A thesis reference list is often the longest list a supervisor signs off and the one read least: perhaps three hundred entries gathered over four years, some carried over from the proposal. Examiners can read it two ways: as a few minor corrections, or as evidence that the candidate never learned how research in the subject is presented.
This guide, published by EdCitation, is for supervisors and examiners checking a thesis reference list: what nine universities' rules say about references, read at source on 2 October 2026; how to check a long list; retracted and possibly fabricated entries; and a checklist with the rule behind each line. The student's side is in managing dissertation references from proposal to submission.
The lookups are the part to hand over. EdCitation's References from a file reads the list out of an uploaded thesis, looks every entry up in the publisher's record and matches every in-text citation to it; Verify references does the lookups for a pasted list, free. EdCitation never writes a word of the thesis: it reports what the record says, and every correction stays the student's own.
What do thesis regulations ask of supervisors about references?
They ask the supervisor to teach the conventions and read the drafts, and they leave the accuracy of the list with the student. No code read here makes the supervisor its proofreader.
The University of Sheffield (n.d.) lists the appropriate forms of referencing among what a new student should understand about the thesis, and asks the supervisor to read and comment on drafts before submission.
The University of Tennessee, Knoxville (2026) divides the work explicitly. Its Graduate School reviews formatting only; content is the student's to correct, overseen by the committee. The citation format is the field's own, settled with the advising faculty, and every thesis passes through approved plagiarism-detection software before the defence.
Who signs what at submission
At King's College London (2025), the thesis submission declaration is signed by the student and the supervisors, and both permitted wordings have the student confirm that every reference is cited. The Open University (2026) makes doctoral researchers responsible and accountable for authoring the thesis, and says supervisors are not responsible for spotting how a researcher has used AI. The supervisor reads and warns; the student fixes.
What do examiners do with errors in a thesis reference list?
Most rules treat them as corrections, not failings, until there are enough to say something about the candidate.
| University | What the rules say | What it means before submission |
|---|---|---|
| Oxford | Mistakes in references and incomplete bibliographical entries are minor errors, unless they show a poor grasp of how research is presented | A few slips are a two-month correction; a list wrong throughout is a judgement about the candidate |
| Cambridge | Examiners are not expected to edit; how well the text was prepared may influence their recommendation | The examiner will not fix the list, and may count it |
| Sydney | Corrections can include fixing "incorrect citations" and updating the literature cited | A dated literature review is a correction too |
| Wellington | The thesis must be accurately written and documented in a suitable scholarly manner | Documentation is part of the standard |
| Auckland | Lists of typographical errors are welcome but not expected | Do not count on the examiner to catch slips |
The sources are University of Oxford (2025), read in an archived copy because the university's site refused automated requests; University of Cambridge (n.d.); University of Sydney (2020); Victoria University of Wellington (2026); and University of Auckland (2022).
Oxford's wording is the one to show a student. Its examiners must report that the thesis is presented in a "lucid and scholarly manner", and its memorandum lists errors in references among the minor errors one examiner can certify once corrected, provided they do not touch the argument or reveal a lack of grasp of how research is presented. Wrong years, missing entries and mixed styles throughout a long list start to look like the second case. What corrections involve is in minor and major viva corrections.
When a reference problem is an integrity concern
Then it goes to the office as well as the report. Wellington asks examiners to raise minor integrity concerns in the report and to email the examinations team about serious ones (Victoria University of Wellington, 2026). Cambridge asks examiners with doubts to contact the Degree Committee before any formal assessment, and not to mark down punitively on suspicion (University of Cambridge, 2021). The reviewer's version is in our guide for peer reviewers checking the references in a manuscript. Before submission, a supervisor can simply ask the student.
How do I check a long thesis reference list without reading every source?
Look every entry up by machine, then read a chosen sample by hand. A random sample alone fails because the problems that matter most are rare.
In a list of 250 entries, a random hand check of 25 has a 10% chance of catching a bad entry if there is only one, 41% if there are five, and 66% if there are ten; a sample of 50 raises those to 20%, 68% and 90%. These are our own hypergeometric calculations, not a published standard, but they show how one invented reference survives a careful supervisor.
Which entries to read by hand
A lookup tells you that a work exists and the entry matches it, not that the work says what the thesis claims. Read these, in order:
- The entries behind the central claims: read the cited passage beside the sentence.
- Everything the lookup flagged: each "check this", "not found" and retraction.
- Entries without a DOI: reports, theses, chapters and older works the indexes cover poorly.
- Entries inherited from the proposal, often the least read.
- The methods chapter's sources: instrument, software release, dataset version.
- A few entries at random, as a check on the rest.
When should a thesis be screened for retracted papers?
