People come to a PDF expecting a document and find something that behaves like a photograph of one. Everything that follows — why you cannot fix a typo, why rearranging pages is instant, why the watermark somebody added is still there — comes from that one gap.
The tool Merge PDF Combine PDFs into one file. Drag to reorder before you save.A PDF is a description of a printed page
A Word file stores a document: paragraphs, headings, a structure that knows one thing follows another. A PDF stores instructions for putting marks on a page — this glyph at these coordinates, in this font, at this size. Then the next one.
It was designed so that a page looks identical everywhere, on any machine, forever. It succeeds at that completely, and the price is that there are no paragraphs in there to edit. There is a sheet of positioned marks.
Once you hold that picture, every PDF operation sorts itself into three tiers, and which tier a job falls into tells you immediately whether it will take ten seconds or ruin your afternoon.
Tier one: moving whole pages
A page is a sealed object. Nothing has to be understood to move one, so all of these are effectively free and none of them touches what is printed:
- Joining files — the pages of one are copied into another, as they are
- Splitting, and removing pages — the ones you keep are copied out untouched
- Reordering — the same objects, listed in a different order
- Rotating — a single number stored against the page saying which way up it is
That last one is worth a sentence of its own, because it is the source of a common and maddening problem. Rotating in a viewer usually changes only what you see; the file goes out the way it came in, and the next person gets it sideways again. Rotation that sticks is written into the page itself.
None of these re-renders anything, which is why the text in the result is still selectable, still searchable, and exactly as sharp as it was. If a tool hands back a merged document whose text has become an image, it re-printed your pages instead of copying them.
Tier two: drawing on top
Adding something new to a page is also cheap, because it is genuinely just more instructions appended after the existing ones.
Page numbers and watermarks both work this way: the original marks stay exactly where they were and new ones are drawn over the top. Nothing underneath is decoded, altered or re-compressed.
Two consequences follow, and both surprise people.
A watermark is a label, not a lock. It sits on top of the page like a rubber stamp. Anyone determined can remove it, and the document underneath is entirely intact. It announces "this is a draft" to an honest reader. It protects nothing from a dishonest one.
A black box over a name is not redaction. This one has ended careers. Drawing a filled rectangle over text adds a rectangle; the text is still in the file, underneath, and selecting or copying that region gets it back. Real redaction removes the underlying marks. If a document has to be safe to publish, the words have to come out — covering them is theatre.
Tier three: changing what the page says
Here is the hard one, and the reason "PDF editor" software is a category at all.
To change a word you have to work out which marks form that word, remove them, and lay down new ones — matching the font, the size, the spacing, and the position of everything after it on the line. There is no paragraph to re-flow, because there are no paragraphs.
Serious editors do attempt this, and they do a decent job on simple layouts. Two things routinely defeat them:
- Fonts are usually embedded in part. To keep files small, a PDF often carries only the glyphs the document actually used. If the letter you want to type was never on the page, it may simply not be in the file — which is why edited text sometimes comes out in a visibly different typeface.
- There may be no text at all. A scan is a picture of paper. There is nothing to edit until the words have been recognised, and the ten-second check that tells the two kinds apart is the same one that decides whether compressing will help.
So the honest order for changing wording, best first:
- Find the original. The Word or Pages document it was made from. Edit that, export again. Everything else is a workaround for not having it.
- Ask whoever sent it. Faster than any software, and it is the answer nobody tries first.
- Use a real editor, and check every page afterwards — particularly the line the edit sat on.
- Recognise the text first if it is a scan, then work on what comes out. Extracting the words is often all anybody actually needed.
The jobs people ask for, sorted
| What you want | Which tier | How it goes |
|---|---|---|
| Combine several files | Moving pages | Seconds, nothing altered |
| Take out a page | Moving pages | Seconds — but not the same as redacting it |
| Fix a sideways scan | Moving pages | Seconds, if the rotation is stored rather than shown |
| Number the pages | Drawing on top | Seconds, original untouched |
| Stamp DRAFT across it | Drawing on top | Seconds, and it is a label rather than protection |
| Get the file smaller | Depends on the kind | Check which kind you have first |
| Correct a typo | Changing the page | Find the original. Genuinely. |
| Remove a name safely | Changing the page | Needs real redaction, not a black box |
One more thing worth knowing
Everything in the first two tiers can be done without the file leaving your computer, because none of it requires much: copying page objects, storing a number, appending some instructions. That is the whole reason these particular jobs are worth doing in a browser rather than uploading a document to somebody else's server.
And these are exactly the documents you would least like to upload. Contracts, statements, medical letters, applications with your address on them. The edit is trivial; the file is not.