Why editing a PDF is so hard

"Just edit the PDF" is a sentence that has ruined a lot of afternoons. The file opens, the text is right there, you can see the typo. And yet every attempt to change it involves some tool doing something odd: covering the old text with a patch, converting the file into something else, or asking for money. It's reasonable to blame the software, but mostly it is coping with a format that was designed, on purpose and for good reasons, not to be edited.

Digital paper

PDF's founding goal was to replace the printed page for exchanging documents: the same page on every machine, indefinitely. To keep that promise the format stores the result of a document rather than its source. Your word processor knows your document as paragraphs, styles and flowing text. The PDF it exports knows only where each glyph landed. So a PDF is closest to a very precise printout, and editing one raises the same questions as editing paper with a razor blade and correction fluid: you can remove marks and add marks, but the page will not re-arrange itself around your change, because the rules that arranged it were left behind in the word processor.

What a PDF doesn't store

Three absences do most of the damage. First, sentences. The file stores runs of glyphs at coordinates; which glyphs form a word, which lines form a paragraph, what order the pieces read in: none of it is recorded unless the file was tagged for accessibility, and even then the tags describe the marks rather than controlling them. An editor that wants to let you click into a sentence must first reconstruct the sentence from the geometry, the same guesswork that makes copied text come out scrambled. It guesses well on a single column of body text and badly on tables, footnotes and anything in two columns.

Second, reflow. In a word processor, adding one word nudges every following line, across pages if need be. A PDF has no machinery for that: every line is independently nailed down. Add a word to a line and the line is now too long, and nothing else will move to make room. Adding a sentence to the middle of a paragraph is the request no PDF editor handles well.

Third, complete fonts. Type a letter the original document never used and there's a fair chance the font inside the file doesn't contain it, because most exporters embed a subset: only the characters actually on the pages, often with the character-mapping table stripped out. Editors respond by substituting a lookalike font, which is why edited text so often sits slightly wrong against its neighbours, one weight off, a hair too wide. The clean fix is supplying the real font file, which is what our editor asks for when it matters.

What editors actually do

Given all that, every PDF editor on the market is some mix of four strategies (two of them the same trick at different levels of care), and knowing which one a tool is using tells you what it will be good at.

Draw on top. Leave the page alone and add a layer: highlights, notes, a signature, a white box with new text over the old. Fast and safe, and the original is untouched underneath, which suits markup and rules it out for redaction. Most free "editors" do nothing beyond this.

Cover and re-set. Sample the page colour, cover the old line, re-type it in a matching font. This is drawing on top with craft: done well, the seam is invisible. It's how our editor changes existing text. The original characters stay in the file unless you tick "Remove what's underneath", which rebuilds the page without them.

Edit the source for real. Parse the content stream, modify the drawing instructions, rewrap the surrounding lines. Acrobat and its desktop-class competitors do this within a recognised block of text, pushing the rest of the page down, with the font caveats above. It is the closest thing to true editing the format allows, though no tool that edits the PDF in place can carry text across a page boundary.

Convert, edit, re-export. Reconstruct a Word document from the PDF, edit it in an editor built for editing, export a fresh PDF. For heavy rewriting this beats everything else, at the price of the reconstruction being a guess: layout shifts, and complex pages come back different. PDF to Word is that first step, and its page states what survives the trip.

When to edit the PDF, and when to convert it

The practical rule falls out of the mechanics. For small, local changes like a typo, a date or a name, edit the PDF directly, with covers or real content edits; it's ten seconds of work. For real rewriting, new paragraphs and restructured sections, convert to Word, rewrite there, and export again. And when you still have the source document, don't edit the PDF at all. Change the original, export a fresh PDF, and let the word processor do the reflowing it was built for.

To see what the frozen page is actually made of, content streams and embedded fonts and the layers above the page, What's actually inside a PDF is the tour.