Checked against the current Windows build. Where an article names a menu, a button or a message, that is the wording in the application, not a paraphrase.
Every tool on the Annotate tab adds something; none of them removes anything
Highlight, Underline, Strike out, Note, Box and Draw are the six tools on that tab, and the line above them says what they have in common: marks on the page, none of which remove anything. Box draws a rectangle at coordinates you drag across. Strike out draws a line through words that stay exactly where they were, spelling exactly what they spelled.
The same holds for whatever you might reach for instead. A highlight in a yellow dark enough to look opaque is still a translucent shape over live text. A white rectangle is a white rectangle. Painting the word out in an image editor and putting the image back does work, and it is a different operation from anything on that tab — it also turns the page into a photograph and costs you the searchable text on the rest of it.
How the words come back, in about ten seconds
Open the file in any reader, drag the cursor across the blacked-out area and press Ctrl+C. That is the whole of it. No tools, no intent, and no skill: somebody copying a paragraph to quote it finds the hidden text by accident.
If you would rather see it written down, creasepoint find contract.pdf --query 4412 prints a page number and a snippet for every hit. creasepoint objects contract.pdf --page 3 lists every object on that page in order, with each text run's content, font, size and bounding box, so the covered words and the rectangle covering them appear as two entries a line apart. Export to Word pulls the text into a document you can read at leisure. Three of those four are ordinary features of ordinary software rather than techniques anybody had to learn.
Cropping and a password are not it either
Cropping changes the crop box, and the crop box is only the part a reader chooses to show. Everything outside it stays in the file, and the tool says so in its own result note every time it runs. Anyone who resets the box, or reads the page's text without rendering it, has the lot.
A password is a different job, done well. AES-256 decides who can open the document and removes nothing from inside it: once the recipient has typed the password they hold every byte, including the bytes under the rectangle. The two get confused because both feel like making something private. Password-protecting a PDF sets out what that does and where it stops.
There is no redaction tool here, and that is a decision
It is the first entry on the limitations page. Researchers have pulled supposedly removed text back out of eleven PDF tools that shipped one, and that is the standard a redaction tool has to clear before it exists here rather than an excuse for it not existing. A tool that removes the glyphs and leaves the same value in the metadata, a form field, an attachment or a page thumbnail is worse than no tool, because it stops people checking.
Note what a Creasepoint save is, too, and what it is not. Most operations rebuild the document, write a single revision and read the bytes back to confirm only one is there. That is a real property. It is not sanitising: cropped-away content, metadata and attachments all survive it, and the local-processing page says so in the same paragraph that describes the check.
What to do instead
Go back to the document the PDF was made from, take the value out there, and export again. It is the only route that ends with nothing to find, and it is usually available: the file was printed from something an hour ago.
If the whole page is the problem, remove the page. Delete pages builds a new document and imports the pages you kept rather than appending a revision to the old one, so what you removed is not sitting inside the result. Extract pages does the same from the other direction, when the recipient needs three pages out of forty.
If it has to be that page and the source is gone, the honest ceiling of this application is creasepoint edit page.pdf --page 3 --delete 12 --out out.pdf, which removes one object and rewrites that page's content. It deals with one object on one page. It does not look for the same value in the metadata, in a form field, in an attachment or drawn inside an image, and it is not a redaction workflow. Check the result yourself with find and objects, on every page, before it leaves your hands.
The two places worth looking next.
Read the matching guide if you want the reasoning rather than the fix, and check limitations before spending any longer on it — some of what people look for here is absent on purpose.