← All posts

You probably don’t need Word: edit the PDF itself

There is a price on an invoice and it is wrong. Or a date in a contract that should say March. Three words, in a document somebody else made, and the only advice anyone gives you is: convert it to Word, fix it there, and export it back to PDF.

So you did that, and the file that came back is not your document any more. The line breaks moved. The logo is in the wrong place, or gone. The table has become a column of tab-stops. The page count changed, which means the page numbers in the footer now disagree with the page numbers on the pages.

You wanted to change three words and you have been handed a rebuilt document to proofread. The round trip is the problem, not your Word skills, and you can usually skip it entirely.

Why the round trip rearranges everything

A PDF does not contain paragraphs. It contains instructions: set this font at this size, move to this coordinate, draw these glyphs, move on. There is no object in the file saying "this is a paragraph", "these six lines are one sentence that wrapped", or "this is a table". A tagged PDF carries some of that, and most PDFs in the wild are not tagged.

So a PDF-to-Word converter has to reconstruct all of it from geometry. It groups glyphs that share a baseline into a line. It guesses that two lines belong to the same paragraph because the leading matches and the second is not indented. It guesses a heading because the type is larger. It guesses a table because some things line up vertically. Every one of those guesses is a place where your document can come out different, and the guesses are made on all four hundred lines, not only on the line you wanted to change.

Then there is the way back, which is worse and gets discussed less. Word does not reproduce a document, it lays one out. It breaks lines using the metrics of the fonts installed on the machine doing the export — and the fonts inside your PDF are not available to it. They are subsets: cut-down font programs carrying only the glyphs that document needed, embedded for drawing, not installable. Word substitutes something with different widths, every line rewraps a fraction differently, and the difference accumulates down the page until a paragraph falls onto the next one. The same subsetting explains why edited text changes typeface, and it is the reason a "lossless round trip" through a word processor is not a thing that exists.

Everything that was not text goes as well: links, bookmarks, form fields, annotations, any digital signature, and the document's own metadata. You changed three words and replaced the file.

What editing the PDF itself actually does

Editing in place works at the level the file is actually written in. The instruction that drew the old glyphs is removed from the page's content stream and replaced with a positioning adjustment of exactly the same width, so anything further along that line stays where it was. Your new words are drawn with the font the document already carries, at the same size, on the same baseline.

Nothing else in the file is touched. The other pages are byte for byte what they were, because nothing re-flowed and nothing was re-laid out — there was no layout engine involved at all. That is the whole difference: the converter rebuilds the document and hopes it lands in the same place, and an in-place edit changes one instruction.

Changing a word is trivial. Changing the shape of the document is the hard part, and it is a different job — see below for when that is the one you have.

What you can realistically change in place

Small, local, surgical things. Which is almost always what people actually want:

The procedure for the ordinary case, changing a wrong figure:

  1. Open the file in our PDF editor and click on the line with the mistake in it. A block that cannot be retyped says so when you hover, and the handful of reasons are all to do with how the text was drawn.
  2. Retype it. If the replacement is longer than what it replaced, the when the text does not fit control decides between shrinking it into the space, wrapping it, or letting it run on. Shrinking is the default for a block of several lines, because that is what holds the paragraph shape.
  3. Apply, download, and read the notes on the result.

Your new words are set in the typeface the document already carries, unless the copy of that typeface inside the file has no shape for a character you typed — in which case the whole line you edited comes back in a substitute, and the notes tell you so. Why the typeface changes, and how to check the line afterwards is a subject of its own.

When Word genuinely is the right tool

Reframing is not the same as pretending. Go through a word processor when the shape of the document has to change:

An in-place editor is not a word processor and ours does not pretend to be. Text does not flow from page to page, paragraphs do not renumber themselves, and a heavily rewritten document is better rebuilt than patched.

If you are going that way, be clear-eyed about what our converter is. PDF to Word produces an HTML document with a .doc extension — Word, LibreOffice and Google Docs all open it and one Save as makes it a real .docx. It gives you paragraphs and headings. It does not give you tables, images or columns, and it says so on every run. That is adequate for rewriting a document from its text and useless for reproducing one.

The answer nobody wants to hear first

If you have the file the PDF was made from — the Word document, the InDesign file, the spreadsheet — change it there and export again. Every route through the PDF is a repair. The source is the actual document, and re-exporting takes a minute against an hour of reconciliation.

Ask two questions and the decision makes itself. How much am I changing? A few words and numbers: edit the PDF. The structure: word processor. Do I have the original? If yes, and you are changing the structure, this is not a PDF problem at all.

What editing in place will not do

One thing to do after editing a form rather than before: once the answers are right, run the file through Flatten PDF. It draws the field values into the page and deletes the fields, so nobody — including you — can retype a figure afterwards. Keep the unflattened copy.

If you want to know what you are dealing with before any of this, our PDF inspector reads the page count, the page sizes, the encryption status and whether there is a real text layer, and it does it in your browser without uploading the file.

When a desktop tool is better

The browser route wins the common case: one figure to correct, a document you did not write, and no appetite for installing anything or explaining to anyone why the page numbers changed.

Tools this post uses: