Did someone help with your thesis? Two checks before you submit

Almost nobody finishes a thesis alone. Supervisors suggest sources. Writing centres restructure chapters. Statisticians run the models. Translators render a draft into a second language. Professional proofreaders are permitted by most institutions and used by a great many candidates. And increasingly, some part of a literature search has been done with an AI tool.

Most of that is legitimate, disclosed and unremarkable. But all of it shares one property that is easy to lose sight of at the end of a long project: the accountability does not transfer. Whoever contributed, you are the person whose name is on the submission and the person who will be asked, in a room, why a particular source supports a particular claim.

This guide is two checks to run before you submit. Neither is about suspecting anyone. Both are about knowing what you are signing.

Why references specifically

Because they are the part of a thesis most likely to have been assembled by more than one pair of hands, and the part where a defect is least visible to re-reading.

A proofreader normalises reference formatting and may merge what look like duplicates. A supervisor suggests three papers you add at the last minute. A research assistant compiles a section's sources into your manager. A translator rebuilds a bibliography in a different convention. Every one of those is a reasonable thing to have happened, and every one of them can produce a citation that no longer resolves to what it names — without anybody doing anything wrong.

You do not have to distrust anyone to want to know that every source in your bibliography exists and says what you cited it for.

Check one: did every citation reach the list?

The completeness question. Every citation in your text should correspond to an entry in your bibliography, and every entry should be cited somewhere in your text.

This is where late help shows up most reliably. Sources suggested near the end are added to the text under time pressure and often never make it into the list — and the reverse happens too, where a helper adds entries to the bibliography for material that was ultimately cut.

Citations in the text                    612
Entries in the bibliography              347

  matched both ways                      331
  cited, no matching entry                 9   ← add these
  listed, never cited                      7   ← cite or remove

Nothing was dropped for being hard to read.

Nine and seven, on a thesis of that size, is an ordinary result rather than an alarming one — a small number of each is what a healthy report looks like. What matters is that you know the number, and that it is not sixty. The mechanics of doing this pass yourself are in our guide to finding missing citations.

Check two: does each source exist, and is it the one you cited?

The authenticity question, and the one that has changed most in the last two years.

A reference can be perfectly formatted, sit in the right place alphabetically, and describe a paper that was never written. This is not hypothetical or rare: generative tools produce plausible reference lists containing entries that do not exist, and a candidate who received a reading list from one has no way to tell by looking. We have written about what to do when that has already happened, and the first step is always the same — find out which ones are real.

The subtler version is a reference that exists but is not the one your citation points at. Two researchers share a surname and a year; a DOI was copied from a stale record; an edition was superseded. The entry is real, the citation is real, and the pair is wrong.

(Okonkwo & Ferreira, 2022)
  matched to   Okonkwo, C. B., & Ferreira, L. (2022). Workflow
               variance in regional centres.
  resolved     against a real record · title and authors agree
  verdict      matched

(Alvarez, 2021)
  matched to   Alvarez, M. (2021). Cohort effects in adherence.
  resolved     no record found for this title and year
  verdict      unverified — check this one yourself

The important column is resolved. A tool that tells you a citation "matches" without showing you what it matched against is asking to be believed. What you want is the evidence beside each pair, so that you can disagree with any of it — and so that the ones it could not verify are visibly separate from the ones it could.

What to do with something that does not resolve

Almost always: find it and fix it. An unresolved entry is far more often a wrong year, a mistyped identifier or an unusual venue than a fabrication. Search for the title directly, check the author's own publication list, and see whether a preprint became a journal article somewhere along the way.

If you genuinely cannot find a source, the answer is not to leave it in and hope. A citation you cannot verify is a claim you cannot defend, and the safe move is to replace it with something you have read or remove the sentence that depends on it. That is a twenty-minute decision now and an unanswerable question in a viva.

And if a number of entries do not resolve — not one or two, but a pattern — that is worth a calm conversation with whoever assembled that part, before submission rather than after. Not as an accusation. As the same thing you would want if the positions were reversed.

What this is not

It is not a test of whether you were helped, and it cannot be used as one. A thesis written entirely alone produces the same categories of finding as one that had four contributors, because these defects come from the process of assembling a long document, not from who did the assembling. Anyone who tells you a reference report reveals authorship is selling something.

Nor is most of it about misconduct. Missing entries, orphans and stale identifiers are quality problems with quality fixes — the institutional case for treating them that way applies just as much to a candidate checking their own work.

But be clear-eyed about the second check, because that is the one it exists for. If a source in your bibliography cannot be found to exist, you are citing work that was never written, and that is academic misconduct regardless of who put it there or whether anyone meant it. Your name is on the submission. Finding it yourself, now, is the difference between a correction you make quietly and a question you cannot answer in a viva.

The point is narrower and more useful than either. You are about to put your name to a document containing several hundred factual claims about what other people wrote. It is worth an afternoon to know that those claims are true.