Page deletion in a PDF is an operation that reveals an important distinction in how the format handles modifications. The PDF specification supports a concept called incremental updates, where changes to a document are appended to the end of the file rather than rewriting the entire structure. When a PDF reader saves a modified document using incremental update, it appends new or changed objects and a new cross-reference table that supersedes entries in the original table. The old data remains physically present in the file but becomes unreachable through the updated cross-reference chain.

When a page is deleted using incremental update, the page tree is modified to remove the reference to the deleted page object, and this modified page tree is appended to the file. However, the actual page object, along with all its content streams, embedded images, and font references, may still exist in the file — they are simply no longer referenced. This means that an incremental delete does not necessarily reduce file size and, in fact, slightly increases it due to the appended update data. The orphaned objects remain recoverable by anyone who examines the raw file structure, which has important implications for documents containing sensitive information.

For true page removal — where the deleted page's data is permanently eliminated — the entire PDF must be rewritten. This process, sometimes called "saving as new" or linearization, rebuilds the file from scratch, including only the objects that are still referenced in the updated document structure. During this rewrite, the cross-reference table is regenerated, object numbers may be reassigned to fill gaps left by removed objects, and all internal references are updated accordingly. The result is a clean file with no traces of the deleted pages.

The page tree must be carefully restructured during deletion. If the page tree uses intermediate nodes with inherited attributes, removing a page might require redistributing those inherited properties to remaining sibling pages. For example, if a parent node specifies a rotation of 90 degrees for all its children and one child page is removed, the remaining children must still inherit that rotation correctly. The page count stored in ancestor nodes must also be decremented to reflect the removal.

Beyond the page tree, deletion affects cross-page features. Bookmarks pointing to deleted pages become dangling references and should be removed or redirected. Internal hyperlinks targeting deleted pages need similar treatment. Page labels — the logical page numbering scheme separate from physical page indices — must be updated so that remaining pages are numbered correctly. A thorough page removal implementation handles all of these secondary effects to produce a structurally sound, clean document.