To split a PDF by document type, go through the file page by page and mark where each new document starts, using the form title, form number, issuer and "page 1 of" cues. Write down a page range and a file name for each document, extract each range into its own file, then confirm the ranges cover every page once.
Why clients send a multi-page PDF with mixed documents
A multi-page PDF with mixed documents is the natural result of how clients scan. A client gathers a stack of paper, scans it at home, and the device saves the whole stack as one file. A W-2, three 1099s and a bank statement end up as pages 1 to 22 of a single PDF.
Clients do this for good reasons. One file is easier to attach to one email or upload once. Scanning each form separately means more files to name and more steps for the client. Some clients are also following an instruction to "send everything in one go", which they read as "one file".
The order of pages inside the file follows the stack, not any logic the firm would choose. Forms are interleaved with envelope inserts and blank back sides. A multi-page bank statement may be split by a 1099 that was lying on top of it. Sometimes a page was fed upside down or twice.
None of this is the client's fault, and asking clients to scan each form separately rarely works for long. The client document collection category covers ways to make sending easier for clients. For the firm, the practical answer is a reliable way to take the combined file apart, which the rest of this guide covers.
Finding where one document ends in a scanned PDF
Finding where one document ends in a scanned PDF means spotting the first page of the next one. The strongest cues are a new form title or form number at the top of a page, a change of issuer name, and a page that says "page 1 of" something. A blank page is a weaker hint.
Form titles and numbers
Tax forms put their identity in a predictable place: the tax year, form number and form name sit together in one block, a layout IRS Publication 1179 also requires on substitute statements. When those details change from one page to the next, a new document has almost always started. The post on which marks on a page identify a tax form's document type shows where each mark sits.
Issuer changes
Two pages can carry the same form number and still be two documents. Two 1099-INTs from different banks are separate forms. Check the payer or issuer name on every form page, not only the title.
Page counters and statement headers
Bank, card and brokerage statements usually number their pages. A page reading "Page 1 of 4" opens a new statement, and the next three pages belong to it. A change in the statement period or account number at the top of a page also marks a new statement, even from the same bank.
Weak cues
A blank page often sits between documents, but it may also be the back of a single-sided form inside one. Treat blank pages as something to note, not as a split point on their own.
How do you split a scanned PDF into separate documents by hand?
Splitting a scanned PDF into separate documents by hand takes two passes. In the first pass you read every page and write a split list: each document's page range and the name it should get. In the second pass you extract each range into its own file. Keeping the two passes apart prevents most mistakes.
Pass one: write the split list
Open the PDF in any viewer that shows page numbers. For each page, note what it is and whether it starts a new document. The table below shows a finished split list for a fictional 22-page scan from a client named Marcus Lee for tax year 2026. All names and account numbers are examples.
| Page | What the page shows | Cue that marks a new document | Split here? | Resulting file name |
|---|---|---|---|---|
| 1 | W-2, Northgate Media Group | Title "Wage and Tax Statement" and year | Yes | Lee 2026 W-2 Northgate Media Group |
| 2 | Blank back of page 1 | None | No, discard | (set aside) |
| 3 | 1099-INT, Harbor Federal Credit Union | Form number 1099-INT, new payer | Yes | Lee 2026 1099-INT Harbor Federal |
| 4 | 1099-DIV, Lakeside Funds | Form number 1099-DIV, new payer | Yes | Lee 2026 1099-DIV Lakeside Funds |
| 5 | 1099-NEC, Example Studio | Form number 1099-NEC, new payer | Yes | Lee 2026 1099-NEC Example Studio |
| 6 | Checking statement, XXXX-4821, page 1 of 4 | "Page 1 of 4", statement header | Yes | Lee 2026-12 Checking XXXX-4821 |
| 7-9 | Same statement, pages 2 to 4 | Counter continues | No | (part of page 6 file) |
| 10 | Blank | None | No, discard | (set aside) |
| 11 | 1098, Example Mortgage Servicing | Form number 1098, new issuer | Yes | Lee 2026 1098 Example Mortgage Servicing |
| 12 | Escrow summary, same servicer | Same issuer, no new form title | No | (part of page 11 file) |
| 13 | Blank | None | No, discard | (set aside) |
| 14 | Consolidated 1099, Ridgeline Brokerage, page 1 of 8 | New issuer, "page 1 of 8" | Yes | Lee 2026 1099 Composite Ridgeline Brokerage |
| 15-21 | Same composite, pages 2 to 8 | Counter continues | No | (part of page 14 file) |
| 22 | Property tax bill, County of Example | New issuer, bill layout | Yes | Lee 2026 Property Tax County of Example |
The list yields eight documents and three set-aside pages. Add up the ranges before moving on: 8 documents covering 19 pages, plus 3 blank pages, makes 22.
