How to extract text from a PDF, and why some PDFs give you nothing

Two PDFs can look identical in a viewer and behave completely differently when you try to get the words out. One hands over every sentence. The other hands over nothing at all, and no setting will change that. The difference is invisible on screen, so it is worth understanding before you blame the tool.

Text layer and OCR, defined

A text layer is the set of characters a PDF stores as text, each with a position on the page. Selecting a word, searching with Ctrl+F or Cmd+F and extracting text all read the text layer and nothing else. A PDF exported from a word processor or a browser has one, because the software knew what every character was when it wrote the file.

OCR, optical character recognition, is the separate job of working out which characters appear in a picture of a page. A scanner or a phone camera produces a picture, so a scanned PDF has pages but no text layer, and extraction returns nothing until something has done OCR. This tool does not do OCR. Recognising glyphs from pixels is a different operation with a much larger download, and the tool’s page lists it as something it does not do.

The quickest test needs no tool at all: open the PDF and try to select a single word. If you cannot, the page is almost certainly an image.

PDF to text, step by step

1. Open PDF to text and choose Add a PDF, or drop it onto the page. Each page appears as a thumbnail. A password-protected file asks for its password first.
2. Press the PDF to text button. There are no settings to choose.
3. Read the first page of the output in the box before anything else. It tells you at once whether the document has a text layer and whether its order matches reading order. Then Copy the text, or Download it as document.txt.

What the output looks like

The tool reads each page’s text in the order the file lists it, joins the pieces with single spaces, and starts the next page after a blank line. Line breaks inside a page are not kept, so every page arrives as one long paragraph. That is a good shape for searching, quoting or pasting into something that will reflow it, and a poor one for anything where the line breaks carry meaning, such as an address block.

There is no page limit. A 300-page report is the same operation as a receipt; the only difference is how long your own processor takes.

Why some PDFs give you nothing, and others a jumble

What you add	What you get	Why
A document exported from a word processor or browser	The text, one block per page	It has a text layer
A two-column paper with a running header	Every word, with the columns partly interleaved	Extraction follows the file’s internal order, not the visual one
A scanned document	Nothing	Each page is an image with no text layer, and the tool does no OCR
A PDF made from photos with Images to PDF	Nothing	That tool draws pictures on pages and adds no text
A PDF shrunk with the Strong or Smallest setting of Compress PDF	Nothing	Both render each page to a JPEG and re-embed it, so the text layer is gone
A password-protected PDF	A request for the password, then the text	The content is encrypted until the file is opened

The scan is the case people hit. Nothing is wrong with the file and nothing is wrong with the extractor: there is no text in it, only pictures of text. When no page has any text at all, the tool says so: it reports that no text was found, names a scan as the likely cause and says it does no OCR. Until October 2026 it returned an empty box with no explanation, which is exactly the confusion this guide exists to clear up. The next two rows are the same failure made on purpose. A document that has been through a rasterising compressor is, as far as any extractor can tell, a scan, which is the trade described in our measurement of lossless PDF compression.

Interleaving is the subtler problem. Reading order is the sequence a person reads the words in, and a PDF does not record it. It stores glyphs with coordinates, and the paragraph you see is something your eye assembles. Most producers happen to list text in reading order, so most files come out fine. In a two-column layout the columns often interleave, and text in headers, footers or floating boxes can turn up anywhere in the stream. There is no reliable general fix without layout analysis, and no browser-side library implements it.

If the PDF is a scan

You need OCR, and this site does not offer it. If the OCR software you use wants images rather than a PDF, render the pages with PDF to JPG at 300 dpi, which is the resolution that tool recommends for OCR.

Text, not a Word document

The same reasoning is why there is no “PDF to Word” here. Recovering columns, tables and headings means inferring structure that was never recorded. Every free tool that claims to do it returns a flat run of text with the tables destroyed, which is the same text with a misleading file extension. This tool gives you the text and says what it is.

Other ways to do it

There is no built-in desktop method on the list we have checked. On the command line, poppler’s pdftotext does the job:

pdftotext -layout input.pdf output.txt

It reads the same text layer, so a scan gives it nothing either.

Checking that nothing was uploaded

Add the PDF and wait for the thumbnails, which is when the page fetches the code that reads PDFs. Then disconnect from the network and press the button. The extraction still runs, because it happens in the tab.

Extract the text with PDF to text. If you want the page breaks marked, PDF to Markdown runs the same extraction and puts a heading before each page.