PDF is the format people use without ever being told what it is. Knowing what it was built to do explains almost everything that is otherwise puzzling about it: why you cannot easily edit one, why text sometimes will not select, and why a Word file converted to PDF does not always look like the Word file.
The problem it was built to solve
In 1993 sending someone a document was unreliable. A Word file opened on another machine could reflow completely, with different fonts substituted, different margins, page breaks in different places, and a two-page document arriving as three.
The cause was that word processor files describe a document semantically (“this is a heading, this is a paragraph, use Garamond at 11 point”) and leave the rendering to whatever software opens it. If that software lacks Garamond, it picks something else, and everything shifts.
PDF’s answer was to describe the page geometrically instead. Rather than “a heading in Garamond”, a PDF says “draw this glyph at this exact coordinate, at this exact size, in this exact colour”, and it can embed the font itself so there is nothing to substitute.
The result is a file that renders identically on any machine, any operating system, any year. That is the entire point of PDF, and every other property follows from it.
What is inside one
A PDF contains, roughly:
- A page tree defining each page and its dimensions.
- Content streams, which are sequences of drawing operators: move here, draw this text, place this image, stroke this line.
- Embedded resources: fonts (usually subsetted to only the glyphs used), images and colour profiles.
- A cross-reference table that lets a reader jump straight to any object without parsing the whole file, which is why a 500-page PDF opens instantly.
- Metadata: title, author, creation date and producing software.
Page dimensions are in points, at 72 to the inch. US Letter is 612 × 792 points, and A4 is 595.28 × 841.89. Those are the numbers behind every page-size option in a converter.
Why editing a PDF is awkward
Because a PDF stores positioned glyphs and no paragraphs, there is often no paragraph to edit. The file may know that there is an “e” at a particular coordinate. It does not necessarily know that the “e” belongs to a word, that the word belongs to a sentence, or that the sentence should reflow if you delete something before it.
This is why PDF editors either do limited local patching or reconstruct a document model by guessing at the structure, with results that vary a great deal by file.
The reliable PDF operations are structural and leave the text alone: merging files, splitting them, reordering, rotating and deleting pages. Those manipulate whole page objects and always work cleanly. That is exactly the set of operations this site offers (merge, split and organize), because they are the ones a tool can do without guessing.
The right workflow is to edit the source document and re-export.
Text, images, and the difference that matters
A PDF page can contain real text, an image, or both, and they look identical on screen while behaving completely differently.
Text-based PDFs were generated from a digital source. The text is selectable, searchable and copyable, and accessible to a screen reader.
Image-based PDFs came from a scanner or a camera. The page is a photograph of text. Nothing is selectable or searchable, and a screen reader gets nothing.
Turning the second into the first is OCR (optical character recognition), a distinct and much harder operation than conversion. Converting an image to PDF wraps the picture in a page and does not read it. If you need searchable text from a scan, you need an OCR tool, and no format conversion will substitute.
PDF/A, and why an archive rejected your file
There are several PDF variants, and one occasionally becomes someone’s problem:
- PDF/A is for long-term archiving. It forbids anything whose rendering could drift (external font references, JavaScript, encryption) and requires all fonts embedded. Government archives, courts and academic repositories often require it.
- PDF/X is for print production, requiring embedded colour profiles and defined trim boxes.
- PDF/UA is for accessibility, requiring proper tagging and reading order.
A general converter produces standard PDF. If a submission system demands PDF/A, that is a specific export setting, usually available in the originating application.
Fonts, and the two ways a PDF can go wrong
Fonts are where PDFs fail in the wild.
Embedded fonts are the normal case: the file carries the glyphs it needs and looks the same anywhere.
Referenced fonts are the failure: the file names a font and assumes the reader has it. Open it on a machine without that font and the reader substitutes, and your layout shifts, which is the exact problem PDF was invented to prevent.
Most modern tools embed by default. If a PDF looks wrong on someone else’s machine, unembedded fonts are the first thing to check.
Fonts are usually subsetted, so only the glyphs actually used are included. This keeps files small and is why adding one unusual character can noticeably change file size.
Why file size varies so much
- Text is tiny. A hundred pages of text can be well under a megabyte.
- Images dominate. One high-resolution photograph can outweigh the entire text of a book.
- Scans are the worst case. Every page is a full-page image, so a 20-page scanned contract can be 50 MB while a 20-page typed one is 200 KB.
- Embedded fonts add 50-500 KB each, subsetting notwithstanding.
If a PDF is unexpectedly huge, it is almost always images, and almost always scans.
What this means in practice
- Converting to PDF fixes the appearance in place. That is the benefit and the cost: it is now a page, not a document.
- Edit the source and re-export. Don’t try to edit the PDF.
- Page operations are safe, and text operations are not.
- A scan is a picture. Searchable text needs OCR, which is a different tool.
- Embed your fonts, which any competent generator does by default.
Every conversion route on this site produces standard PDF with fonts embedded, built in your browser by pdf-lib, and nothing is uploaded. Start from the job board, or read why Word documents do not convert perfectly.