Compress PDF

Yes. The tool rewrites the internal structure of the file: it packs the objects into object streams, compresses streams that were stored uncompressed and rebuilds the cross reference tables. Images, fonts and text are not re-encoded, so nothing about the visible page content changes.

Because it was already stored efficiently. A PDF can only shrink if it still has slack to remove, such as uncompressed streams, duplicate objects or a loose cross reference table. Many files, especially small ones written by modern apps, have none of that, and the object streams plus linearization add a few hundred bytes of their own overhead. In that case the output is slightly larger but still a valid document, and the original remains the better copy to keep.

Yes. Text, links, bookmarks, form fields and annotations are carried over as they are. This is a structural rewrite, not a scan to image conversion, so the file is never flattened and the text layer stays intact.

Not directly, because the engine has to decrypt a file before it can rewrite it. Unlock the PDF first, then compress the unlocked copy. If you would rather keep it locked, protect the compressed file again afterwards.

No. The PDF engine is qpdf compiled to WebAssembly and it runs inside the page. Your file is read from disk into memory, rewritten there and handed back to your browser as a download, so nothing is transmitted and there is no upload size limit.

Standard packs the objects into compressed streams at compression level 6. Maximum uses level 9 and additionally linearizes the file, which arranges it so a reader can display the first page before the whole download finishes. Linearization and the stronger compression add a small amount of overhead, so Maximum is not automatically the smaller output.