How to check your thesis references: from citations to DOIs

The reference list is where a committee or reviewer most quickly judges how careful a thesis is. A missing source, a wrong year or a publication that does not exist erodes trust in the rest of the text. These errors are also more common than people think: a review of medical journals found that 16.9% of citations were not supported by the source they cited (Baethge & Jergas, 2025). The steps below are ordered to catch these errors before you submit.

1. Match citations to the reference list

Start with a two-way match: every citation in the text needs an entry in the reference list, and every entry needs at least one citation in the text. In an author-date system you can list all citations by searching for years in parentheses; in a numbered system, search for square brackets. Then tick them off in the reference list one by one.

The most common errors this match turns up:

• A different year in the text and in the list (for example, mixing up a preprint year and the year of print publication).
• An author’s name spelled differently in the text and in the list, especially names with non-English characters.
• Sources left over from earlier drafts: removed from the text but still in the list.
• Citations added to the text later that never made it into the list.
• Two papers by the same author in the same year (2020a, 2020b) mixed up.

2. Verify the DOIs

Open each DOI in your list through doi.org and check that the title, authors and year of the publication it leads to match your entry. A single wrong digit can send a DOI to an entirely different paper, and you will not notice on your own, because the link still works. For journal articles without a DOI, search by title on Crossref’s search page. Not every publication has a DOI; some books, theses and journals do not use them. In that case, confirm that the source exists through a library catalogue or the publisher’s website.

3. Weed out retracted papers

A paper can be retracted after publication because of errors, data problems or ethical violations. Relying on a retracted paper as evidence is one of the first things a committee or reviewer will raise. There are a few ways to check:

• The Retraction Watch database: one of the most comprehensive lists of retracted publications, openly available through Crossref.
• The publisher’s page: a retracted article’s page usually carries a retraction notice. Look for corrections (errata) and expressions of concern too.
• Your reference manager: some reference managers, such as Zotero, flag retracted items in your library.

If you need to mention a retracted paper on purpose (for example, because you discuss the controversy around it), say clearly in the text that it has been retracted.

4. Does the source actually say that?

A perfectly formatted reference list does not mean the citations are accurate. The hardest and most important check is whether each claim you cite is really in the source. Common problems:

• Citations based only on the abstract that do not match the content of the paper.
• Citations copied from another study’s reference, where the original was never opened.
• Generalising something the source said in a narrow context.
• Missing or wrong page numbers for direct quotations.

Rereading every cited source takes time, so start with the central claims your argument depends on. For how the literature should connect to your argument, see the guide on how to write a literature review.

5. Make the citation style consistent

APA, Chicago, IEEE or another style: which one you use is decided by your graduate school’s thesis guide or the journal’s author guidelines, so follow your own institution’s current version. Whatever the style, the whole list must follow the same rules. Author name formats, italics, whether journal names are abbreviated, how DOIs are written, the use of "et al." and the ordering of entries are where inconsistencies creep in most often. Reference managers such as Zotero, Mendeley or EndNote make this easier, but imported metadata is often incomplete or wrong, so always check the output by eye.

6. Watch out for AI-invented citations

Chat-based AI tools can produce sources that look real but do not exist. In one experiment, 19.9% of citations generated by GPT-4o pointed to publications that do not exist (Linardon et al., 2025). These sources are often convincing: the title fits the topic, the authors really work in the field and the journal is real; only the paper itself is missing. Or the DOI exists but leads to a different publication.

The rule is simple: never cite a source you have not opened and read. Find every source an AI tool suggests in Crossref, Google Scholar or a library catalogue and read the text yourself. Rules on using AI in thesis writing differ between universities, so check your institution’s policy as well; more in the guide to AI for thesis writing.

Quick checklist

• Does every citation have an entry, and every entry a citation?
• Are author names and years the same in the text and the list?
• Does every DOI lead to the right publication?
• Are you relying on any retracted paper?
• Have you read the sources behind your central claims yourself?
• Is the style consistent throughout and in line with your institution’s guide?
• Have you confirmed that every AI-suggested source exists?

Checking references is one step of checking the whole thesis; for the other steps, see how to check your thesis.

tezkeys does the mechanical part of this list automatically: it compares your reference list with Crossref records and shows citations missing from the list, entries never cited in the text, retracted publications and DOIs that point to another publication. This pre-check is free and needs no sign-up. Reading a source to see whether it really supports your claim is still up to you.