PDF to text

Two different problems with the same symptom, and it's worth knowing which one you have.

If you get nothing at all, the PDF is a scan. It's pictures of pages, and there is no text inside it to extract. That needs recognition, which is a different job.

If you get text but the line breaks are wrong, that's normal and it's a consequence of how PDFs store things.

A PDF has no paragraphs

It stores text as positioned fragments: this string of characters, at these coordinates, in this font. There's no structure saying which fragments form a sentence, or where one paragraph ends and another begins.

So extraction reads the fragments and infers the layout from the coordinates. A noticeable vertical gap probably means a new paragraph. Fragments on the same line probably belong together. Usually right, sometimes not.

That's why extracted text often has line breaks at the end of every visual line rather than every paragraph. The tool can see where a line ended on the page; it can't always tell whether the sentence continued.

If you get one break per line, removing the empty lines or joining them in an editor is the quickest fix.

Columns come out interleaved

The commonest disappointment. A two column page reads across, so you get a line from the left column, then a line from the right, alternating down the page.

There's no reliable general fix, because working out that a page has columns means understanding the layout rather than reading it. For a page that matters, extracting each column separately is the practical answer.

Academic papers, newspapers and annual reports are all affected.

Other things that come out oddly

Tables lose their structure and become rows of words with the column relationship gone.

Ligatures. Some fonts store fi and fl as a single character, and if the PDF doesn't include the mapping back to plain letters, find comes out as nd or with an odd symbol.

Hyphenation. A word broken across two lines stays broken, so you get inter- and national separately.

Headers and footers appear on every page, in the middle of the flow, since they were on the page and extraction reads the page.

Reading order follows the order the text was written into the file, which is usually but not always the order it appears on the page. Documents produced by unusual software can come out shuffled.

Telling a scan from a real document

Open it in any reader and try to select a line of text. If a selection highlight appears, there's text in there. If your cursor draws a box instead, it's a picture.

Searching is the other test. If Ctrl+F finds a word you can plainly see, the text exists.

Some PDFs are both: a scan with recognised text stored invisibly behind the image. Those extract fine, and the quality depends entirely on how good the original recognition was.

When you need the layout as well

If you want the words and the arrangement, plain text is the wrong target. Converting to a Word document keeps some structure, though it produces loose text boxes rather than a properly flowing document, which is why this site doesn't offer that conversion. Nothing browser-based does it well.

For quoting, searching or feeding text into something else, plain text is exactly right and this is the fastest route to it.

Nothing is uploaded

The extraction happens in your browser. That matters here more than on most tools, because the PDFs people pull text out of are contracts, statements, reports and academic papers, and plenty of them aren't meant to leave the machine they're on.