Zotero Full-Text Search: Find Any Word Inside Your PDFs

Zotero full-text search works on the contents of your PDFs, but only in two places: the quick search box set to "Everything", and an Advanced Search using the "Attachment Content" condition. The default quick search mode does not look inside files at all, which is why so many people with a thousand papers in Zotero conclude the software cannot read them. It can. This guide covers the three search modes and what each one actually matches, how the full-text index is built and where it quietly stops, the diagnostic order to run when a phrase you can see on screen returns nothing, and what changed in Zotero 10.

The three quick search modes

The search box at the top-right of the center pane has three modes, and you switch between them by clicking the magnifying glass icon inside the box. Per Zotero's own documentation, they match different things:

• "Title, Year, Creator" matches against those three fields, plus publication titles.
• "All Fields & Tags" matches against all fields, as well as tags and text in notes.
• "Everything" matches against all fields, tags, text in notes, and indexed text in PDFs.

Only the third one reaches into your files. If you have never touched this setting, a search for a phrase that appears on page 14 of a paper will return nothing, because Zotero is comparing your phrase against titles and author names.

The mode is sticky. Set it once to "Everything" and it stays there, which is what most people want. The cost is speed: "Everything" has more to compare against, and in a large library the search runs on every keystroke. Zotero's documented fix is a keyboard trick worth memorizing. Type a quotation mark " at the very beginning of the search field, and the search waits until you press Enter instead of running as you type. The quotation mark is not part of your query; it is a signal to defer.

Attachment Content, and why you want it

Quick search answers "does this phrase appear anywhere." Advanced Search answers narrower questions, and it is where the full-text index gets its more precise interface. Open it with the magnifying glass button in the quick-search bar, then choose Attachment Content as the condition. That is the same index "Everything" uses, exposed as a field you can combine with others.

The combination is the point. Attachment Content contains "endogeneity" AND Date is after 2020 AND Collection is "Dissertation ch. 3" is a question quick search cannot express. A few mechanics from the documentation make these searches behave:

• Match "all" or "any". Items appear only if they satisfy every criterion by default. Switch Match to "any" for an OR.
• Search subcollections pulls in items filed below the collection you named.
• Include parent and child items of matching items is the one people miss. Your PDF is a child attachment of the reference item. When the text matches the child and your other condition matches the parent, this option is what lets both halves count toward one result.
• Wildcards. The % character substitutes for zero or more characters, so W% Shakespeare matches "W Shakespeare", "W. Shakespeare" and "William Shakespeare". A lone % matches any field that has content at all, which is a fast way to find records missing a DOI.

Save any of these with the Save Search button and it becomes a saved search in your left pane, with its own icon. Saved searches store the criteria and not the results, and they update continuously, so a saved search for Attachment Content containing your dissertation's key term keeps catching new PDFs as you add them.

What Zotero 10 changed

Zotero 10 arrived on August 17, 2026, and search got the largest share of the work. This matters for a practical reason: the documentation pages that rank highest for these questions were last updated in June 2026, so two things they say are now out of date. From the Zotero version history:

• Advanced search was redesigned. It opens directly above the items list from a button in the quick-search bar, and it can filter any view, including collections and saved searches.
• Nested condition groups are supported. The older documented method for a Boolean query like (a OR b) AND (c OR d) was to build two saved searches and then search for items belonging to both. That workaround is still described on the Searching page. In Zotero 10 you can nest the groups directly.
• You choose a result level: top-level items, attachments, annotations, or notes, with conditions automatically mapped across levels. For full-text work this is the standout. Searching thousands of PDFs for a phrase and getting back the annotations rather than the parent items is now a setting.
• Terms typed in the quick-search bar automatically populate advanced search with the equivalent conditions, so you can start loose and tighten without retyping.
• New conditions include Annotation Type, Color and Author, Attachment Storage Type, counts of annotations, attachments, notes and tags, and new "is empty" / "is not empty" operators.
• Searches now ignore accents and typographic characters, so "cafe" matches "café", and a word typed with a straight apostrophe matches the same word set with the curly one that word processors insert. Anyone who has lost twenty minutes to a smart quote will appreciate this one.
• Full-text content searches are much faster, and indexing of new attachment content is faster and more reliable.
• Multiple collections, saved searches, or even libraries can be selected at once in the collections pane and viewed together. The Searching page still says it is not currently possible to search in multiple libraries at one time. In Zotero 10, it is.

Zotero 10.0.1, released a week later on August 24, fixed mangled full-text index statistics after searching in settings, which is worth knowing if you looked at your index numbers in late August and found them implausible.

How the index actually works

Zotero does not read your PDFs at search time. It builds a full-text index in advance, and your search runs against that index. Indexing happens automatically in the background when Zotero is idle, which has a consequence worth stating plainly: a PDF you added four minutes ago may not be findable yet, especially if you have been working continuously since.

Three limits shape what ends up in that index.

File type. Only PDF, EPUB and HTML content plus plain text files can be indexed. Word documents and OpenDocument files cannot. If your library is half .docx, half of your library is invisible to full-text search and no amount of reindexing will change that.

