

Reference List Checklist: 9 Checks Before You Submit
Before you submit, run nine checks on your reference list, starting with these three. Every in-text citation has an entry, and every entry is cited (American Psychological Association [APA], 2022). Each entry agrees, field by field, with a bibliographic source such as PubMed or with the original (International Committee of Medical Journal Editors [ICMJE], 2026). Every DOI opens the paper the entry describes. For a revision, use the revised manuscript checklist instead.
- Match citations and entries both ways.
- Check each entry against the published record.
- Open every DOI.
- Look for retractions.
- Read each source you cite.
- Label preprints, in-press papers, and unpublished work.
- Read the journal's guide for authors.
- Check author lists and issue numbers.
- Verify every reference an AI tool suggested.
Examples were checked against Crossref, PubMed, or doi.org on November 7, 2026.
1. Is every citation in the list, and every entry cited?
APA requires an entry for every work cited and a citation for every entry, which can sit in the text, a table, a figure, a footnote, or an appendix; author spellings and years must match (APA, 2022). In an author-date style like APA, search the manuscript for each entry's first-author surname, captions and table notes included. The citation (Walters & Wilder, 2023) should lead to exactly one entry:
Walters, W. H., & Wilder, E. I. (2023). Fabrication and errors in the bibliographic citations generated by ChatGPT. Scientific Reports, 13, Article 14045. https://doi.org/10.1038/s41598-023-41032-5
In ICMJE (Vancouver) style, references are numbered in order of first mention; a reference found only in a table or figure legend is numbered from the text's first mention of that table or figure (ICMJE, 2026).
2. Does each entry match the published record?
ICMJE wants references verified using a bibliographic source like PubMed, or the original sources (ICMJE, 2026). Compare every field, down to each author: a title match can pass an entry whose volume or DOI is wrong.
The Walters and Wilder entry above follows the article page (Sci Rep 13, 14045): volume 13, article number 14045, no issue or page range. Crossref adds issue 1, which the journal doesn't print, so the entry leaves it out (APA, 2025b). For a journal that numbers its articles, APA writes "Article" plus the number in place of the page range (APA, 2025b). Missing an author, date, or page numbers? See missing reference information.
3. Does every DOI open the right paper?
Putting doi.org/ in front of a DOI turns it into a link that resolves to the latest information about the item (DOI Foundation, n.d.), and APA includes the DOI whenever a journal article has one (APA, 2025b). Open each one and check that the landing page shows your entry's title, journal, and year.
A DOI that resolves is not automatically the right one. For example, https://doi.org/10.1038/s41598-023-41032-5 opens Walters and Wilder's article on nature.com. Change the ending to 41033-4 and it opens a 2023 Scientific Reports paper on OLED microdisplays; change it to 41022-5 and doi.org finds no such DOI. If a DOI fails or opens the wrong paper, the DOI checker guide shows what to check.
You can hand checks 2 and 3 to a tool for a first pass. Paste the reference list into PerfectRefs; you don't upload the paper. The Existence check searches Crossref, OpenAlex, and other scholarly databases for each entry, flags the ones it can't find, and has a free daily allowance. The Style check compares your authors, title, journal, year, volume, issue, pages, and DOI field by field against the database record and shows the record's values next to a source link; one click accepts a fix. It costs 2 premium credits per reference, and new accounts start with 10. It works on the list alone, so check 1 is still yours.
4. Have any of your sources been retracted?
ICMJE makes authors responsible for checking that no reference cites a retracted article, except to refer to the retraction. For journals indexed in MEDLINE, search PubMed for "Retracted publication [pt]" (ICMJE, 2026). ICMJE also asks journals to label a retracted article as retracted in all its forms, so opening each article at its DOI is a second check.
To discuss a retracted paper, APA gives the original article first, then the retraction's year, journal, volume, issue, and page numbers (here an article number) in parentheses (APA, 2025b). This PLOS ONE article was retracted about two months after it appeared; its Crossref record lists the notice as an update:
Liu, M.-J., Xiong, C.-H., Xiong, L., & Huang, X.-L. (2016). Biomechanical characteristics of hand coordination in grasping activities of daily living. PLOS ONE, 11(1), Article e0146193. https://doi.org/10.1371/journal.pone.0146193 (Retraction published 2016, PLOS ONE, 11[3], Article e0151685)
In the text: (Liu et al., 2016).
5. Have you read each source, and does it support your sentence?
Cite only works you have read (APA, 2022), and be able to attest that each reference supports its statement; cite original research directly whenever possible (ICMJE, 2026). For each citation, find the passage, table, or result that backs your sentence. If you know a study only from another paper, read the original or use APA's secondary-source form (see APA 7 in-text citations).