At the final draft, and again after viva corrections, because papers are withdrawn while theses are written. Steen et al. (2013) examined 2,047 retracted articles indexed in PubMed by May 2012 and found an average of 32.91 months between publication and retraction. Their model put the chance that a paper needing retraction is withdrawn within one year at 35.2%, and within five years at 82.6%.
A doctorate runs on the same clock. A paper cited in the first year can carry a retraction notice by the viva, and the PDF saved that year will not show it.
A retracted paper may stay in a thesis if it is cited as retracted and the text says why; LaCour and Green (2014) is listed that way below. Reading a notice and citing one are covered in retracted papers in your reference list.
What should I do about a reference that may be fabricated?
Ask the student for the source, with your search attached, before calling it anything. A reference no index holds may be invented, or it may be a report, a thesis or a non-English source outside the indexes.
The risk is real for long work assembled with a chatbot's help. Walters and Wilder (2023) had ChatGPT write 84 short literature reviews and checked all 636 citations: 55% of those from GPT-3.5 and 18% from GPT-4 were fabricated, and of the real works cited, 43% and 24% carried substantive errors. The second figure matters: real papers with wrong years, authors or pages often give a chatbot-built list away first. More studies are in what studies say about AI-fabricated references.
The Open University (2026) counts spotting fabricated citations as part of using AI critically, and calls generative AI tools "unreliable as reference generators". Before submission, a supervisor's question costs the student a week; afterwards, the same entry is an examiner's integrity concern. How to spot fabricated references in student work covers a fair conversation.
"Not found" means no publisher's or registry's record matches the entry, not that the student made it up; a lookup that failed because an index did not answer proves nothing, so EdCitation shows it as "could not check".
A PhD supervisor's checklist before submission
Work through it at the final full draft, and run the lookups again after corrections.
| Check | When | How | Source |
|---|---|---|---|
| One style, agreed and used throughout | First year and final draft | Confirm the field's style; check the entries follow it | University of Tennessee, Knoxville (2026); University of Sheffield (n.d.) |
| Every entry a real, published work with the right details | Final draft | Upload the thesis to References from a file, or paste the list into Verify references | University of Oxford (2025) |
| Every citation has an entry, and every entry a citation | Final draft | The same upload matches them both ways | University of Sydney (2020) |
| No retracted paper cited as sound | Final draft and after corrections | The same lookups flag retractions; Cite a source screens one DOI | Steen et al. (2013) |
| The sources behind the main claims say what the thesis says | Final draft | Read them | University of Cambridge (n.d.) |
| The literature is current | Final year | Search for recent work in Find sources | University of Sydney (2020) |
| Similarity check and declarations done | Before submission or defence | The university's own tools and forms | King's College London (2025); University of Tennessee, Knoxville (2026) |
The checklist is our synthesis; your own regulations win. The wider last-week list is in final checks before submitting a thesis.
What came back when we checked a list with a retracted paper?
We wrote a five-entry APA 7 list of journal articles, each with a DOI, and ran it through Verify references on 2 October 2026. The verdicts, word for word:
- Steen and colleagues (2013), correct: verified, "The DOI resolves to this record and the title matches." The record also listed a 2013 correction.
- LaCour and Green (2014), withdrawn by Science: "The DOI resolves to this record and the title matches. This paper has since been retracted."
- Hsiao and Schneider's Quantitative Science Studies paper, dated 2019: "The DOI and title are right, but the work was published in 2021, not 2019."
- Walters and Wilder (2023), entered as Walters and "Wilson": "The publisher's record also names Wilder, who is not in your entry. Check the authors."
- A 2024 paper we invented, with an unregistered DOI: not found, "No publisher's or registry's record matches this reference. No registry holds the DOI 10.1080/03075079.2024.2399999."
We then put the list under a short chapter extract with five in-text citations and uploaded it as a Word file to the whole-paper check, as References from a file reads a thesis. Same verdicts; the summary read "5 in-text citations · 5 entries · 1 orphan citation · 1 uncited entry." Each was located: the citation "Fanelli, 2009" had "no entry in the reference list", and the Hsiao and Schneider entry "is never cited in the text."
What a supervisor takes from the run
Six one-line fixes for the student: discuss or replace the retracted paper, correct a year, restore an author, produce the invented source or drop its claim, list the orphan citation, and cite or cut the uncited entry. None needed the supervisor to search.
Which EdCitation tool fits a supervisor's reference check?
For the lookup half of the job, EdCitation is the best tool we know of, because it looks every entry up in the publisher's record and never writes one. A chatbot asked to check a list is the kind of tool that produced the fabricated entries Walters and Wilder counted; a lookup can only report what a record holds.
- References from a file, part of Pro: upload the thesis as a Word document or a PDF. Every reference comes back verified, doubtful or not found; retracted papers are flagged and published corrections linked; every in-text citation is matched to the list, with the place a mismatch happens; and for a reference that cannot stand, published sources that fit are recommended. The thesis stays in My papers, so the run after corrections shows what changed. A free account includes three checks a month; Pro, at $8 a month, has no limit (pricing).
