crossref#
Crossref is the largest DOI registration agency for scholarly publications. Its free search at search.crossref.org lets you paste a full reference and returns the closest matching records with their DOIs and metadata. If nothing close comes back, the reference may be wrong or the item may not be registered with Crossref (some books, reports and older articles are not).
PubMed#
For biomedical and life-science references, search PubMed by title (in quotation marks) or by author and year. A match gives you a PMID and the citation as indexed. PubMed also labels retracted publications and links to the retraction notice.
Books, reports and preprints#
- Books: search your library catalogue or WorldCat by title or ISBN, and check the edition and publisher.
- Reports and grey literature: find the document on the issuing organisation's own website.
- Preprints (arXiv, bioRxiv, SSRN and others): check whether a peer-reviewed version has since been published, and cite that if your course or journal prefers it.
- Google Scholar: useful for finding a title quickly, but its metadata is sometimes wrong. Confirm details at the publisher's page.
Step 3: Check for retractions and corrections#
A paper can be real and correctly cited but no longer stand. Check:
- The publisher's page. Retracted articles are normally marked as retracted, often with a watermark on the PDF and a link to the notice. Look also for corrections (errata) and expressions of concern.
- The Retraction Watch database. A large, searchable database of retractions across publishers. Since 2023 it has been owned by Crossref and made freely available.
- PubMed, for biomedical papers, which flags retracted publications.
If a source has been retracted, do not cite it as evidence. You may still cite it when the retraction itself is the subject, and then you should say that it has been retracted.
Step 4: Verify the quote or claim#
This is the check that catches the most serious errors. For each place you cite a source:
- Open the full text, not just the abstract.
- Find the passage that supports your sentence. For a direct quotation, check every word and the page number.
- Ask whether the source supports your claim as you have worded it. Watch for overstatement ("proves" for "suggests"), dropped qualifiers (a result in mice presented as a result in people), and claims the paper only mentions while citing someone else.
- If you are citing a claim that the paper attributes to another study, find and read that original study, or cite it as a secondary source following your style's rules.
Step 5: Format it correctly#
What to do when a reference fails a check#
- It does not exist. Remove it. Do not replace it with a similar-sounding real paper unless you have read that paper and it supports your sentence.
- The details are wrong. Correct them from the publisher's record, and check that the content still fits your claim.
- It has been retracted. Remove it as evidence, and look for other work that supports the point, or rethink the point.
- It does not support the claim. Reword your sentence to match what the source says, find a source that does support it, or cut the claim.
- You cannot get the full text. Ask your library for an interlibrary loan or contact the author. Citing a paper from its abstract alone is risky, because abstracts leave out the qualifications.
If you only read a study as described in another paper, cite it as a secondary source using your style's format (APA, for example, uses "as cited in"), and cite only what you read.
Red flags and a quick routine#
Be especially careful with a reference when:
- it came from an AI chatbot, a search snippet or another paper's reference list rather than from the source,
- the title is suspiciously well matched to your exact argument,
- the journal name is slightly off from a real journal's name,
- the DOI prefix does not belong to the journal's publisher, or the page numbers do not fit the volume,
- you cannot find the full text anywhere, including through your library.
| Check | How | Pass if |
|---|---|---|
| Exists | Resolve the DOI, or search Crossref, PubMed or a catalogue | A matching record is found |
| Details | Compare with the publisher's page | Authors, year, title, journal, volume, pages match |
| Status | Publisher page, Retraction Watch database, PubMed | No retraction; corrections noted |
| Claim | Read the cited passage in the full text | It supports your sentence as worded |
For a longer project, build these checks into your workflow as you collect sources rather than leaving them to the end. The literature review checklist includes them, and our guide to writing a literature review shows where they fit.
KEEP GOING
https://moremind.ai/en/guides/how-to-check-a-citation-is-real