PDF compression explained: why your file is big and what actually shrinks it

You compress a PDF, and sometimes it shrinks by 90%. Sometimes it barely moves. That inconsistency isn't a bug — it's because PDFs aren't one kind of file, they're a container that can hold very different kinds of content, and each kind compresses differently.

What's actually taking up the space

• Scanned pages — each page is really just a photograph, often saved at a much higher resolution than needed for screen reading. This is almost always the biggest opportunity for savings.
• Embedded fonts — a PDF that embeds a full font file for every typeface used can carry real weight, especially with several fonts in one document.
• Vector graphics and text — actual selectable text and line-drawn graphics are usually tiny; a 40-page text-only report might be a few hundred kilobytes.
• Duplicate objects — the same logo or background image repeated on every page, stored inefficiently, adds up fast in badly-generated PDFs.

Why compression tools re-render pages as images

A properly built PDF compressor would go object by object — recompress this embedded JPEG, subset that font to only the characters actually used, remove duplicate resources — the way desktop software like Acrobat does. Most in-browser tools, including this one, take a different approach: they render each page as an image and re-encode it at a lower quality, then rebuild the PDF from those images.

This trades some sharpness — especially on small text — for a much smaller, much simpler operation that a browser can actually do quickly. For scanned documents, this is close to what a proper compressor would do anyway, since scanned pages are already images. For text-heavy documents, it's a blunter tool: you'll get a smaller file, but selectable text becomes part of a flattened image, which also means it stops being searchable or copyable.

Picking a compression level

If the document is a scan (a contract someone signed and photographed, for instance), a higher compression level is usually safe — you're not losing much you didn't already lose when it was scanned. If the document has real, sharp text you care about keeping crisp and searchable, start at the lowest compression setting and only increase it if the file is still too large.