Getting the words out without the layout
Sometimes the words are all you want. Quoting a clause in an email, pasting a paragraph into a document you are writing, feeding a report into a note-taking tool, counting how long something actually is — none of these need the page design, and fighting the formatting to get at the text is a familiar waste of time.
Extraction pulls the text content out as plain text. Copying from a PDF viewer often works, but on multi-column documents it interleaves the columns, and across page breaks it drags in headers and footers. Extracting handles the whole document in one pass and gives you something you can clean up in a text editor rather than page by page.
It is also the fastest way to search a document properly. Browser and reader search boxes are fine for a single term, but if you need to find every instance of five different terms, having the text in an editor is far quicker than repeating a search five times.
Why a scanned document produces nothing
This is the most common surprise, and the explanation is simple. A scanned PDF is a photograph of a page. There is no text in it — there are pixels arranged to look like text, and no amount of extraction will find characters that were never stored.
The test takes a second: open the PDF and try to select a line with your cursor. If a selection highlight appears over the words, there is real text and extraction will work. If your cursor draws a rectangle over the page instead, it is an image.
The fix is OCR, which examines the image and recognises the shapes as characters. It is a genuinely different operation with genuinely different reliability — good OCR on a clean scan is highly accurate, and OCR on a creased fax at an angle is not — but it is the only route from a picture of a page to text you can use.
What extraction loses, and how to compensate
Plain text has no fonts, no sizes, no bold, no italics. That means heading structure disappears, and a document whose meaning depended on typography — a form, a table of contents, a price list — can come out as a run of words whose relationships are no longer obvious.
Tables suffer most. A table is a visual arrangement, and flattened into text it becomes a sequence of cell contents whose row and column structure has to be reconstructed by eye. For anything tabular, extracting to a spreadsheet format will serve you far better than extracting to text.
Reading order can also surprise you on complex layouts. Text is extracted in the order it appears in the document's internal structure, which for a straightforward document matches reading order and for a magazine-style layout with sidebars and pull quotes may not. Skim the beginning of the output before trusting the rest.