Blanked in place, byte for byte
A PDF depends on a table of byte offsets pointing at every object in the file. Rewriting the structure risks breaking that table and producing a file that will not open. Instead, this replaces the metadata values with spaces of the exact same length, so every offset in the document remains correct and the file opens exactly as before.
What is covered
The Author, Title, Subject, Keywords, Creator and Producer fields, the creation and modification timestamps, and any XMP metadata packets embedded in the document.
What it will not do, and why
If a PDF has been saved incrementally, its earlier versions are part of the file body rather than its metadata. Removing them means rewriting the document, which is a different and riskier operation. The tool detects this and tells you, so you can re-export the file instead. Printing to PDF from a viewer flattens it reliably.