Zotero, Mendeley or EndNote: which reference manager

What a reference manager actually does

Three jobs. It captures a source's metadata from a database or a web page; it stores that record with the PDF and your notes; and it inserts citations into your document, formatted to a style, with a reference list that builds itself.

The third job is the one people know about and the least important. Anyone can format a reference by hand. What you cannot do by hand is keep four hundred sources organised across three years, find the paper you half-remember from last February, or renumber a Vancouver reference list after inserting a citation on page twelve.

The second job is where the real value sits for a dissertation. A manager that holds the PDF and your notes together means your reading is searchable — you can find the sentence you highlighted eight months ago in ten seconds, which is the difference between a literature review you write and one you reconstruct.

The comparison

Cost, ownership and what happens if you stop paying are the things worth deciding on. Feature lists converge; those three do not.

Why Zotero for most people

It is free permanently, not free until you need more than a token amount of storage. Your library is a folder of files on your own machine in an open format, which means no vendor can change the terms on it and you can move it anywhere.

Its browser connector is genuinely better than the alternatives at reading a page and producing a correct record, including from library catalogues, newspaper sites and government publications that the others handle poorly. Since bad metadata is the main source of bad references, this matters more than it sounds.

It supports over ten thousand citation styles in CSL, an open format you can edit yourself, which is what you need when your institution has its own Harvard variant that differs from every published one — as most do.

The trade-off is polish. Zotero looks like a tool built by librarians, because it was. Some of the interface is unlovely and group features are less smooth than commercial alternatives.

When EndNote is the right answer

Three cases. Your department requires it and provides the licence, in which case use it and stop deliberating. You are doing a systematic review, where EndNote's deduplication and screening workflow is genuinely ahead. Or you are working with a supervisor whose library and shared groups are already in EndNote.

EndNote's Word integration is the most robust of the three under stress — a two-hundred-page thesis with eight hundred citations and a lot of tracked changes is where the others start to struggle.

The cost is the cost, and the licence usually ends with your enrolment. Export your library to RIS before you graduate; doing it afterwards means reinstalling software you no longer have a licence for.

Mendeley's position

Mendeley was the student favourite for years and remains competent. Its position now is mainly institutional: if your research group, lab or supervisor runs on Mendeley groups, the friction of being the only person on something else outweighs any feature difference.

It is owned by Elsevier, which is a publisher, and some researchers have views about that. It is a reasonable consideration and not a decisive one for coursework.

The practical caution is storage. The free tier fills quickly once you are keeping PDFs, and a library that will not sync is worse than no library at all.

The others worth knowing about

Two alternatives come up often enough to name. Paperpile is a paid, Google-Docs-native manager that is excellent if you write in Docs and nowhere else. JabRef is a free, BibTeX-native option that is the natural choice if you are writing in LaTeX.

If you are in mathematics, physics, computer science or engineering and writing in LaTeX, the question is different: your citations are BibTeX entries and the manager's job is to produce a clean .bib file. Zotero does this well through Better BibTeX; JabRef does it natively.

Ignore anything that only formats citations and does not store your library. A citation generator is a different tool for a different, smaller problem.

The metadata problem nobody warns you about

Every manager imports metadata from somewhere, and the somewhere is often wrong. Publisher records have typos, database records lose subtitles, and Google Scholar in particular produces records with the wrong year, missing author initials and journal names abbreviated inconsistently.

The consequence is a beautifully formatted reference list full of small errors, which markers read as carelessness rather than as software behaviour — reasonably, since you submitted it.

So build the check into your workflow: whenever you add a source, look at the record once. Authors, year, title capitalisation, journal name in full, volume, issue, pages, DOI. Twenty seconds per source, done as you go, versus a miserable evening before submission.

Set the style before you write, not after

Install your institution's own style file if it publishes one — most libraries do, precisely because their Harvard variant is not the same as anyone else's. If not, pick the closest published variant and check its output against your handbook on a test document.

Test four cases before you trust it: a two-author paper, a paper with three or more authors, a chapter in an edited book, and a web page with no author. Those four expose almost every template difference that matters.

Changing style later is a single dropdown, which is the whole point. What is not easy is fixing four hundred records whose metadata was wrong from the start.

Moving between managers

All three export and import RIS and BibTeX, so switching is possible and mostly painless. What survives is the metadata; what may not survive is your folder structure, tags, annotations and attached PDFs.

Export the library, import it into the new manager, then spot-check twenty records across different source types. Annotations made inside a manager's own PDF viewer are the most likely casualty, which is an argument for keeping substantive notes in the note field rather than as highlights.

Do this before a dissertation rather than during one. A manager change mid-thesis with live citation fields in a Word document is the kind of afternoon nobody plans for.

The habits that matter more than the software

Whichever you choose, these five make the difference between a manager that helps and one that becomes a second problem.

• Add the source the moment you decide to read it, not when you decide to cite it.
• Check the metadata once, on entry.
• Write a two-line note in the record saying what the source argues and why you kept it. In eight months this is what makes the library searchable.
• Use tags for themes and folders for chapters, not both for the same thing.
• Back up the library separately from your cloud sync, because a sync error propagates.

And do not trust it blindly

A reference manager is a tool with defaults, and it does not know your handbook. The final reference list still needs a human pass: check that et al. thresholds match your style, that the hanging indent is right, that no entry is a bare URL, and that nothing appears in the list that is not cited in the text.

Markers deduct for the output, not the process. 'Zotero formatted it' has never once been accepted, and it is not an unfair rule — the same standard applies to a spellchecker.