PDF Merge vs Split: When to Use Each
Merge combines, split extracts. Learn when each saves time, how to reorder safely without upload, and how to keep sensitive PDFs local.
Combines multiple PDFs into a single file with reordering, page count, and local processing — no upload.
merge multiple PDF files into one document →When to merge
Merge when the deliverable is one document read in sequence: a cover letter with a CV and a portfolio, an annual report assembled from chapter files, or a batch of single-page scans belonging to one expense claim. The common case is twelve invoices that arrived separately and need to be one submission. The second reason to merge is ordering: once pages are in one file you can be certain of their sequence, which matters for anything a reviewer reads in order. The merge tool here takes up to 30 files, reads each one to count its pages, lets you reorder with the arrow buttons or drop one, and shows the running total of files, pages and bytes before you commit. Nothing is uploaded: the file is read with the File API into an ArrayBuffer, the pages are copied into a new document, and the result is handed to you as a blob download. Client invoices and contracts with account numbers in them can be processed this way without a third party ever seeing the contents.
When to split
Split when you need a subset rather than a union. The frequent reasons are practical: a 200-page catalogue must go out as several attachments under an email limit, only pages 4 to 6 need signing and the appendix should not travel with them, or a scanned book has blank pages and a duplicated insert to remove. Extracting a subset also lets you handle the two halves of a job differently, keeping text pages as vectors while converting only the image-heavy ones. The split tool takes one file, reads the page count, and accepts a range expression: a single page as 5, a range as 3-7, or a mixture as 1, 3-5, 8. Pages outside the document's range are ignored rather than causing an error, and repeated entries are collapsed, so 3, 3, 4 resolves to pages 3 and 4. The selected pages are copied into a new document and the original is untouched. One caution matters more than it sounds: splitting is not redaction. The content of excluded pages is absent from the new file, but nothing has been removed from the original.
How the local PDF tools work
Both tools use pdf-lib, a pure JavaScript library, loaded on demand with a dynamic import so it does not weigh down the rest of the page. Merging creates a new document, loads each source in turn, copies every page with copyPages, and adds them to the target. Splitting does the same with a selected index list. The output is a Uint8Array wrapped in a Blob and passed to a download, so the bytes never touch a network request. Two behaviours of the library surprise people. Bookmarks are not carried over: the outline lives in the document catalog, not on the pages, and copyPages duplicates page objects only, so a merged file starts with no outline even when every source had a good one. Document metadata is also not inherited: the merged output takes pdf-lib's own producer string rather than the title and author of the first file, so set the title yourself if the file is going somewhere. Page sizes are preserved exactly, so a file mixing A4 and Letter comes out mixed; resize before merging if your consumer needs uniform pages. Password-protected files fail at load and must be unlocked first.
Merging is not compression
This is the most common misunderstanding about merging. Concatenating files copies content, so the output is close to the sum of the inputs and is not smaller than any of them. If a document needs to shrink, merging is the wrong operation and will not deliver it, however the order is arranged. Reducing size means changing what is stored, and each option costs something. Rendering pages to images at a chosen scale and re-embedding them as JPEG gives the largest saving and destroys the text layer, so the result is no longer searchable or selectable; the PDF to image tool here offers 1x through 4x, which is 72 to 288 DPI, with JPEG output at a fixed quality. Downsampling images inside an existing PDF preserves the text layer but permanently reduces resolution. Reducing images before building the PDF keeps quality where it matters and avoids carrying a 300 DPI scan into a document read on a laptop. Whichever route you take, keep the uncompressed original: a file recompressed three times is visibly worse than one compressed once, and the detail cannot be recovered.
Workflow cheat sheet
Assembly: name scans in date order as you create them, merge them in one pass with the reorder controls, then export under a new name with a suffix rather than overwriting anything. Keep audit copies separate and untouched. Extraction: for a document that must go out in pieces, split on ranges you chose deliberately, such as 1-10, 11-30, 31-50, rather than accepting a default page count, and check each part against whatever size limit the recipient imposes. Mixed sources: if part of the material is images, convert those to a PDF first with the images to PDF tool, which offers A4 and Letter in both orientations plus a fit-to-image mode, then merge that with the text documents. Before exporting anything you intend to file, open the result and confirm the page count and order, because a merge that silently produced the wrong sequence is the failure you will not notice until it is too late. If you need to check the visual result, render a page or two to an image and look at it.
Verifying privacy yourself
A privacy claim is worth about as much as the check behind it, and the check takes under a minute. Open the developer tools, switch to the Network panel, clear it, and perform the operation. The list should stay empty of any upload. That is the only evidence that matters, and you gathered it rather than being told it. A second check is stronger: load the page, turn the network off, and repeat the merge. It should still work, which can only happen if the processing is local. Note the scope of what you have verified. Some tools on a site do use the network, IP lookups and breach checks for instance, so a claim that everything is offline is usually false; what matters is whether the specific tool you are about to use with this specific document sends anything. A third consideration is what leaves your machine afterwards: the file you download is a real file with real contents, so local processing controls the transport, not the storage.