Character and page caps. The Search pane of Zotero settings has a maximum characters to index per file, set by default to 500,000, which Zotero describes as roughly 100,000 words or 180 to 200 pages of content. The Searching page states the shipped defaults as 500,000 characters and 100 pages. For most journal articles neither cap is close to binding. For a 600-page monograph, the back half of the book is simply not in the index, and a search for a term that appears only in the final chapter will come up empty while the book sits right there in your library. Setting the character value to 0 disables full-text indexing entirely.

Whether the file contains text at all. This is the big one, and it gets its own section below.

You can check any individual file. Select the attachment item in your library and look at the Indexed: field in the right pane. For the library as a whole, the Search pane of settings reports four numbers: how many files are completely indexed, how many are partial, how many are not yet indexed, and the total word count of the index. A large Partial count usually means you are hitting the page cap on long documents. A large Unindexed count means either Zotero has not caught up yet or those files have no extractable text.

When the search finds nothing

Work through this in order. Each step is cheap and rules out a whole class of cause.

1. Check the mode. Is quick search set to "Everything"? This is the cause maybe half the time, and it costs one click to eliminate.
2. Check that one file's Indexed field. Select the PDF attachment, read the right pane. If it says the file is not indexed, the problem is the index, not your query.
3. Try to copy the text out of the PDF. Open the file, select the phrase you searched for, copy it, paste it somewhere. Zotero's documentation recommends exactly this test, and it is decisive. If you cannot copy the text, the page is an image, and no search tool will find words in it until the file is put through OCR.
4. Reindex the one item. Right-click the attachment and choose Reindex Item.
5. Rebuild the whole index from the Search pane of settings, using Rebuild Index. Zotero names two situations where this is the right move: after running OCR on a large number of files, and when "Everything" or Attachment Content is returning results you know are wrong. On a large library this takes a while, so it is a step 5 and not a step 1.
6. Raise the caps if the term lives deep inside long books, then rebuild. The index grows, though Zotero notes it usually occupies a relatively small amount of space.

If a file has valid, copyable text and still refuses to index after a rebuild, that is the point to ask on the Zotero forums rather than keep rebuilding.

Scanned PDFs need OCR, and Zotero does not do it

A scanned chapter is a photograph of a page. There are no characters in the file, only pixels arranged to look like characters, and Zotero has nothing to index. Optical character recognition is the process that reads those pixels and writes real text back into the PDF.

Zotero has no built-in OCR, so this happens before the file reaches your library, or through a plugin. Cornell University Library's Zotero plugins guide covers the options: OCR turns PDFs into machine-readable documents that can be highlighted and copied, Apple computers now have built-in OCR for downloaded PDFs, and the zotero-ocr plugin runs Tesseract from inside Zotero, with setup the guide describes as more complex than other plugins. Illinois State University's Milner Library guide makes the related point that a readable, OCR-enabled PDF is also what lets Zotero pull metadata out of a document to build the record in the first place, and suggests converting static PDFs with a tool such as Adobe Acrobat Pro or Foxit PDF Editor.

The order matters. OCR the files first, then run Rebuild Index. Doing it the other way round indexes the empty versions and tells you nothing has changed.

Searching a library of several thousand

Large libraries hit different problems than small ones, and most of those problems live in the index.

Add the quotation mark prefix to defer searching until you press Enter. Keep Zotero open and idle occasionally, since that is when indexing runs; a laptop that is closed the moment work ends never gives the indexer a window. Watch the Partial number in your index statistics, because that is where long PDFs quietly truncate. Zotero 10's multi-library selection removes the old constraint of one library at a time, which matters most when you routinely search a shared group library alongside your own.

Sync is worth understanding here, because it surprises people. Your library data syncs separately from your attachment files, and the full-text index lives locally. A second machine that has synced your records but not your PDFs has nothing to search inside. Our guide to using Zotero covers the split between data sync and file sync, and the storage quotas that decide how many PDFs actually travel with you.

One honest limit: Zotero's full-text search is a word and phrase search. It is not a semantic search, and it will not find "unemployment" when your PDF says "joblessness". If your workflow depends on concept-level retrieval across a large corpus, that is a plugin or an external tool, and our comparison of Zotero alternatives covers managers that approach search differently.

Common mistakes and how to avoid them

• Leaving quick search in the default mode. The mode persists once changed, so this is a one-time fix that resolves the most common complaint.
• Judging the index by one file. A single scanned PDF proves nothing about the library. Read the four index statistics in settings instead.
• Rebuilding the index before checking whether the text is copyable. A rebuild on a scanned file produces the same empty result, slower.
• Expecting .docx files to be searchable. They will never be indexed. Convert to PDF if the contents need to be findable.
• Assuming a long book is fully indexed. Page and character caps truncate it silently, with no warning on the item.
• Searching immediately after a bulk import. Give the indexer idle time, then check the Unindexed count before concluding anything is broken.
• Mistaking a library problem for a search problem. Items that have vanished from the pane entirely need a different diagnosis; see recovering a lost Zotero library.

Quick-reference table