Pass two: extract each range
If your PDF editor can extract a page range into a new file, use that. The open-source command-line tool qpdf does it with its --pages option, documented in the qpdf command-line reference. Each line below extracts one range from the example scan:
qpdf --empty --pages lee-scan.pdf 1 -- "Lee 2026 W-2 Northgate Media Group.pdf"
qpdf --empty --pages lee-scan.pdf 6-9 -- "Lee 2026-12 Checking XXXX-4821.pdf"
qpdf --empty --pages lee-scan.pdf 14-21 -- "Lee 2026 1099 Composite Ridgeline Brokerage.pdf"
The --empty option starts from a blank file, and the range after the input file picks the pages to copy into it.
How do you split a PDF with different names in one pass?
Splitting a PDF with different names in one pass works only if the names are decided before the split. A tool that splits by page count cannot know that pages 6 to 9 are a checking statement, so it gives the parts numbers. The fix is to write the split list first and drive every extraction from it.
The qpdf reference shows what a count-based split produces. Its --split-pages option writes each group of pages to a separate file and names the files by page range, so a 12-page input split in pairs becomes files ending 01-02, 03-04 and so on. The same page notes that outlines and other document-level features of the original are not kept when splitting this way. Those names say nothing about what each part holds, and the groups cut across documents of different lengths.
Driving the split from a list avoids both problems. Keep the split list as a plain table with three columns: first page, last page and file name. Each row becomes one extraction, whether you type it into a PDF editor's extract dialog or run it as a qpdf line like the ones above. Because the name is in the row, every file comes out named correctly and nothing needs renaming afterward.
Use one naming pattern for every client. The pattern in the example puts the client, then the year or period, then the form and the issuer. A consistent pattern keeps files sorting in the same order in every folder.
Before closing the original scan, compare the file count against the split list. Eight rows should produce eight files. If one is missing, a row was skipped, and the original still holds the pages.
Splitting a PDF by document type automatically
Splitting a PDF by document type automatically means the software decides where each document starts by reading the pages, instead of cutting every so many pages. It labels each page with a document type, and a change in that label, or in the issuer or period, becomes a split point. Pages with the same labels in sequence stay together.
The difference from a page-count split is the input to the decision. A count-based split looks only at page numbers. A type-based split looks at what is printed on each page, which is the same set of cues a person uses: form titles, form numbers, issuer names and statement headers. The post on rules, templates and AI document classification compared explains the methods software uses to reach a label.
Automatic splitting has the same weak spots as a manual one. A blank page, a continuation page with no header or a statement interrupted by another form can each produce a wrong break. Good tools show the proposed breaks so a person can confirm them before files are created.
Within Schema CPA, every page of an upload is classified by document type, issuer and period, so a long mixed PDF becomes a list of the documents inside it, each with its pages. The list of document categories Schema CPA reads shows which kinds of pages it recognizes.
Whether automatic splitting is worth it depends on volume. A firm that receives a few combined scans a season can split them by hand with a split list. A firm receiving hundreds will spend real staff time on pass one alone.
Edge cases when splitting PDFs: blank pages, back sides and stapled statements
Edge cases when splitting PDFs are the pages that look like a break but are not, or hide a break that is there. Blank pages, the back sides of forms and statements stapled together cause most wrong splits. Each needs a quick check before you commit a page range to the split list.
Blank pages
A duplex scanner records both sides of every sheet, so a stack of single-sided forms produces a blank page after each one. Set blank pages aside rather than deleting them, and count them in your page total. Before setting a page aside, zoom in. A faint page can be a light scan of a real document.
Back sides with printed instructions
Many tax forms carry recipient instructions on the back or on a separate sheet. For substitute forms such as the 1099-INT and 1099-DIV, Publication 1179 requires the recipient copy to include the applicable recipient instructions, and allows them on a separate sheet of paper. These pages belong with the form in front of them. They carry no figures, but splitting them into their own file creates a document that looks unidentified.
Stapled statements
A client who staples two months of statements together and scans them produces one run of pages with two "page 1" headers. Split at the second header, even though the pages look continuous. The reverse also happens: a statement fed out of order may show page 3 before page 2. Keep its pages together and reorder them inside the range.
Rotated and skewed pages
A page scanned sideways or at an angle is harder to read and easier to misjudge as a new document. Fix orientation before writing the split list. The post on what goes wrong in scanned client documents covers rotating and cleaning up pages first.