Most PDF jobs are small: put three files together, pull out two pages, turn a stack of photos into one document, produce something printable. None of that needs installed software, and none of it needs to leave your computer.
Here is what a browser can do, and, more usefully, what it cannot.
Merging and splitting
A PDF is a container of pages. Merging copies pages from several containers into a new one; splitting copies a subset into a smaller one. Both operations move the page objects as they are, so text stays selectable, images keep their resolution, and nothing is re-compressed.
That also means mixed page sizes survive. An A4 cover letter followed by an A3 drawing stays exactly that.
Two things do not survive, in any tool: form fields and digital signatures. A signature covers a specific document, and the merged file is a different document, so the signature cannot carry over. That is mathematics, not a limitation of the software.
Making a PDF from images
Scans, receipts and screenshots usually arrive as separate images and need to be one document. Images to PDF embeds JPEG and PNG data directly, without re-encoding, so quality is untouched and the file ends up roughly the size of the images together.
The page size choice matters more than it looks. Fitting each page to its image is right for anything read on screen. A4 or Letter with a margin is right for anything printed or uploaded to a form, because a form that expects a page and receives a 4000 by 3000 pixel page will do something unhelpful with it.
If the result is too heavy to email, resize the images to about 2000 pixels on the long side first. A printed page cannot show more detail than that anyway.
Creating a PDF from scratch
The shortest reliable route is the one built into your operating system: print, then choose “Save as PDF” instead of a printer. Every current browser on Windows, macOS, Linux, Android and iOS can do it, and you get the exact page size your system uses.
That is how the invoice generator works. It renders the document as a page and lets your browser write the PDF, which is more reliable than generating one in JavaScript and produces a smaller, properly sized file.
What a browser genuinely cannot do well
Being honest about the boundary is more useful than pretending there isn’t one.
Optical character recognition. Turning a scanned page into searchable text needs a real OCR engine. Browser-based OCR exists and is slow and mediocre for anything but clean, straight, high-contrast scans.
Heavy compression. Shrinking a large scanned PDF properly means re-sampling every embedded image and possibly re-encoding to JPEG 2000. Tools that do this well are desktop software.
Editing existing text. A PDF stores positioned glyphs, not paragraphs. Changing a sentence means reflowing everything after it, which is why “PDF editors” either approximate it or convert to another format and back.
Real redaction. Covering text with a black box is drawing, not removing. Proper redaction deletes the underlying content, and you should verify afterwards by trying to select the text.
For those four, use proper software. For the other ninety percent of PDF work, a browser is faster, and the file never leaves your machine.
The habit worth having
Ask what a PDF contains before you send it. A merged bundle of documents carries every page you added, including the one you meant to remove, and the metadata of whichever tool made it. The tools here clear the producer and creator fields when writing a file, which is a small thing that costs nothing and occasionally saves an awkward conversation.