Scanned PDF vs Text PDF: Why One Compresses So Much Better
Two PDFs can look identical on screen and compress completely differently, because a PDF doesn't store "a page" — it stores either real text and vector instructions, or a photograph of a page. Those are very different amounts of data.
A text PDF stores instructions, not pixels
When a PDF is generated from a Word document, a webpage, or any "real" digital source, each page is a set of drawing instructions: "place the character A at this position, in this font, at this size." Text stored this way is tiny — a whole page of dense text is often a few kilobytes of actual instruction data, because letters and words repeat constantly and compress extremely well.
A scanned PDF stores a photograph of the page
A scanner or a phone camera doesn't know what a letter "A" is — it captures a grid of pixels representing however the page happened to look, shadows, paper texture and all. That page becomes one embedded image, typically hundreds of kilobytes to a few megabytes depending on scan resolution, and every one of those pixels has to be stored or compressed as image data — there's no shortcut, because the PDF has no idea there's text on it at all.
Why this changes what compression can do
Compressing a text PDF mostly can't shrink the text itself much further — it's already tiny and already compressed by the PDF format's built-in stream compression. Any size reduction on a text PDF usually comes from images or fonts embedded alongside the text, not the text itself.
Compressing a scanned PDF is fundamentally an image-compression problem: re-encoding each page's embedded photo at a lower quality or resolution, the same tradeoffs that apply to compressing any photograph apply here. This is also why scanned PDFs typically shrink dramatically under compression while already-small text PDFs barely change size — there's simply more image data to work with.
How to tell which kind you have
The simplest test: try to select and copy text from the PDF. If you can select individual words, it's a text PDF. If clicking and dragging just selects the whole page like an image (or does nothing), it's scanned — and running it through OCR (optical character recognition) is the only way to make its text genuinely searchable and selectable, compression won't add that capability.