You convert a PDF to Word expecting a document you can edit. Sometimes you get exactly that. Other times you get text boxes scattered across the page, broken columns and fonts that shifted everything half a line. The tool is often the same in both cases. The input was not.

PDF was never designed to be edited

This explains almost everything else. A PDF describes where marks appear on a page: this glyph at these coordinates, in this font, at this size. It is a description of an appearance, not a structured document.

Word works the other way round. It stores structure, headings, paragraphs, tables, lists, and decides appearance from that structure at display time.

Converting one to the other means reconstructing structure that the PDF never recorded. A converter looks at positions and spacing and infers that these lines are probably a paragraph, that this row of aligned text is probably a table. Good converters infer well. None of them are reading structure that is actually there.

Converting PDF to Word is not a translation. It is an educated reconstruction of intent from layout.

Two kinds of PDF, very different outcomes

Digitally generated PDFs were exported from a real document, from Word, from a design tool, from a report generator. The text exists as text, with real font information. These convert well, often almost perfectly.

Scanned PDFs are photographs of paper. There is no text at all, only an image of text. Converting one without optical character recognition gives you a Word file containing a picture. With OCR you get real text, plus whatever errors the recognition made, and essentially none of the original layout.

You can tell which you have in two seconds: try to select a line of text in the PDF viewer. If it highlights, it is digital. If nothing selects, it is a scan.

What survives and what does not

  • Usually fine. Body paragraphs, headings, simple bulleted lists, single column layouts.
  • Often needs a fix. Tables, especially without visible borders. Two column layouts. Footnotes.
  • Rarely survives. Complex magazine style layouts, text wrapped around images, heavy use of unusual fonts, forms.

Getting the best result

  1. Find the original. Obvious, frequently overlooked. If the PDF came from a Word file, asking the sender for that file beats any conversion.
  2. Check selectability first. Thirty seconds spent here tells you whether to expect a clean result or a rebuild.
  3. Convert the pages you need. Extracting three pages from a ninety page report is faster and produces a cleaner file than converting all of it.
  4. Expect a cleanup pass. Budget a few minutes for fixing tables and spacing. A converter that promises none is overpromising.

When you only need the text

Worth asking what you actually want. If you need the words rather than the document, copying the text straight out of the PDF and pasting it into a clean Word file often beats conversion. You lose the formatting, but you also lose all the broken formatting, and typing fresh headings takes less time than repairing inherited ones.

Keeping the file on your machine

Documents are the category where uploading matters most. Contracts, invoices, medical letters and anything with personal details in it are exactly the files that should not be handed to an unknown server for processing. Browser based conversion keeps the file local for the same reasons set out in the post on converting media without uploading. For documents, the argument is stronger still.