Skip to content

What merging, splitting, rotating and cropping actually do to a PDF

Which page operations change a single value, which ones build a new file, and why cropped content, bookmarks, form fields and file size behave the way they do.

Published

Merging, splitting, rotating, cropping and numbering pages feel like physical operations: staple, tear, turn, trim, stamp. Inside the file they are something else. Some of them change a single number, some build an entirely new document, and one of them hides content without removing it. Knowing which is which explains most of the surprises people run into afterwards — a file that is bigger than expected, bookmarks that vanished, or text that is still there after it was “cut off”.

A PDF is a tree of page objects

A PDF is not a sequence of pictures. At its root is a catalog, which points to a page tree, which lists the pages in order. Each page is a small dictionary that says how big the page is, which drawing instructions to run, and which fonts and images those instructions use. Simplified, a single page looks like this:

12 0 obj
<< /Type /Page
   /Parent 2 0 R
   /MediaBox [0 0 595 842]
   /Rotate 0
   /Resources << /Font << /F1 7 0 R >> /XObject << /Im1 9 0 R >> >>
   /Contents 13 0 R
   /Annots [14 0 R]
>>
endobj

/MediaBox is the page size in points (595 × 842 is A4). /Contents holds the actual drawing commands. /Resources names the fonts and images those commands need, which live elsewhere in the file as separate objects. /Annots lists annotations such as links and form-field widgets. Things that belong to the document as a whole — bookmarks, the form definition, title and author — hang off the catalog, not off any page.

That split between page-level and document-level information is the key to everything below.

What each tool changes

ToolWhat it does to the filePage content re-encoded?
MergeCreates a new, empty document and copies every page of each source file into it, in the order you set.No
SplitCreates a new, empty document and copies the selected pages into it, in their original order.No
RotateOpens the original and changes the /Rotate value of each page you turned.No
CropOpens the original and sets the /CropBox and /MediaBox of the current page to the area you drew.No
Page numbersOpens the original, embeds one Helvetica font, and adds a short text-drawing instruction to every page.No, only added to

None of these tools redraws a page as an image, so text stays selectable and images keep their quality. The difference that matters is the first column of verbs: creates a new document versus opens the original.

Rotation is a number, not a redraw

The rotate tool reads each page’s current rotation and adds the quarter-turns you chose, so a page that was already at /Rotate 90 and is turned once more is saved as /Rotate 180. The drawing instructions are not touched. A viewer reads the number and turns the page as it displays it.

This is why rotation is lossless and instant, and also why it only works in 90-degree steps: /Rotate accepts multiples of 90 and nothing else. Straightening a scan that is two degrees crooked would mean rewriting the content itself. Each click in the tool turns a page clockwise; three clicks give the anticlockwise result.

Cropping hides content; it does not delete it

The crop tool changes the page boxes of the page you are viewing. After cropping, that page carries something like:

/MediaBox [48 120 520 790]
/CropBox  [48 120 520 790]

Viewers draw only what falls inside that rectangle. The content stream is unchanged, so every word and image that was outside the rectangle is still in the file. Anyone can extract it with a text-extraction tool, or make it visible again by widening the boxes in a PDF editor.

The practical consequence is about privacy, not layout. Cropping away an account number, an address block or a margin note does not remove it. If the purpose is to stop someone reading something, cropping is the wrong tool. Dropping whole pages with split does remove those pages from the output; anything on a page you keep stays, including text that looks covered.

Page numbers are drawn on top

The page numbers tool embeds Helvetica once and appends a drawing instruction to each page: the text, 12 points, black, 20 points in from the edge you chose. Existing content is not moved or inspected, which is why a number can land on top of a footer.

The position is calculated from the page’s width and height, measured from its lower-left corner, without taking account of /Rotate or a crop. Two things follow:

  • On a page that carries /Rotate 90 or /Rotate 270, “bottom centre” is the bottom of the unrotated page, which appears as a side edge once the viewer turns it. Check rotated scans after numbering.
  • On a page that was cropped away from its lower-left corner, the number can fall outside the visible area. Number the document first, then crop.

In the “1 of n” format, n is always the number of pages in the file, whatever starting number you set.

Merging and splitting build a new file

The merge tool and the split tool both start from an empty document and copy pages into it. Copying a page brings along everything that page refers to: its content, and every font and image named in its resources. What it does not bring is anything that lives at document level.

What does not come across

  • Bookmarks. The outline belongs to the catalog. The tools copy pages only, so the output has no bookmarks.
  • Form fields. A fillable form needs a document-level form definition that lists its fields. That definition is not copied, so fields may still be visible but may not behave as a form. Two source files with a field of the same name would conflict in any case.
  • Document properties. Title, author and similar metadata are not carried over from the source files.
  • Internal links. A link that jumps to another page can break, particularly in a split where its target page was not kept.

Rotate, crop and page numbers work differently: they open the original document and save it again, so its bookmarks and other document-level parts are still there.

Why the output can be bigger than expected

Fonts are copied per source file and are not merged with one another. If five documents each embed the same typeface, the merged file contains five copies of it. A merged file therefore tends to be about the size of its inputs combined, not smaller, even when the documents look alike.

Splitting can surprise in the other direction. Some programs give every page one shared resource list naming every font and image in the document. Copying a single page from such a file copies everything on that list, so a one-page extract can be far larger than one page of content would suggest. Checking the file size before sending is the only way to know.

Digital signatures

All five tools write out a complete new file. A certificate-based signature covers the exact bytes of the signed file, so it will no longer validate after any of these operations. Organise the pages first and sign last.

Why password-protected files are refused

An encrypted PDF stores its page content and text strings in encrypted form. The PDF library these tools use cannot decrypt, so it rejects the file when loading it rather than producing something half-readable. This includes files that open in a reader without asking for a password but restrict printing or editing: those restrictions are also implemented as encryption. Removing the protection in a PDF reader you trust, and saving an unprotected copy, is the only route.

All of this happens in your browser tab; the files are not uploaded. If you want to check that for yourself, the guide to what local processing really means shows how to watch the network while a tool runs.

Before you send an assembled document

  1. Do the operations in a sensible order. Merge and split first, then rotate, then add page numbers, then crop, and sign last. Merging after numbering repeats numbers; cropping before numbering can hide them.
  2. Page through the whole file. Confirm the order, and that each page you rotated displays the right way up.
  3. Search for what should not be there. If you cropped or dropped anything sensitive, use the viewer’s search, or select all text and paste it somewhere, to confirm it is really gone. Cropped areas will still be found.
  4. Test anything interactive. Click a few links and try a form field. If the recipient needs to fill in a form, send that form as its own file.
  5. Rebuild bookmarks if they matter. A long merged document without an outline is hard to navigate; add one in a PDF editor if readers will need it.
  6. Check the size and the properties. Look at the file size against the recipient’s attachment limit, and set the title in a PDF editor if it will be shown to the reader.

Tools referred to in this guide