Deleting a page looks like the simplest thing a document editor does. It is, right up to the moment the file comes out the same size as it went in — which is the normal result, not a fault, and is worth understanding before you send that file to anybody.
How to delete pages from a PDF
Select every page first, then rebuild
Tick all the pages you want gone before removing any of them, so the numbering cannot shift under you, and save a document made from the pages you kept. If the file does not shrink, the pages are still in it.
The two ways a page can go
A PDF is a set of numbered objects and a table at the end saying where each one lives. Pages are objects; so are the images on them, the fonts they use and the text itself.
The cheap way is to leave every object where it is and append a new table saying the document now has fewer pages. This is what the format was designed for and it is genuinely clever: saving a change to a 400 MB file takes a moment, because almost nothing is written. The page stops appearing. Every byte of it stays.
The thorough way is to make a new document, copy across the pages you are keeping, and write that out. Anything only the removed pages referred to is not copied, because nothing points at it any more. The file shrinks, and the page is gone in the sense most people mean.
Both are correct behaviour and no dialogue tells you which you got. The tool here does the second — it builds a fresh document from the pages you kept — and prints the resulting size next to the page count so the difference is visible rather than asserted.
The test that needs no trust
Look at the byte count before and after. A twenty-page scan where you removed ten pages should come out near half the size. If it came out slightly larger, the removal was an annotation and the pages are still in the file, readable by anyone who opens it with something other than a viewer.
That matters in one situation and is harmless in every other. Sending the trimmed file to somebody who should not see the removed pages is the situation. Tidying a document for your own use is not, and the cheap save is the better trade there.
Why the second deletion takes the wrong page
This is the mistake that costs people an hour, and it has nothing to do with PDFs specifically.
You want pages 3 and 5 gone. You delete page 3. The document now has one fewer page, and everything after the gap has moved up: the page that was 4 is now 3, and the page that was 5 is now 4. Delete page 5 and you have taken what used to be page 6.
The fix is not care, it is order. Decide the whole selection against the original numbering, then remove everything at once — which is also why deleting from the bottom upwards works, since a page never moves when something after it goes. A grid of thumbnails you tick makes the trap structurally impossible, and it also means you are looking at the pages rather than at a list of numbers you wrote down earlier.
What deleting does not do
It does not redact. A black rectangle drawn over a paragraph is a rectangle in front of a paragraph. The words remain, selectable, copyable and findable. Removing the whole page removes them with it, which is why deletion is a real redaction for a whole page and no help at all for a line in the middle of one. For anything smaller, the page has to stop containing text — converting it to an image is the blunt version that actually works.
It does not remove what the file knows about itself. The author name, the software that produced it and the creation date are properties of the document, not of any page, and they survive every page you take out. They live somewhere else entirely.
It does not renumber anything printed on the page. A page with “4” typed in its footer still says 4 after the pages before it are gone. Page numbers drawn into the content are pictures of numbers; adding a fresh set is the only thing that fixes the sequence.
Deleting, splitting and reordering are three answers
They get confused because all three start with picking pages, and they end in very different places.
Delete when the pages are surplus — the blank backs a duplex scanner adds to every odd-length original, the cover sheet on an incoming fax, the pages of a contract you are not party to.
Split when both parts matter. A scanner run that swallowed six separate documents is one file that should be six, and nothing in it is waste.
Reorder when nothing should go and the sequence is wrong — the commonest reason being a document fed into the scanner back to front, which produces a perfectly complete file in exactly the wrong order.
The short version
Select every page you mean to lose before removing any of them. Save a rebuilt document rather than an amended one, and check the byte count to see which you got. Remember that a deleted page takes its own contents and nothing else — not the document’s properties, not the numbers printed on other pages, and not anything you merely covered up.
All of it happens in your browser here, on a file that is never sent anywhere, which also means you can trim the same document three ways and compare the results before deciding which to keep.