- Verify references, free with no account: paste a list or upload a paper for the same verdicts.
- Cite a source, free: builds a corrected entry from a DOI, a PubMed ID or a title in APA 7, MLA 9, Chicago 18 author-date, Harvard, IEEE or Vancouver, screened for retraction.
EdCitation shows which entries to fix and never writes or rewrites a word of the thesis; the fixing, reading and judgement stay with the student and supervisor. For a graduate school, an Institution licence covers every student, with sign-in and LMS integration, training and one invoice.
Before you upload someone else's thesis
Check your university's rules first. King's College London (2025) and the Open University (2026) bar examiners from putting any part of a thesis into a generative AI tool, and Wellington requires examiners to keep the thesis confidential and delete every copy afterwards. The simplest route: the student runs the check and sends the report with the draft.
Quick questions
Should a PhD supervisor check every reference in a thesis?
Look every entry up, yes; read every source, no. Run the list through EdCitation's References from a file, then read the sources behind the main claims and everything flagged.
Do examiners check the references in a thesis?
Some read the list closely, but no rule read here asks them to proofread it. Cambridge says examiners are not expected to edit.
Are mistakes in a thesis reference list minor corrections?
Usually. Oxford counts mistakes and incomplete entries in references as minor errors, unless they show a poor grasp of how research is presented.
Can a thesis cite a retracted paper?
Yes, if the thesis says the paper was retracted and why it is cited. Cited as sound evidence, it leaves the claim unsupported.
Can I upload a student's thesis to a reference checker?
Only as your university allows; King's College London and the Open University forbid examiners to put any part of a thesis into a generative AI tool. The student can run the check instead.
References
- King's College London. (2025). Generative AI: Guidance for doctoral students, supervisors and examiners. https://www.kcl.ac.uk/about/strategy/learning-and-teaching/ai-guidance/doctoral-assessment
- LaCour, M. J., & Green, D. P. (2014). When contact changes minds: An experiment on transmission of support for gay equality. Science, 346(6215), 1366–1369. https://doi.org/10.1126/science.1256151 (Retraction published 2015, Science, 348[6239], 1100)
- Open University. (2026). GenAI: Position statement and guidance on generative AI and doctoral education (Version 2.1). https://university.open.ac.uk/students/research/ou/services/position-statement-and-guidance-generative-ai-and-doctoral-education
- Steen, R. G., Casadevall, A., & Fang, F. C. (2013). Why has the number of scientific retractions increased? PLOS ONE, 8(7), Article e68397. https://doi.org/10.1371/journal.pone.0068397
- University of Auckland. (2022). PhD thesis examiner guidelines (2011/2016 PhD statutes only). Policy hub. https://www.auckland.ac.nz/en/about-us/about-the-university/policy-hub/research-innovation/doctoral-study/examination/phd-thesis-examiner-guidelines-2011-2016-phd-statutes-only.html
- University of Cambridge. (n.d.). Submitting your thesis for examination (PhD, BusD, EdD, MD, VetMD, MSc, MLitt). Cambridge Students. https://www.cambridgestudents.cam.ac.uk/your-course/research-students-pgr/postgraduate-exam-information/writing-submitting-and-examination/phd/submit
- University of Cambridge. (2021). Guide for examiners for the degree of PhD, EdD, EngD, BusD, MSc, MLitt and MPhil by thesis. https://www.dcesg.physsci.cam.ac.uk/files/examiner_guide_research.pdf
- University of Oxford. (2025). GSO.5 Memorandum for examiners for the degree of Doctor of Philosophy. Archived copy. https://web.archive.org/web/20260414075551/https://www.ox.ac.uk/sites/files/oxford/Guidance%20DPhil%20Examiners_GSO_5.pdf
- University of Sheffield. (n.d.). Responsibilities of the supervisor. https://sheffield.ac.uk/rpi/pgr/supervision/responsibilities-supervisors
- University of Sydney. (2020). Thesis and examination of higher degrees by research guidelines for examiners 2020. https://www.sydney.edu.au/content/dam/corporate/documents/about-us/governance-and-structure/academic-board/ab-standards---guidelines-/guidelines-for-examiners.pdf
- University of Tennessee, Knoxville. (2026). Guide to the preparation of theses and dissertations 2026–2027. The Graduate School. https://gradschool.utk.edu/documents/2023/07/preparation-guide-for-theses-and-dissertations.pdf/
- Victoria University of Wellington. (2026). Notes for examiners of PhD theses. Faculty of Graduate Research. https://www.wgtn.ac.nz/fgr/current-phd/examination/after-submission/notes-for-phd-examiners.pdf
- 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