PRACTICAL GUIDES · PDF and documents
How to extract text from a scanned PDF
A PDF can display words without containing text you can copy. Follow the route from identifying a scan to extracting useful text and checking recognition errors.
By Puffinmint · Updated
1. Check whether the PDF already contains text
Try selecting a sentence and searching for a word in your PDF reader. If copying works, use PDF to text directly. When the page is only an image, optical character recognition may be necessary.
Empty or garbled copied text is not proof of a scan: restrictions and character encoding can also cause problems. Test a simple page and compare it with the original.
2. Prepare a readable page and choose its language
Keep the original. Check rotation, contrast and page edges; missing lines cannot be recovered with certainty. If you can scan again, avoid blur and shadows. Start with one or two representative pages.
In OCR PDF, select the document language and page range. The available languages are Italian, English, Spanish, French, German, Japanese, Korean, Polish and Turkish. Portuguese is not currently offered; the interface language does not change this list.
3. Extract text from the recognized PDF
Run OCR PDF and download its output PDF, which includes a recognized text layer. Open it and test search and selection. Then open PDF to text, select this new PDF, choose the pages and copy the text or download TXT.
OCR uploads the document to the server when processing starts; PDF to text extracts text in your browser. Check each tool’s current limits and file handling details. A TXT file does not preserve the page layout.
4. Proofread against the original
Check the original and extracted text line by line, especially names, dates, amounts and similar characters such as O/0 and I/1. Check reading order across columns and tables as well as individual words.
The synthetic example below contains three recognition mistakes. Correct them using the original, then read the whole passage again. Text comparison can highlight edits between versions; it cannot verify a transcription against the document for you.
- All required pages are present in the right order.
- Names, figures, symbols and separators match the original.
- The text still reads in a sensible order after pasting into its destination.
Reviewing one extracted line
- Starting values
- Unchecked OCR: B-I02 · 2026-11-O2 · 1 25O
- Result
- Checked original: B-102 · 2026-11-02 · 1 250
Keep both copies. These corrections come from checking the original, not from automatically guessing missing characters.
Sources and method
Frequently asked questions
Does OCR automatically make a PDF accessible?
No. Recognition and reading order may still be wrong. Text recognition, document tags, structure and accessibility testing are separate checks.
Can I recover an unreadable word automatically?
Not reliably. Obtain a clearer scan or check the source document instead of guessing missing words or numbers.