How to make a scanned PDF searchable: OCR or text extraction
If you cannot select a sentence in a PDF and search (Ctrl+F on Windows/Linux
or Cmd+F on macOS) cannot find a word, do not start by converting the file
to Word or TXT. First find out whether the PDF contains a text layer. That
determines the right operation:
| What the PDF does | Next step |
|---|---|
| Text can be selected and searched | Use PDF to text if you need a separate TXT file |
| Pages behave like photos and search is empty | Run PDF OCR to create a searchable PDF |
| You have separate page photos | Assemble them in the document scanner, then run OCR if needed |
| Search works but the text contains many mistakes | The file already has weak OCR; ordinary extraction will not repair it |
A one-minute text-layer test
Open the PDF in an ordinary viewer and search for an uncommon word on the first page: a surname, contract number or company name. Then drag across one line.
If individual words or characters are selected rather than a rectangular area, the layer already exists. To export it, open PDF to text. The tool reads the existing layer locally in the browser. It does not run OCR, so an image-only scan produces a clear no-text result rather than guessed text.
If the whole page behaves like an image, you need OCR. OCR analyzes the page image, recognizes characters and places text behind the original visual layer. The page still looks like the scan, but you can search, select and copy words.
Create the searchable PDF
- Open recognize PDF text.
- Select Russian, English or Russian + English.
- Upload one scanned PDF and wait for processing.
- Download the searchable PDF.
- Search the result for a surname, a date and one word printed in small type.
OCR is server-side: the file is sent to pdfy over HTTPS, and the source and result are deleted after processing. The current limits are 20 MB, 15 pages and 180 seconds. If company policy does not allow the document to leave the device, use an approved desktop OCR workflow instead.
What changes OCR quality
Straight, high-contrast scans with legible type work best. Blur, fold shadows, slight rotation, bleed-through and handwriting all increase recognition errors.
If you still have separate photos, prepare them in the document scanner: reorder pages, crop margins, rotate images and adjust brightness or contrast. This preparation stays in the browser and does not run OCR. It creates a cleaner PDF that you can then submit for recognition.
Check these details closely:
- surnames, addresses and item codes;
- dates, totals and decimal separators;
- lookalikes such as
O/0,S/5andI/l/1; - tables and multi-column layouts;
- pages with stamps, signatures or a weak background.
A searchable PDF is useful for retrieval, but it is not proof of accuracy. If the extracted values feed a contract, report or calculation, compare them with the page image.
When OCR is the wrong step
If text is already selectable, OCR adds work and can make copying worse. Go straight to text extraction.
If your real goal is a spreadsheet, OCR is only the first stage. You still need table reconstruction and a manual check. The guide to extracting a table from PDF to Excel explains why a visible grid is not the same thing as real spreadsheet cells.
Before you finish, test search on the first and last page, copy one line with accented or non-Latin characters, and compare critical numbers with the scan. Those three checks expose weak OCR faster than scrolling through the whole document.
FAQ
- How can I tell whether a PDF is a scan?
- Try selecting a sentence or searching for a distinctive word in the viewer. If the page behaves like one image and search finds nothing, the file probably has no text layer.
- How is OCR different from extracting text from a PDF?
- Extraction reads a text layer that already exists and can save it as TXT. OCR recognizes letters in page images and adds a new searchable layer to the scan.
- Does pdfy run OCR locally?
- No. The OCR tool sends the PDF to the pdfy server over HTTPS. The source and result are deleted after processing. Extracting an existing text layer to TXT runs locally in the browser.
- Does OCR turn the PDF into a fully editable document?
- No. The result is a PDF with a recognized text layer for search and copying. It is not a reconstructed DOCX, and names, numbers and tables still need checking.