Split PDF by Content

Split a PDF by content into separate documents

Split PDF by Content separates a PDF that holds several documents — a stack of papers fed through a scanner in one go, or files that were merged together — back into individual PDFs. Instead of noting page numbers and typing ranges, you let AI look at each page and propose where a new document begins: a new letterhead, a different sender, an invoice number, “Page 1 of 3”.

You stay in control. The proposal is shown as page thumbnails grouped into documents, with every split marked and its probability shown; uncertain boundaries are highlighted so you can check them first. Click between two pages to add a split, click a split to remove it, optionally let AI suggest a file name for each part, then download them all as a ZIP.

The pages are read, drawn and rebuilt in your browser. Scanned pages are read with OCR on your device.

How to split a PDF by content

1. Drop the PDF. Each page is read (with OCR for scans) and shown as a thumbnail. For very large batches, set Pages to include.
2. Press Find document boundaries with AI. Or stay offline: press Split at “Page 1 of …”, or click between pages yourself.
3. Check the amber marks first — those are the boundaries the AI was unsure about. Click between two pages to split, or click a “New document starts at page …” bar to join.
4. Optionally press Name the parts with AI to name each part from its content, such as 2026-03-05_Invoice_Sharma-Office-Supplies-Pvt-Ltd. Edit any name by typing over it.
5. Press Download PDFs (ZIP), or use the download icon on a single document.

Split PDF by Content features

How the boundaries are found

For every page the tool takes the beginning (about 700 characters) and the end (about 300 characters) of its text. For each pair of neighbouring pages, TypeSafe’s Jev model answers one yes/no question: does this page begin a new, separate document, or continue the previous one? Signals such as a fresh letterhead, a title, a new sender or reference number, and numbering that restarts at 1 point to a new document; running text, continued tables, “page 2 of 3” and repeated headers point to a continuation.

The answer for each gap is a probability. Gaps at 50% or above become splits; gaps between 30% and 70% are flagged for review whichever way they fall. Large batches are sent in parts of up to 200 pages that overlap by one page, so every gap is judged.

When to check the result carefully

• Poor scans: if OCR cannot read a page, the AI sees almost no text there. Blank separator sheets and photos without text need a manual decision.
• Look-alike documents: twelve monthly statements from the same bank, or several invoices from one supplier, share letterheads; the invoice number, date and “page 1” are what tell them apart, so review the amber marks.
• Non-English documents work, but less reliably than English.
• Long batches: up to 300 pages at a time. Split larger files into chunks with a page range.
• Know the page ranges already? Split PDF is quicker. To reorder or delete pages first, use Organize PDF; to rename many separate PDFs, use Rename PDFs by Content.

Privacy

Your file never leaves the device: thumbnails, OCR and the new PDFs are all produced in the browser. Pressing “Find document boundaries with AI” sends the beginning and end text of the selected pages; pressing “Name the parts with AI” sends the text of each part, up to 12,000 characters. Neither is stored by Galaxy of Tools.