Managing Citations Across a Long Thesis Without Losing Track

A thesis is the longest document most researchers ever write, and it is usually written over years, in pieces, on more than one computer, with chapters that move, merge and split. Citations are where that history shows. Duplicate entries, broken references, quotations whose page numbers nobody can find and claims attributed to the wrong paper are all symptoms of a citation workflow that grew without a plan. This guide describes a workflow that holds up over a long project: how to structure a reference library, how to link notes to sources, how to keep citations stable through drafting and revision, and where a source-linked research assistant such as Yila fits alongside your reference manager.

Choose one library and one citation key scheme

The first decision is to have a single reference library for the whole thesis. It can be in Zotero, Mendeley, EndNote, JabRef or a plain BibTeX file under version control, but it should be one place. Separate libraries for each chapter seem tidy at first and become a source of duplicates and inconsistencies as chapters borrow from each other.

Decide early on a citation key format, for example author surname, year and first significant word of the title: smith2021tutoring. Stable, readable keys make it easy to cite while drafting, to search your notes and to spot duplicates. If your reference manager generates keys automatically, fix the format in its settings and avoid regenerating keys later, because changing a key breaks every place where it was cited.

Capture complete metadata at the moment of saving

Most citation problems begin when a reference is saved with incomplete or wrong metadata. Import from a DOI, a publisher page or a database record rather than typing details by hand, then check the fields that are often wrong: author names with diacritics, conference names, page ranges, and the distinction between preprint and published versions.

Preprints deserve particular care. If you cite a preprint and the paper is later published, decide whether to update the reference to the published version. The published version may differ, and page numbers or section structure may change. Record which version you actually read; if you update the citation, re-check any quotation or specific claim against the new version.

Attach the PDF you read to the library entry. Months later, when a reviewer asks where a statement comes from, the copy you annotated is far more useful than a fresh download.

Link notes to sources, not just to topics

Reading notes are where citations go wrong most often. A note such as "tutoring improves retention, several studies" is useless when you come to write, because you no longer know which studies or what they showed. Every note that contains a claim should carry the citation key and, ideally, a page or section reference.

A practical structure is one note per source, containing a short summary, the key findings with locations, your assessment of the evidence and links to related sources. Topic notes, such as a note on "peer learning and retention", can then reference the source notes rather than restating findings. When you draft, you move from topic notes to source notes to the original PDF, and each step keeps the trail intact.

Distinguish clearly between direct quotation, close paraphrase and your own interpretation. Mark quotations with quotation marks and an exact page number in your notes from the start. Retro-fitting page numbers at the end of a thesis is slow and error-prone.

Cite while drafting, not afterwards

Write citations into the draft as you go, using keys or the reference manager's plugin, even in rough drafts. A placeholder such as "(ref)" or "[cite]" always seems temporary and often survives into the submitted version. If you are unsure which source supports a sentence, flag the sentence explicitly, for example with a comment, rather than inserting a likely-looking citation. A wrong citation is worse than a missing one, because it looks finished.

For long theses written in Word, use the reference manager plugin consistently and avoid copying formatted citations between documents, which can turn live fields into plain text. For LaTeX, keep one bibliography file for the whole thesis and include it from every chapter.

Keep chapters consistent when they move

Chapters change shape during a thesis. A literature review chapter is split into two, a methods section becomes an appendix, a paper published from chapter four is reworked back into the thesis. Each move risks orphaned or duplicated citations.

Before major restructuring, run a check: compile the bibliography and look for undefined keys, unused entries and near-duplicates. Most reference managers can find duplicates by DOI or title. When a chapter is based on a published paper, make sure the thesis version cites the same sources the paper did, and decide how to cite the paper itself according to your institution's rules.

Audit before submission

Plan a citation audit a few weeks before submission. The audit has three parts. First, a technical check: every in-text citation resolves to a bibliography entry, every entry is cited, and the style matches the required format. Second, a metadata check: scan the bibliography for missing years, inconsistent capitalisation, truncated author lists and preprints that have since been published. Third, a content check on a sample of claims: pick a set of sentences with citations, especially quantitative claims and direct quotations, and verify each against the source.

The content check is the one people skip, and it is the one that matters most to examiners. A misattributed finding or a quotation that does not appear on the cited page undermines confidence in everything else.

Where an assistant helps

A reference manager stores sources and formats citations; it does not check whether a sentence in your draft is actually supported by the paper it cites. That gap is where a source-linked assistant can help. Yila is designed to turn papers and data into cited reviews, figures and slides, keeping each statement tied to the passage it came from. Used on a set of sources you have already selected, it can produce a draft summary in which every claim links back to a specific location, which makes the content check in your audit much faster: you follow the link and confirm the wording.

The assistant does not replace the library or the citation keys. It works best as a layer on top of a clean library, where the sources are the ones you chose and read, and where its output is checked before any sentence enters the thesis.

A compact checklist

• One reference library and one stable citation key format for the whole thesis.
• Import metadata from DOIs or databases and check authors, venues and versions.
• Attach the PDF you actually read.
• One note per source, with citation keys and page references for every claim.
• Mark quotations and exact pages from the start.
• Cite while drafting; flag uncertainty instead of guessing.
• Check for undefined, unused and duplicate entries before restructuring.
• Run a technical, metadata and content audit before submission.

Conclusion

Citation management in a thesis is less about software than about habits that survive years of drafting. A single library, stable keys, notes linked to precise locations and a planned audit make the difference between a bibliography that holds up under examination and one that has to be rebuilt in the final month. A source-linked assistant like Yila can make the checking faster, but it rests on the same foundation: every claim traceable to the source you read.