6. Are preprints, in-press papers, and unpublished work labeled?
For a preprint, the citation should make clear that the work is a preprint, and if it has since been published, cite the published article instead whenever appropriate (ICMJE, 2026). For example, the Crossref record for the medRxiv preprint https://doi.org/10.1101/2023.03.16.23286914 links it to a journal article, so cite that:
Gravel, J., D'Amours-Gravel, M., & Osmanlliu, E. (2023). Learning to fake it: Limited responses and fabricated references provided by ChatGPT for medical questions. Mayo Clinic Proceedings: Digital Health, 1(3), 226-234. https://doi.org/10.1016/j.mcpdig.2023.05.004
In the text: (Gravel et al., 2023). For a preprint still unpublished, see citing an arXiv preprint.
For work accepted but not yet published, APA puts "in press" where the year would go (APA, 2025a); ICMJE uses "in press" or "forthcoming" (ICMJE, 2026). If it has since come out, update the entry.
For work submitted but not accepted, APA has you give the year the work was produced and keep "submitted for publication" out of the date (APA, 2025a). If your journal follows ICMJE (Vancouver style), cite it in the text as "unpublished observations," with written permission from the source (ICMJE, 2026).
7. What does the journal's guide for authors say?
Read the journal's reference rules; they can differ from the style manual. Elsevier tells authors to check the journal's Guide for Authors; journals offering its Your Paper Your Way option accept references "in any style or format, as long as the style is consistent," and say so at the start of the guide (Elsevier, 2026).
PLOS ONE sets its own rules: ICMJE (Vancouver) style, the first six authors and then et al., journal abbreviations from NCBI databases, and the DOI as in its example, "doi:" plus the number, not shortened and not a URL (PLOS, n.d.). Here is the Walters and Wilder entry laid out like PLOS's published-article example, with the abbreviation from PubMed:
Walters WH, Wilder EI. Fabrication and errors in the bibliographic citations generated by ChatGPT. Sci Rep. 2023;13: 14045. doi: 10.1038/s41598-023-41032-5.
8. Do author lists and issue numbers follow your style?
- APA lists up to 20 authors by surname and initials; from 21 on, it keeps the first 19, adds an ellipsis, then gives the final author (APA, 2025a). Other styles cut the list sooner; see when to use et al.
- APA always includes the issue number for a journal article (APA, 2025b). If the journal prints none, as with Scientific Reports, leave it out.
9. Did you check every reference an AI tool suggested?
ICMJE's rule from check 2 covers these too: verify each against a bibliographic source or the original (ICMJE, 2026). Treat a chatbot's list as leads and run every entry through checks 2 and 3. Walters and Wilder (2023) found fabricated citations at rates of 55% for GPT-3.5 and 18% for GPT-4, and substantive errors in 43% and 24% of the real ones. Hallucinated citations explains why the chatbot can't vouch for its own list; desk rejection covers what a reference to a nonexistent work can cost.
FAQ
What should I check in my reference list before submitting?
Start with three: citations and entries match both ways, each entry matches the published record, and each DOI opens the right paper (APA, 2022; ICMJE, 2026). Then check for retractions and the journal's own rules.
How do I find out if a source has been retracted?
Search PubMed for "Retracted publication [pt]" if the journal is in MEDLINE, and open the article at its DOI: ICMJE asks journals to label retracted articles (ICMJE, 2026).
Do I have to use the journal's reference style?
It depends. Elsevier journals with Your Paper Your Way accept any consistent style (Elsevier, 2026); PLOS ONE asks for ICMJE (Vancouver) style (PLOS, n.d.).
References
- American Psychological Association. (2022, July). Basic principles of citation. APA Style. https://apastyle.apa.org/style-grammar-guidelines/citations/basic-principles
- American Psychological Association. (2025a, October). Elements of reference list entries. APA Style. https://apastyle.apa.org/style-grammar-guidelines/references/elements-list-entry
- American Psychological Association. (2025b, August). Journal article references. APA Style. https://apastyle.apa.org/style-grammar-guidelines/references/examples/journal-article-references
- DOI Foundation. (n.d.). Digital Object Identifier (DOI) system. Retrieved November 7, 2026, from https://www.doi.org/
- Elsevier. (2026, June 15). How should I prepare the references in my manuscript? https://www.elsevier.support/publishing/answer/how-should-i-prepare-the-references-in-my-manuscript
- International Committee of Medical Journal Editors. (2026). Recommendations for the conduct, reporting, editing, and publication of scholarly work in medical journals. https://www.icmje.org/icmje-recommendations.pdf
- PLOS. (n.d.). Submission guidelines. PLOS ONE. https://journals.plos.org/plosone/s/submission-guidelines
- Walters, W. H., & Wilder, E. I. (2023). Fabrication and errors in the bibliographic citations generated by ChatGPT. Scientific Reports, 13, Article 14045. https://doi.org/10.1038/s41598-023-41032-5
Original article by PerfectRefs. To republish it, credit PerfectRefs and link to this page. Do not edit it or use it commercially.
