How to Split a PDF by Question (Not by Page)

Here is the mismatch: “Split PDF” is one of the most searched PDF operations, and virtually every tool that offers it splits by page. But when the goal is questions — a test paper with three questions per page, one of which crosses the page break — page-splitting is the wrong cut.

This guide covers how to split a PDF by question instead, and why the difference is bigger than it sounds.

Page-split vs question-split

A sample exam page tells the whole story:

Page 1:  Q1 · Q2 · Q3
Page 2:  Q4 · Q5 (starts here, continues →)
Page 3:  Q5 (continued) · Q6

The page-split output is not “almost right” — it is the wrong unit, and Q5 being cut in half is not a edge case but what long questions naturally do.

The method: box each question, export all

The reliable workflow is a question-based cropper in your browser:

1. Load the PDF. It renders locally — nothing is uploaded (which matters for unadministered exams; see is it safe to upload a PDF to an online tool?).
2. Draw a box around each question, top to bottom. Include the question number and marks.
3. For a page-crossing question, drag one continuous box from the bottom of the first page into the top of the next. The tool clamps it per-page and stitches the fragments into one file in reading order — a ⇔ marker flags any stitched crop so you can verify it.
4. Export. All splits come down at once as numbered files (Question-1.pdf, Question-2.pdf, …) in a single ZIP.

The full drawing walkthrough — including filename conventions and tips for building a reusable library — is in how to crop questions from a PDF test book and how to build a question bank from past papers.

When question-splitting is the right tool

• Reusing exam questions — one file per question drops straight into a question bank or an LMS.
• Building a Google Forms quiz — each question becomes one image upload.
• Sharing a single question — send Q7 to a student without sending the whole answer-bearing paper.
• Reassembling a new test — pick and reorder question files instead of hunting through pages.
• Scanned papers — works with no text layer, since selection is visual.

What about OCR-based “question detection”?

Some tools claim to auto-detect questions in a PDF. On clean, digital files it can be passable. On real-world papers — scans, two-column layouts, diagrams, sub-parts labeled a/b/c — detection errors are common, and every missed boundary means silent data loss you discover after the test.

Manual box-drawing takes about five minutes for a full paper and gets the boundaries right the first time. For confidential exams, it also keeps the file in your browser rather than on a detection server. Precision beats automation here.

Do it now

Load a real test paper into the question splitter, box every question including one that crosses a page break, and export. Six questions in, you will never reach for a page-split tool again.