Building a bibliography you can defend

The list is a set of claims

A reference list is not decoration. Every entry asserts: this work exists, I consulted it, and it supports what I attached it to. Three claims per line, and you are answerable for all of them.

That framing settles most questions about how to do this. A reference you did not open makes a false claim, whether it came from a model, a friend’s essay, or a paper’s own reference list copied without checking.

Zotero, in ten minutes

Zotero is free and open source, and it does the job better than EndNote, which universities pay for. Install the desktop app and the browser connector.

Then the workflow is: read a paper, click the connector, and the reference plus the PDF is saved. Write in Word, LibreOffice or Google Docs with the Zotero plugin, insert citations as you write, and generate the bibliography in whatever style your department demands. Changing style is one dropdown, which matters more than it sounds when a department switches from Harvard to APA between years.

Two cautions. Zotero’s automatic metadata is usually right and sometimes wrong: check chapters, older books and anything from an unusual site. And keep your notes in Zotero attached to the item, so the four-line note from the previous lesson lives with the reference rather than in a separate file you will lose.

Free alternatives: JabRef if you use LaTeX, Mendeley free tier, or a plain BibTeX file maintained by hand, which is entirely respectable.

Why models get citation formatting wrong

Styles are rule systems with many exceptions, and a model reproduces the general pattern while missing the specific case: a chapter in an edited volume, a translated work with two dates, a government report, a dataset, a preprint, a court judgment. The output looks right and is subtly wrong in the way that costs marks in a coursework rubric with a presentation criterion.

Zotero applies the actual style specification, which is maintained by people who care about the edge cases. Use the tool.

One legitimate use of a model here: explain why this reference is formatted this way in APA when you cannot work out what your source type is. Explanation, not generation.

Citing an AI tool

If you used a model in a way your institution asks you to declare, cite it in whatever form they specify. Most style guides now have a form for it, and most treat it as a personal communication or software rather than a source, because it is not retrievable: nobody else can go and read what it told you.

That unretrievability is exactly why a model can never be a source for a factual claim. A citation exists so a reader can check. A conversation nobody else can open is not checkable, so it supports nothing.

What you cite, therefore, is the use, as a disclosure of method, and never as evidence for a fact. The fact needs a real source underneath it.

Before you submit

A fifteen-minute check that has saved a great many people:

1. Every in-text citation appears in the reference list, and vice versa. Zotero does this; check anyway.
2. Every DOI resolves.
3. For every reference: can you say, out loud, what it argues and which page you used?
4. Any reference where the answer to 3 is no comes out. Not gets checked later. Comes out.
5. Every direct quotation matches the source word for word, including punctuation.
6. Nothing is cited from a summary you read rather than the source itself.

If that check removes half your bibliography, that is a finding about the essay, and it is much better to discover it now.

The reference list as a reading plan

One reframing makes all of this less painful. Build the bibliography as you read, from the first day, rather than assembling it the night before. Every paper you read gets saved and annotated at the moment you read it, with the four-line note and the page numbers. The reference list then writes itself, and it contains exactly the works you actually used.

Students who assemble a bibliography at the end are doing something different and much harder: reconstructing, under time pressure, which source said what. That reconstruction is where fabricated and misattributed citations enter an otherwise honest essay.

The standard, stated plainly

Cite what you read. Read what you cite. Say where in it.

That standard predates every tool discussed in this course by several centuries, and none of them changed it. What changed is only how easy it has become to produce a list that looks like it meets the standard without meeting it.