Why Zotero and EndNote exports sometimes miss the required style

When output looks wrong, the instinct is to blame the style file. Often the style is doing exactly what it was written to do with incomplete or misclassified data. But style files can also be outdated or represent a different local variant, so the useful first question is not "is Zotero broken?" It is "which layer supplied the wrong instruction or value?"

A CSL style is a set of instructions: put the journal abbreviation here, italicise the volume, use sentence case for the title. If the item in your library has no journal abbreviation stored, the instruction has nothing to place, and the style either falls back to the full title or leaves a gap. Neither is a bug in the style.

Reference managers automate formatting; they do not guarantee that imported metadata is accurate. Automation makes consistent data consistent and bad data consistently wrong.

• Metadata layer: author, title, date, container, volume, issue, pages, DOI and language must describe the source correctly.
• Type layer: the item must be a journal article, chapter, conference paper, dataset or other correct type so the style knows which fields apply.
• Style layer: the installed CSL or output style must match the requested edition and any journal-specific variant.
• Document layer: live citation fields must still be connected, refreshed and free from manual edits that the next update will overwrite.

A quick test separates the first two layers from the third. Open the library item and compare its raw fields with the source. If the author or date is already wrong there, switching styles cannot repair it. If the fields are correct but punctuation or ordering differs from the target example, inspect the selected style and variant.

Journal abbreviations are a frequent source of non-compliance in medical and chemistry submissions. Vancouver-style outputs commonly need NLM abbreviations and ACS journals may require CASSI forms. A general shortening rule cannot reliably derive either controlled abbreviation from every full journal title.

The field depends heavily on what a publisher page, database or site translator exposes. Common failure modes include an empty field, the full title copied into the abbreviation field, punctuation from the wrong convention, or an old title retained after a journal rename. Two records for the same article can therefore differ simply because one came from a publisher page and the other from a database export.

Zotero can auto-abbreviate in Index Medicus/MEDLINE form when the field is blank, which helps for medicine and does not solve ACS, whose abbreviations carry periods that the medical convention omits.

Zotero's documentation also explains an important precedence rule: if a Journal Abbr field is populated, it is used; automatic MEDLINE abbreviation is generally a fallback when that field is empty. A bad stored value can therefore override a better automatic result. Inspect the item before assuming the abbreviation engine failed.

• For NLM forms, verify the journal in the NLM Catalog rather than removing periods by eye.
• For chemistry, verify the target journal's instructions and CASSI form; not every house style follows the same punctuation.
• For renamed journals, use the title and abbreviation appropriate to the cited issue, not automatically the journal's current branding.
• For unidentified titles, do not invent an abbreviation. Many styles prefer the full title when no authoritative short form exists.

• Case. APA wants sentence case for article titles; MLA and Chicago want title case. Managers store the case that was captured. Databases export in every combination, and a style cannot reliably recase a title it cannot parse - especially with proper nouns, acronyms and species names in it.
• Author name parsing. "van der Berg", hyphenated surnames, Chinese names given family-name-first, and single-name authors all get split wrongly on import, and the style then formats the wrong halves.
• Item type. A conference paper saved as a journal article will be formatted as a journal article, flawlessly and incorrectly. Preprints, theses, standards and datasets are the usual casualties.
• Style currency. The Zotero style repository is community-maintained. A style can lag an edition change - APA 6 to 7 removed publisher location and changed the et al. threshold - and an out-of-date style produces consistent, obsolete output.
• Identifiers and locators. A DOI stored with a URL prefix, an article number placed in the pages field, or an online-first record missing its eventual volume can produce duplicates or malformed locators.
• Dates and editions. Web publication, issue publication and access dates are different fields. Books also need the cited edition, not merely the latest edition found online.
• Language and protected case. Styles cannot reliably infer that an acronym, gene symbol, place name or non-English title must retain unusual capitals unless the metadata marks it correctly.

Duplicates make these errors harder to see. One library record may hold the corrected authors while another, cited in the document, still has the bad title. Merge duplicates inside the manager rather than deleting one blindly, then refresh the document and confirm that every citation still points to the surviving item.

Manual edits in the rendered bibliography are fragile while citations remain live. The next refresh can overwrite them because the manager correctly regenerates output from the unchanged library data. If a correction should persist across drafts, make it in the item or style. Reserve output-only edits for a controlled final copy.

For an active manuscript, fixing metadata at source is the durable move: correct the library item and the style can render it properly everywhere that item appears. Prioritise cited records rather than attempting to clean an entire lifetime library before a deadline.

For a finished manuscript with a hundred or more references, use a risk-based audit. Check every item for missing author, year, title, source and locator; then sample formatting by source type and scrutinise the fields the target style treats specially. A hundred journal articles do not require the same checks as a list mixing datasets, standards, chapters and websites.

• Refresh first. Resolve citation-manager warnings and confirm the bibliography is generated from the current document.
• Validate required fields. Sort or filter for blank dates, creators, container titles, DOIs and page/article numbers.
• Compare difficult examples. Use the target guide to check a many-author article, chapter, web page and any field-specific source.
• Choose where the fix belongs. Library correction for reusable truth, style correction for a rule, document correction only for a final one-off output.
• Freeze only a copy. If citations must be unlinked or converted to static text, preserve an editable version with live fields.

Fix the source of truth while a document is live. Correct a static output only when you understand that later library refreshes will not inherit the change.