Why scanned PDFs are different
A PDF is a container. Some PDFs contain real text (letters you can select and copy); others contain only images of pages — what a scanner or a phone app produces. For an image-only page, a summarizer that just reads the text layer sees an empty document.
When you upload a PDF here, the tool first checks how much readable text each document has. If the text layer is missing or too poor, it runs optical character recognition (OCR) on the pages, then summarizes the recognised text. The result panel shows an OCR badge so you know which path was used.
How to summarize a scanned PDF
- Upload the scanAny PDF from a scanner, a multifunction printer or a phone scanning app.
- Wait a few extra secondsOCR takes longer than reading a text layer: count roughly 2-4 seconds per page.
- Choose what you needSummary, key points, questions, or Extract text to get the recognised text as a .txt file.
- Check names, dates and figuresOCR can confuse similar characters (0/O, 1/l, 5/S), especially on low-quality scans.
Getting good OCR results
- Resolution: 200-300 dpi is ideal. Below 150 dpi, small print becomes unreliable.
- Straight pages: skewed or curved pages (book spines) reduce accuracy.
- Contrast: black text on white paper works best; faded thermal paper and coloured backgrounds are hard.
- Handwriting is generally not recognised reliably.
- Mixed PDFs are fine: if most pages have text, the text layer is used.
Common scanned documents people summarize
- Signed paper contracts and leases.
- Letters from administrations, banks and insurers.
- Old reports, theses and archives digitised from paper.
- Medical or school records received as scans (check them carefully).
- Book chapters scanned for a course.
Frequently asked questions
How do I know if my PDF is scanned?
Open it and try to select a word. If you can't select text, or the selection covers the whole page like an image, the PDF is scanned. The tool detects this automatically anyway.
Why only 40 pages with OCR?
OCR is computationally expensive. To keep the free tool fast for everyone, scanned PDFs are OCR'd up to their first 40 pages. Documents with a text layer can be up to 300 pages.
Which languages does the OCR recognise?
English, French, German, Spanish, Italian and Portuguese are recognised best. Other Latin-alphabet languages often work with more errors; non-Latin scripts in scanned images are not reliably recognised yet.
Can I just get the recognised text?
Yes. Use the Extract text tab and download the text as a .txt file, page by page.
Is a photo of a document OK?
Phone scanner apps that produce a flattened, cropped PDF work well. A raw photo saved as PDF works if the page is flat, sharp and well lit.