Finding and Merging Duplicates in Zotero

Zotero’s Duplicate Items view (in the left-hand pane of the desktop app) automatically groups records it judges to be the same source, so you can select them and merge. For a small personal library that mostly works. It breaks down fast for a systematic review library assembled by merging several people’s exports, database downloads, and reference-manager migrations — because Zotero’s matching algorithm is narrower than most researchers assume, and because merging itself has a hard rule that silently blocks the exact case a systematic review produces most often: the same paper saved twice as two different item types.

How Zotero actually decides two records are duplicates

Per Zotero’s own documentation, the Duplicate Items algorithm works in two tiers. First, it checks for a match on title, DOI, or ISBN. If none of those line up — or a record is missing them — it falls back to a looser check: publication years within one year of each other, plus at least one shared author (matched on last name and first initial only). Zotero’s own docs describe this as a starting point, noting the algorithm “will be improved in the future to incorporate other fields.” It is a real feature, and it catches the common case (the same PDF saved twice from the same database), but it is not the exhaustive check a systematic reviewer needs before reporting a final included-study count.

Where the Duplicate Items view actually breaks down

Same paper, different item type: flagged, but not mergeable

The case that hits systematic-review libraries hardest is a paper saved once as a Preprint (from bioRxiv, arXiv, or medRxiv during initial screening) and again as a Journal Article once the published version comes out — or a conference paper later expanded into a journal article. If the title and authors are close enough, Zotero’s Duplicate Items view does group these two records. But merging two items requires them to be the same item type — Zotero has no way to hold fields from two different item-type schemas in one merged record, so the Merge option is unavailable until you manually change one record’s item type to match the other. This is not a bug so much as an unadvertised precondition: the pane shows you a duplicate pair and offers no obvious next step, so it looks broken rather than blocked.

The fix is to change the type first: open the older or less-complete record, use the Item Type dropdown at the top of the right-hand pane to change it to match its pair (e.g. change the Preprint to Journal Article), fill in any newly-required fields Zotero exposes for the new type (volume, issue, pages), and only then select both records and merge. Do this before you merge, not as a cleanup step after — a mismatched item type is also why an otherwise-correct record can silently disappear from an export filtered by item type.

DOI or title mismatch: not flagged at all

The inverse failure is quieter and more dangerous for review accuracy: two records that are genuinely the same paper but don’t share a matchable title, DOI, or ISBN never enter the Duplicate Items pane in the first place. This happens routinely in a systematic-review library because:

• A preprint is deposited with its own DOI (a Crossref or DataCite DOI issued by the preprint server) that is entirely different from the DOI the publisher later assigns to the peer-reviewed version.
• A manuscript’s title changes between submission and publication — a common outcome of peer review — so the preprint title and the published title no longer match closely enough for Zotero’s title comparison to catch it.
• One record was imported without a DOI at all (a common gap in older database exports or manually-entered records from a hand-searched reference list), so there’s nothing for Zotero’s primary check to compare.

None of these trigger the fallback year+author check reliably either, since a preprint and its published version can carry different years (deposited one year, published the next) and the fallback still requires the years to be within one of each other. A pair split by two full publication years — not uncommon during a long peer-review cycle — will not be flagged by either tier of the algorithm.

Records split across My Library and a group library

Duplicate detection also runs within a single library only — it does not compare items in your personal library against items in a shared group library, even if both are open in the same Zotero window. A systematic-review team that has some members importing directly into a shared group library and others importing into their own library before copying items over will accumulate cross-library duplicates that the pane never surfaces.

A manual reconciliation pass for a merged review library

Before finalizing a study count for a PRISMA flow diagram, treat the automatic Duplicate Items view as a first pass, not a final answer, and run a manual check on top of it:

1. Sort by title, not by date added. In the main library view, click the Title column header. Records for the same paper under slightly different titles (a preprint’s working title vs. the published title) will usually still land near each other alphabetically, which a date-added sort will scatter.
2. Add a DOI column and scan for blanks. Right-click any column header, choose DOI from the list, and sort by it. A record with a blank DOI is exactly the kind the automatic check is most likely to have missed against a DOI-bearing sibling — check each one by hand against the title/author list nearby.
3. Tag ambiguous pairs instead of guessing. Where you’re not sure two records are the same study, apply a shared color-coded tag (e.g. “verify-duplicate”) rather than merging on a guess or leaving it unresolved — this keeps the review auditable and lets a second team member confirm before anything is merged or discarded.
4. Align item type before merging, every time. If Duplicate Items groups a pair but the Merge action is unavailable, that is the item-type mismatch above — fix the type first (see the fix described earlier), then merge.
5. Use Related for pairs you deliberately keep separate. A preprint and its published version are sometimes worth keeping as two linked records rather than merging — for instance, if your protocol cites the preprint’s earlier deposit date for a registration timeline. Zotero’s Related tab lets you link two items without merging them, which keeps both discoverable without duplicating them in your reported study count.
6. Re-run Duplicate Items after any bulk import. Importing a second search’s results, a co-reviewer’s exported library, or a citation-management-software migration each re-triggers new duplicate pairs — treat the pane as something to re-check after every merge event in the review, not a one-time pass at the start.

Merging correctly once you’ve confirmed a real duplicate

Once two or more records are genuinely the same source and share an item type, select them (in the Duplicate Items view, or anywhere in the library with the same item type held down via Ctrl/Cmd-click), right-click, and choose Merge Items. Zotero lets you pick which record is the master (its field values win by default) and shows an icon beside any field where the records disagree, so you can pick the better value field by field rather than losing information from the non-master record. Collections, tags, and notes from every merged record are retained on the resulting single item, and any existing citations in a manuscript that pointed at a merged record continue to resolve correctly.

Frequently asked questions

Can I mark two records as “not a duplicate” so Zotero stops suggesting them?

Not directly — Zotero’s Duplicate Items view has no persistent dismiss action, so a pair that looks like a match but genuinely isn’t (two distinct papers by the same close-collaborating authors published the same year, for instance) will keep reappearing in the pane. Moving on past it is the normal workflow; there’s no “false positive” flag to set.

Does merging duplicates delete my citations in an in-progress manuscript?

No. Merging consolidates the underlying library record, but citations already inserted via Zotero’s word-processor plugin continue to point at the surviving merged item, since Zotero resolves citations by the item’s internal key rather than by which specific pre-merge record you originally cited.

Is there a way to force-merge two items of different item types?

Not in Zotero itself — the item type has to match before the Merge option becomes available, since a merged record can only carry one item type’s field schema. Third-party plugins exist that automate changing the non-master item’s type as part of the merge, but that is a plugin behavior, not a built-in Zotero option.

See also: Zotero for Researchers, Zotero setup and troubleshooting, citation management software for a research group, sharing a Zotero library, the RIS file format, Better BibTeX for Zotero, systematic review protocol template, ASReview for systematic-review screening, and the Research Tools hub.