asantool
Everyone has opened a PDF, tried to change one wrong digit in it, and found there was nothing to click. It is one of the most common complaints about the format, and the answer is not that PDF editors are badly made.
It is that a PDF is not a document. It is a description of a printed page. Almost everything else follows from that.
Where it came from
In 1991 John Warnock, then at Adobe, wrote a short internal paper about a problem nobody had solved: sending a page to someone else and having it arrive looking the same.
That sounds trivial now. It was not. A file opened on a different machine would find different fonts installed, a different printer driver, a different screen, and would reflow to fit — the layout you spent an afternoon on arrived somewhere else as something else. His proposal, the Camelot project, was to stop sending the document and start sending the result: a fixed description of the finished page, with the fonts carried inside it, that any machine could draw identically.
That became PDF, released with Acrobat 1.0 in 1993. It was Adobe’s proprietary format for fifteen years and became an open ISO standard in 2008, which is why every operating system can open one today without paying anyone.
Read that history and the format’s behaviour stops being annoying and starts being the point. PDF was designed so that nothing would move. Editing is the thing it was specifically built to prevent.
What is actually inside one
Open a Word document and the file contains ideas: this is a heading, this is a paragraph, this list has four items, this table has three columns. The structure is written down.
Open a PDF and the file contains instructions for drawing marks:
Put the glyph “W” in Helvetica at 11 points at coordinates (72, 690). Put “h” at (79.1, 690). Put “a” at (84.6, 690).
That is genuinely it. There are no paragraphs in a PDF, because nothing in the file ever said “paragraph”. There are no sentences, no headings and no columns. There are glyphs, each with a position, plus embedded fonts, images, and vector drawing commands.
When you open a PDF and see paragraphs, you are doing the assembling. Your eye groups glyphs that share a baseline into a line, and lines with consistent spacing into a block. The file did not tell you that. You inferred it, and so does any software that wants to work with the text.
Why the text runs together sometimes
Here is a small, concrete example of what “no structure” costs, from our own converter.
A PDF splits a line of text into separate runs wherever the font changes. The sentence “made between Acme and you”, with “Acme” in bold, is stored as three runs — and the file records no space between them, because a space is not a character the page needed to draw. The gap was made by moving the pen.
Our first version worked out whether a space belonged by counting characters and estimating a width. It was wrong often enough to produce “made betweenAcme” every time a bold word followed a normal one. The fix was to stop estimating: PDF records the exact advance width of every run, so the gap can be measured rather than guessed.
That is a single space, in one sentence, and it took a measurement to get right. Scale that up and you have the whole problem of reading a PDF.
The two kinds of PDF, and how to tell them apart
This is the single most useful thing to know about the format, because it decides what is possible.
A born-digital PDF was exported from something — Word, Pages, a browser, a design tool. It contains real glyph instructions. The text is in there.
A scanned PDF is a photograph of paper in a PDF wrapper. It contains one big image per page and no text at all. What looks like writing is pixels arranged to look like writing.
The test takes two seconds:
Open the file and try to drag-select a word. If it highlights, you have a text layer and most things are possible: search, copy, convert to Word, reflow. If nothing highlights, it is a scan, and the only way to get text out is optical character recognition — software that looks at the picture and guesses the letters.
That is also why a scanned PDF is enormous. A page of text stored as text is a few kilobytes. The same page stored as a photograph is a few megabytes, because it is a photograph.
So what can you actually do with one?
Quite a lot, as long as you are not trying to retype the middle of a sentence.
Put things on top of it. Signatures, form answers, notes, highlights, dates. This is what nearly everyone means by “editing a PDF” and it works perfectly, because adding marks to a page is exactly what the format is for.
Rearrange the pages. Merge, split, delete, reorder, rotate. Pages are self-contained, so all of this is lossless — nothing is re-encoded and the text stays text.
Convert it. To Word, to images, or from images and Word into PDF. Useful, and imperfect in the direction of PDF-to-anything for the reasons above.
Make it smaller. Usually by re-compressing the images inside it, which is why compression does a great deal for a scan and very little for a page of text that is already a few kilobytes.
What you cannot honestly do in a browser is rewrite the text already in the document. That needs to re-subset the embedded fonts and reflow everything around the change, which requires the full document model — it is why the tools that manage it are desktop applications with a price attached. Any free web tool claiming to do it is almost always doing the first thing on this list and calling it editing.
Three things that confuse people
Fillable forms. Some PDFs have real, clickable fields — an AcroForm. Most of what gets emailed around does not; it is a flat page that merely looks like a form. That is why you often cannot click into the blanks, and why typing on top of the page is the reliable way to fill one in.
PDF/A. A stricter version of the format for long-term archiving. It forbids anything that might not render identically in fifty years — external font references, encryption, embedded video. If an institution asks for PDF/A, they are asking for a file that will still open in 2076.
Password protection. PDF supports two different things with one name: a password to open the file, and permission flags that ask readers politely not to print or copy. The first is real encryption. The second is a request that any determined piece of software can ignore, so do not treat it as security.
The short version
A PDF is a finished page, not an editable document. It was built in 1993 to stop things moving, it succeeded, and the frustration people have with it is the direct consequence of it working as designed.
Once you know which kind you are holding — text or photograph — almost every question about what you can do with it answers itself.
Every tool linked here runs in your browser. The file is read from your disk and written back to it, and never uploaded, which for a contract or a bank statement is the part that matters most.
