Choosing the right operation
Most PDF problems take one of a small number of shapes, and picking the matching operation saves a lot of trial and error. If the file is too large for an email attachment, that is compression — and the fix is almost always in the images rather than anywhere else. If you need part of a document, that is splitting or page extraction. If several documents must become one, that is merging, where page order and orientation are the things to check afterwards.
If the text cannot be selected or searched, the document is a scan and needs OCR, not a converter. If the file opens but will not let you print or copy, it has an owner password, which is a permissions flag rather than real encryption. If it will not open at all, it has a user password and genuinely is encrypted.
Getting this diagnosis right first matters because several of these look similar from the outside. A 40 MB PDF might be forty pages of high-resolution scans, in which case compression helps enormously, or four hundred pages of text with embedded fonts, in which case it will barely move and splitting is the better answer.
What compression can and cannot do
PDF compression is mostly image compression. A document's bytes are typically dominated by embedded photographs and scans, and the levers are resolution and encoding: downsampling a 600 ppi scan to 150 ppi cuts the pixel count to a sixteenth, and re-encoding as JPEG at moderate quality removes most of what remains.
This is lossy and not reversible. Downsampling discards pixels permanently, so keep the original if the document might later need printing at full size. Text and vector graphics, by contrast, compress losslessly and are already efficiently stored — which is why a text-only PDF barely shrinks and there is nothing to be gained by trying harder.
Fonts are the other meaningful contributor. Embedding a full font family adds hundreds of kilobytes, and subsetting to only the glyphs actually used typically reclaims most of it. A document that is unexpectedly large despite having no images is usually carrying fully embedded fonts.
Handling documents safely
PDFs are frequently the most sensitive files people handle, which makes the habit of uploading them to whichever free converter appears first in search results a genuine risk. The file reaches that operator's servers, and what happens next depends on a privacy policy few people read.
Everything in this section runs locally in the browser. The document is read with the File API, processed in memory, and written back as a download — no request carries the contents anywhere, which you can verify in the network panel.
One habit is worth adding regardless of tooling: before sharing a document publicly, check what is in it beyond the visible page. Metadata carries author names and software versions, earlier revisions can persist in the file after edits, and content hidden by cropping or covered by a black rectangle is generally still recoverable.