Cropping a PDF doesn't remove anything

Someone sends you a scanned contract with a colleague's home address printed along the bottom margin. You crop it off, send the file on, and consider the matter closed.

It isn't. The address is still in the file, and getting it back takes about ten seconds.

What a crop actually does

Every PDF page carries two rectangles. The MediaBox is the sheet the page was authored on. The CropBox is the part of that sheet a reader chooses to display.

Cropping sets the CropBox. Nothing is deleted, nothing is redrawn, and the page content — every glyph, every image — sits exactly where it always did. A reader simply shows you less of it.

This is a sensible design. It means cropping is lossless and instant, that you can undo it, and that a printer can still reach the bleed area outside the trim. It is the right behaviour for the thing cropping is for: framing a page.

It is the wrong behaviour entirely for hiding information, and the interface gives you no hint of the difference. The cropped page *looks* like the content is gone.

How the content comes back

Anyone can reset the CropBox to match the MediaBox. Plenty of PDF tools do it with a menu item. Text under the crop can also be selected by dragging past the visible edge in some readers, or pulled out wholesale by any text extractor, which does not consult the CropBox at all.

If you have ever run a text extraction over a cropped document and been surprised by what came out, this is why.

The same trap, one step along

Drawing a black rectangle over text has the identical problem and feels even safer, because you watched the black box appear. The text is underneath it, in the content stream, fully selectable. This has embarrassed government departments, law firms and newspapers on the public record, repeatedly.

The rule is simple enough to keep in your head:

If you can still see how the page was built, so can everyone else. Hiding is a display decision. Removing is a change to the file.

What removal requires

Genuinely removing content means rewriting the page so that the content is not in it. In practice that means one of two things: editing the content stream to drop the specific text-drawing operators, which is delicate and easy to get subtly wrong, or rendering the page to an image after covering the area, so what remains is pixels with nothing behind them.

The second is what the redaction here does. It costs you the text layer on the affected pages — they stop being searchable, and screen readers can no longer read them — which is a real loss and worth understanding before you accept it. But it is the honest version. The content is not hidden; it is not there.

What to use when

• Trimming margins, framing a scan, tidying a page for print — crop. It is lossless and reversible, which is exactly what you want.
• Removing a name, an address, a figure, a signature — redact, and check the output before it leaves your machine.

The distinction matters most precisely when you are in a hurry, which is also when the crop tool is closest to hand.