A PDF describes a page, not a document
This is the one idea that explains almost everything people find surprising about PDFs.
A Word document stores a document: paragraphs, styles, a flow of text that rearranges when you change the margins. A PDF stores a page description: put this glyph at this coordinate, in this font, at this size; draw this line here; place this image there.
Adobe designed it in 1993 for a specific problem — a document that looked identical on every machine, whatever fonts or software were installed. Describing exact positions rather than intent solves that completely.
It also creates every limitation below.
Why editing is awkward
There are no paragraphs to reflow, because the PDF does not know it has paragraphs. It knows where each character was placed.
Insert a sentence and something must recompute the layout — which means reconstructing intent the file never recorded. PDF editors do this by inference, which is why the results are variable and why editing the original and re-exporting is nearly always better.
Why some PDFs have no text at all
A scanned PDF is a photograph of a page wrapped in a PDF container. The file holds an image; the letters exist only as arrangements of pixels, exactly as they would in a photograph of a book.
The two-second test: open it and try to select a word. If individual words highlight, there is a text layer, and extracting the text will work. If you can only drag a rectangle over the page, it is a scan, and no amount of extraction will find words that are not there — that needs optical character recognition.
Why PDFs get large
Embedded images, almost always. A PDF assembled from ten phone photographs is roughly the size of those ten photographs, because the images are stored inside it.
Embedded fonts add a smaller, fixed cost — and explain something counterintuitive: splitting a PDF into single pages often produces files that total more than the original, because each page now needs its own copy of the fonts it uses.
What this site can and cannot do with PDFs
Honestly stated, because the gap matters:
Can — merge, split, extract pages, delete pages, rotate, compress, convert pages to images, extract text, and build a PDF from images.
Cannot — recompress a PDF’s images while keeping its text selectable, edit text, fill forms, sign, or run OCR on a scan. Compressing works by rasterising, which is ideal for scans and lossy for text documents. How to make a PDF smaller sets out what actually works with what is here.
All of the above run in your browser using pdf-lib and PDF.js. Nothing is uploaded, which matters more for PDFs than for any other format on this site — they are the files most likely to contain a signature, a bank balance or a medical result.