BlurPDFHow to Extract Questions from a Scanned PDF (No Text Layer Needed)
A scanned exam PDF defeats every text-based tool. You open it, try to select a question, and the cursor drags over a flat image — no text layer, no copy, no OCR, no extraction.
The good news: you never needed the text. You needed the questions — and questions are visual regions of a page. This guide covers the method that works on scans, photographed papers, and print-only PDFs.
Why text tools fail on scans
A PDF has two possible kinds of content:
- Text — selectable characters with positions. Copy-paste, search, and OCR all operate here.
- Raster image — the whole page as pixels. What a scanner, a phone camera, or a photocopier produces.
A scan with no OCR pass is pure raster. Text tools have nothing to grab. And when OCR is run, it introduces its own problems for exam content: equations come out mangled, diagrams lose meaning, two-column layouts interleave, and any error becomes a wrong question in your bank.
The pixel-based approach sidesteps all of it.
The pixel-based method
Since a question is a region of the visible page, treat it that way:
- Load the scanned PDF into a question cropper. It renders the scan in your browser — nothing is uploaded, which matters for exams that have not been administered (why that matters).
- Box each question in reading order. Include the question number and marks; leave answers out.
- Stitch page-crossing questions with one continuous drag across the page boundary — scans split long questions across pages exactly like digital PDFs do.
- Export the ZIP. Each question becomes its own print-resolution PDF (
Question-1.pdf,Question-2.pdf, …).
No text recognition happens anywhere in that flow — which is precisely why it cannot fail the way OCR does. The same workflow, in more detail, is in how to crop questions from a PDF test book.
What if you need editable text too?
Cropping gives you the questions as images — perfect for Google Forms quizzes, question banks, and reassembled papers. If you genuinely need to rewrite the wording, run OCR as a separate step on the cropped question (a single-question OCR pass is far more accurate than OCR on a whole multi-column paper, because the layout is trivial).
That order — crop first, OCR after — beats OCR-then-crop: small images recognize better, and you only OCR the questions you actually want.
Tips for scanned papers specifically
- Scan at 300 DPI or better. Below that, small text gets fuzzy and crops look soft when printed.
- Straighten first. A skewed scan makes question boundaries ambiguous. Any image editor’s rotation fixes it in seconds.
- Photographed paper? Flat lighting and a square overhead angle matter more than camera megapixels. Straighten, then proceed identically.
- Verify the stitches. Scans often have page-edge shadows that tempt over- or under-dragging. Open any
⇔-marked crop and confirm it reads continuously.
Do it now
Take your worst scanned exam — the one every tool has refused — load it into the cropper, and pull out three questions. If it works on that, it works on everything.
Frequently asked questions
Why can't I copy text from my scanned PDF?
Because a scan is an image of a page, not text. Unless OCR (optical character recognition) has been run on it, there is no text layer to select or copy. OCR can add one, but its accuracy depends on scan quality, handwriting, equations, and diagrams — and it is often the wrong tool when what you actually want is the questions as they visually appear.
How do I extract questions from a scanned exam?
Use a visual, pixel-based approach: load the scanned PDF into a browser question cropper, draw a box around each question, and export them all at once. Because the selection is on the rendered image, no text layer is required — the method works identically on scans, photographed papers, and digital PDFs.
Do I need OCR to split a scanned PDF into questions?
No. OCR is for getting editable text; question extraction is about separating regions of the page. A cropper that renders the scan and crops the drawn rectangles needs no text recognition at all — it moves the pixels, not the characters. Skip OCR unless you genuinely need to edit the wording.
Can I extract questions from a photographed paper?
Yes — the same pixel-based method works, as long as the photo is readable and reasonably straight. Load it (many croppers accept images or image-based PDFs), box each question, and export. If the photo is skewed, straighten and crop it in any image editor first so question boundaries are clean.
Ready to try it?
Open the tool and see for yourself — open DevTools → Network and watch zero file requests.
Open the free tool →