Academic Writing

Citation Managers in 2026: Zotero, Mendeley, EndNote, and What Actually Works

A researcher's desk with a laptop showing a reference library beside stacks of annotated journal articles and a coffee cup.

A colleague of mine, a historian in her early forties, opened her laptop last spring to find that several hundred of her PDFs had been quietly locked behind a login wall. Her citation manager had been updated. The sync had triggered some new licensing check. The library was still there, in the sense that the metadata was visible, but the actual files — annotations included — wanted her to sign in and, eventually, upgrade. She had been using this software since graduate school. She did what a lot of scholars have done in the past three years: she exported what she could, swore at the screen for a while, and migrated to Zotero.

This is the story of citation management in 2026, compressed. The tools have converged in features and diverged sharply in trust. What you pick now is less a question of which app has the best PDF viewer and more a question of whose business model you are willing to hitch a twenty-year research habit to.

Zotero remains the honest answer. It was built at the Center for History and New Media at George Mason University in 2006, originally as a Firefox extension, by a group of historians and digital humanists led by Dan Cohen and Sean Takats. In 2013 the project moved under a nonprofit, the Corporation for Digital Scholarship, which still maintains it. The software is open source, the storage format is a straightforward SQLite database you can open with other tools, and the sync server is optional, if you don’t want to pay Zotero for cloud storage, you can point it at your own WebDAV or Nextcloud and it will not argue with you. In 2026 the desktop application is in its seventh major version, with a rewritten PDF reader and note editor that finally match the best proprietary alternatives. Browser capture still works on almost every journal site. The plugin ecosystem is the best in the category; Better BibTeX alone is worth the cost of switching, which is zero. If you are starting a PhD this year, or if you are trying to decide what your lab standard should be, Zotero is the choice that will not betray you.

Mendeley is the cautionary tale. It launched in 2008 out of London, built by a team that included Victor Henning and Jan Reichelt, and for a brief period it was genuinely exciting, a well-designed desktop app with a social layer that let you see what other researchers in your field were reading. Elsevier bought it in 2013 for a reported figure in the tens of millions. Anyone who had been paying attention to Elsevier’s relationship with open access braced themselves, and over the following decade the braces were justified. The original desktop app was discontinued in 2022 and replaced with Mendeley Reference Manager, a thinner application with fewer power-user features. Accounts are tied to Elsevier’s single sign-on, which means your library lives inside the same ecosystem as ScienceDirect and SciVal. Export is possible but awkward. The social features that once distinguished the product have been stripped back. It is not unusable in 2026, but it is a product being maintained, not developed, and every interaction carries a faint suggestion that you should also be signing up for something else.

EndNote is the institutional default in a great many universities, and that is mostly what keeps it alive. It was originally written in 1988 by Richard Niles in Berkeley, sold to ISI Researchsoft, then to Thomson Reuters, and now lives under Clarivate Analytics alongside Web of Science. The current version, EndNote 21, is competent. The Word plugin, Cite While You Write, is the reason most senior faculty will not switch; they have fifteen years of manuscripts wired to it, and the thought of remapping those documents is physically painful. Fair enough. For a new user in 2026, though, EndNote is hard to recommend. The license is expensive, the interface carries the sediment of every UI fashion since the Clinton administration, and the company’s incentive structure, Clarivate sells citation metrics, journal rankings, and research analytics, is oriented toward institutional procurement, not toward the individual scholar at a desk. It will do the job. So will a hammer.

Beyond the big three, the landscape has genuinely shifted. Paperpile, which began as a Chrome extension focused on Google Docs integration, has matured into a full-featured manager with a clean web app and a reasonable subscription model. For researchers who live inside Google Workspace, and who do not need offline access as a hard requirement, it is the most pleasant option on the market. The PDF reader is good. The metadata extraction is unreasonably accurate. The sync is fast. The main caveat is exactly the one Mendeley users are now learning: you are renting, not owning, and if Paperpile is acquired by a larger company tomorrow, you will have the same migration problem my colleague had.

Citavi, long popular in German-speaking academia for its knowledge-organization features, not just references but quotes, categories, and an argument-mapping layer, was acquired by Lumivero in 2021, and the transition has been rough. The macOS and Linux versions remain second-class citizens. For dissertators who want a tool that does more than cite, though, Citavi’s category tree still has no real equivalent.

Among newer entrants, two are worth naming. ReadCube Papers, which absorbed the original Papers app from Mekentosj, has settled into a polished reading-focused product that works well for researchers who want Smart Citations and a strong discovery layer. Bookends, a single-developer Mac app maintained by Jon Ashwell since the 1980s, is small, fast, idiosyncratic, and quietly beloved by a certain kind of humanities scholar. It is not going to replace your lab’s workflow, but if you write alone on a Mac, it is worth trying.

The honest caveats. No citation manager will fix sloppy note-taking, and none of them, not even the good ones, will save you from a literature that has grown faster than any individual can read. The tools help with storage and formatting. They do not help with understanding, and it is worth being suspicious of features that imply they do. Several products now advertise AI summarization of PDFs and “chat with your library” functionality. Some of this is useful for triage. Much of it is the same outsourcing trap that appears everywhere else in academic software: it feels like thinking, and it isn’t. If you want AI involved in your reading, keep it narrowly scoped, extracting citations, cleaning metadata, generating a draft summary you then verify, and read the paper yourself. The same principle that applies to writing assistants in academic work applies here.

A few practical recommendations. If you are starting fresh, install Zotero, pay the small storage fee if you use many PDFs, and install Better BibTeX on day one. If you are on Mendeley and have been meaning to move, move now; the export-and-import path to Zotero is well documented and takes an afternoon. If you are on EndNote because your coauthors are on EndNote, stay, but keep a parallel Zotero library for your own reading, because someday your coauthors will not be the people deciding what you use. If you live in Google Docs and value your time more than your money, try Paperpile for a month. And whichever tool you pick, periodically export your library to BibTeX or RIS and keep the file somewhere you control. A citation manager is infrastructure. Infrastructure should never be something you cannot leave.

Photo via Unsplash.