Search across many PDFs at once
Last reviewed: September 23, 2026 · Markdown version
Ctrl‑F in a PDF reader searches the file you have open. With forty files, that means opening them one at a time, searching each, and hoping you remember which one you were up to.
Choose a folder of PDFs below and this page searches all of them together. Each match comes back with the file it is in, the page number, and the sentence around it.
…or drag them here. Sub-folders are included when you choose a folder.
It keeps a pointer to the folder — not the files, not the text inside them, not the words you
search for — so a later visit can search it again without you finding it first. One click forgets
it. Chrome and Edge only.
Separate several words with spaces — a page counts as a match if it contains any one of them. Matching is on whole words and ignores capitals: run does not find running, and it does not guess at similar words.
English printed text only, read in your browser, slower. About 6 MB of OCR engine is fetched from
this site — never from anyone else — the first time it runs. Off, scanned pages are listed as not
searched.
The files never leave your browser. Nothing is uploaded and no network request carries them anywhere — the reading and the searching both happen on your own machine, in this page. The one thing this page can keep is a pointer to a folder you chose, and only if you tick the box asking it to: never the files, never the text inside them, never the words you searched for. One click forgets it.
A gap you can see beats a gap you cannot. These are listed so that "no matches" never quietly means "never read".
This page reads the text stored inside each PDF. A scanned page — a photograph of words — has nothing in that text to search. With the OCR box off it is reported above rather than counted as a page with no matches; with it on, the page is read by OCR in your browser and any match on it is labelled OCR.
What the results mean
"Matched" — the word is in the file's text
The file stores its words as text and one of yours is among them. The page number counts from the first page of the file, which is not always the number printed on the paper: the PDF standard lets a document label its pages in any style, and says "Page labels and page indices need not coincide" (PDF 32000-1:2008, the PDF standard as published by Adobe, §12.4.2, read 23 September 2026).
"No text layer" — the file is a scan
The pages are pictures. Nothing inside marks any of those pixels as letters, so there is nothing for a plain text search to match. Tick Also read scanned pages with OCR and this page reads those pages by OCR, in your browser, before searching them — English printed text only, and any match found that way is labelled OCR. For what OCR is and what it cannot read: what OCR is, and how to make a scanned PDF searchable. To check a single file on its own, use the searchable-PDF checker.
"Needs a password to open"
Under the PDF standard a reader first tries an encrypted file's default, empty password, and should "prompt for a password" only "If this authentication attempt fails" (same standard, §7.6.3.1). A file locked with a real user password therefore needs that password before anything in it can be read; this page cannot ask for one, so it lists the file as skipped. A file restricted only by an owner password — printing or copying limits — opens with no prompt, is searched here normally, and DocFind indexes it the same way.
The honest limits
- OCR is optional, English-only, and can misread. Off, scanned pages cannot be searched here and are named in the skipped list rather than passed over silently, because a scan that reports "no matches" is one of the most misleading answers a search can give. On, they are read by OCR in your browser — English printed text only — using the open-source Tesseract engine. Its own documentation warns that on a skewed page "The quality of Tesseract’s line segmentation reduces significantly", that it "works best on images which have a DPI of at least 300 dpi" (Improving the quality of the output, Tesseract documentation, read 23 September 2026), and that handwriting is not what it is for: "it won’t work very well, as Tesseract is designed for printed text" (FAQ, Tesseract documentation, read 23 September 2026). So a miss on an OCR page is not proof the word is absent. Every match found by OCR is labelled so you can tell.
- Whole words only. "run" will not find "running" and "running" will not find "run". It does not guess at similar words or related words. That is deliberate: a match you can trust is worth more than a match you have to check.
- A word broken across two lines with a hyphen is stored as two pieces and will not match as one word.
- A folder can be remembered; nothing else can. There is no library and no index here. In Chrome and Edge you can tick Remember this folder and the page keeps a pointer to it, so a later visit can search the same folder without you finding it again — it does not keep the files, their text, your searches, or any OCR. Every search still re-reads every file from scratch, OCR still runs from scratch every time, and one click forgets the folder. The browser's folder picker this relies on is listed as supported in Chrome and Edge and unsupported in Safari and Firefox (MDN Web Docs, the browser-compatibility table for the Window directory picker, read 23 September 2026), so the box is not shown in Safari or Firefox.
- How long it takes is your machine's business, not ours. A large set of long documents is real work, and it is done in this tab; you can stop it at any point.
- PDFs only. Word documents, spreadsheets and plain-text files in a folder are left out of the search.
When the folder stops fitting in a browser tab.
DocFind does this on a whole library instead of a page-load: it indexes your documents once — scanned pages included, read by OCR on your device once rather than on every visit — and searches them again any time without re-reading them, and tapping a result opens that PDF at that page with the word highlighted. It uploads nothing either: the index lives on the phone, and the iPhone app ships with no network capability at all. iPhone and Android.