Merge PDF

A PDF file is not really a linear stream of pages in the way a plain document is — it is a graph of objects, where a page references the fonts, images and content streams it needs, and those objects can be shared between pages within the same file. Merging works by copying the object graph of each source document into a new combined file and rewriting the page tree so it lists every page from every document in the chosen order, while preserving each document's internal references so nothing renders broken.

This is why merging is generally lossless in the sense that matters: the pages themselves — their text, images, layout and formatting — are copied exactly as they existed in the source files, not re-rendered or re-encoded. What does not automatically carry across is document-level structure that exists above the level of individual pages, which is where the real limitations of merging show up.

Bookmarks (the outline tree used for navigation in a PDF reader), form field associations, and document-level metadata such as the title and author are structures that describe the document as a whole rather than any single page, and a merge that combines several such structures from different source files has no single correct way to reconcile them — three documents each with their own bookmark tree do not have an obvious combined tree, and three different declared authors do not have an obvious combined author. Rather than guess, client-side PDF assembly generally drops these structures, which is a limitation of merging PDFs client-side broadly rather than something specific to any one implementation.

What is preserved without exception is the visible content: text stays selectable and searchable, images stay at their original resolution, and inline annotations drawn directly on a page carry over with it, because those live at the page level rather than the document level.

Because the documents merge in exactly the list order shown, the practical workflow is to add every file first, then use the reorder controls to arrange the list before clicking merge, rather than trying to predict the final order while adding files one at a time. For a large batch — assembling a dozen scanned statements into one file, for instance — it is worth a moment to verify the order against something external, like the file names or dates, since a merge that finishes in the wrong order still succeeds without any warning that the sequence is wrong.