Can't copy text from a scanned PDF? OCR makes it searchable
You can read every word on the page with your own eyes. But Ctrl+F finds nothing, and dragging the mouse selects nothing. That is because the page is not text at all — it is a photograph of text. OCR is the step that reads the photograph and lays real text over it, so search, select and copy all start working.
Why a scanned PDF has no text inside
A scanner — or a phone camera — does not read your document. It photographs it. The PDF you get is a neat wrapper around that photograph: pages that look like text, made of pixels, with no letters stored anywhere inside the file. Your reader shows you the picture faithfully, but there is nothing for it to search, select or copy.
The two-second test: open the PDF, press Ctrl+F, and type a word you can plainly see on the page. “0 results” means you are looking at a picture. A born-digital PDF — one saved from Word or a form — passes the same test instantly, because its text was text from the start.
What OCR actually does
OCR — optical character recognition — looks at the photograph and works out which letter each shape is. Then it places an invisible layer of real text over the page, lined up with the picture. The page looks exactly the same as before. Underneath, it is now a document a computer can read: Ctrl+F finds words, the mouse selects sentences, and copying pastes actual text instead of nothing.
Make a scan searchable, step by step
Step 1 — straighten first, if the pages lean. Crooked pages are the commonest cause of bad recognition. If your scan sits at an angle, run it through Deskew PDF first. Straight, sharp pages OCR far better than slanted, soft ones — a minute here saves ten later.
Step 2 — open OCR PDF and add the file. OCR PDF takes one scanned PDF and reads it in your browser. Pick the document's language — English, Urdu, Hindi, Arabic and French are supported. The recognition data for a language (about 0.6–2 MB) downloads once, the first time you use it, and stays in your browser afterwards.
Step 3 — choose the quality. Best renders pages at 300 DPI — use it for small print, dense pages and anything official. Faster renders at 200 DPI and is fine for big, clear text when you are in a hurry.
Step 4 — download and test it. Open the result and repeat the two-second test: Ctrl+F a word you can see. It should be found, highlighted, on the page. Select a sentence and copy it somewhere — real text should arrive.
Step 5 — want the words as a file? Run the searchable PDF through PDF to Text to pull the text out as a plain file. And if your “document” is a single photo rather than a PDF, OCR Image reads the photo directly.
The honest limits
- Handwriting reads poorly. OCR is built for printed text. A handwritten page will come back with many mistakes — sometimes mostly mistakes
- Blurry and angled scans make errors. The recogniser guesses at unclear shapes, and a guess can turn a 3 into an 8
- Names, numbers and dates need an eye-check. Whatever the document is for — an application, a bank, a court — compare the important figures against the picture yourself before you rely on them. OCR is a very fast reader, not an infallible one
Used with those limits in mind, it changes what a scan is: the archive of photographed pages on your drive becomes a library you can actually search. And because compression turns pages into pictures too, OCR is also the step that gives a compressed PDF its text back.
Give the scan its text back
OCR PDF reads your scanned pages in the browser and hands back the same document — searchable, selectable, copyable.
Open OCR PDFA worked example: an archive scan becomes a searchable file
Say a forty-page scanned agreement has been sitting in a folder for a year, and today someone asks which page mentions the renewal date. Without a text layer, that question costs you forty pages of reading. With one, it costs a Ctrl+F. Here is the file's journey. The two-second test first: Ctrl+F, type a word you can see — “0 results”, as expected. The pages lean slightly, the way phone scans do, so the file goes through Deskew PDF before anything else, because recognition quality is decided by the pages you feed it.
Then OCR PDF, language set to the document's own, quality on Best — 300 DPI, because this is an official document with small print. The recognition data downloads once, the file is read on the device, and the result comes back looking identical: same pages, same photograph of every page. Underneath, each page now carries its text. Ctrl+F finds the renewal clause on page twenty-three. A paragraph copies as real text. And because the text now exists, the file can go further — through PDF to Text for a plain-text copy, or on to conversion, the way a scanned table becomes a spreadsheet in our PDF table to Excel guide.
When the OCR comes back wrong, work through it in order
- Whole pages of nonsense. Usually the language, not the file. A document recognised with the wrong language selected comes back as confident gibberish. Check the language matches the document's — English, Urdu, Hindi, Arabic and French are supported — and run it again
- Good text, leaning pages. The order was reversed: the scan needed Deskew before OCR, not after. Straighten the original scan, then re-run recognition on the straight file — the text layer is rebuilt from the better pages
- Small print full of errors. The Faster setting renders at 200 DPI, which is fine for big, clear text and unkind to small print. Re-run on Best (300 DPI) for dense or official pages
- Mostly right, numbers wrong. That is OCR's known weak spot, not a broken file: unclear shapes get guessed, and a guess can turn a 3 into an 8. The text layer is good for searching; names, numbers and dates still get eye-checked against the picture before anyone relies on them
Before you upload: the 60-second check
- The two-second test failed first — Ctrl+F found nothing in the original
- Slanted pages were deskewed before OCR, not after
- The right language was picked for the document
- You used Best quality (300 DPI) for small or official print
- Afterwards, Ctrl+F finds a word you can see on the page
- Names, numbers and dates were eye-checked against the picture
Fair questions
Why can I see the words but not copy them?
Because the page is a photograph. A scan stores a picture of the page, not the letters themselves, so there is no text for your reader to select or search. OCR reads the picture, works out the letters, and adds a real text layer over it — after that, copying and searching work normally.
Will OCR change how my document looks?
No. The visible page stays the same photograph it was. OCR adds an invisible text layer underneath it, aligned with the words. What changes is what you can do with the document: find, select, copy, and feed it to other tools.
Does it work for Urdu, Hindi or Arabic documents?
Yes — pick the document's language before you run it. English, Urdu, Hindi, Arabic and French are supported. The language's recognition data downloads once (about 0.6–2 MB) and then works offline in your browser.
Can OCR read handwriting?
Poorly, honestly. OCR is built for printed text. Neat printed pages read well; handwriting comes back with many errors and should not be relied on. If the page matters, treat OCR output as a draft and check every line against the original.
Is my document uploaded to your server?
No. OCR PDF runs entirely in your browser — the recogniser works on your own device, and your document never leaves it.
My document is a single photo, not a PDF. Can OCR still read it?
Yes. That is what OCR Image is for: it reads a photograph directly and returns its text, without the PDF step in between. The same honesty applies — printed text reads well, handwriting poorly, and the language choice matters. If the photo is one page of a longer document, you may still prefer the PDF route so the whole document stays together.
One page came back garbled when the rest were fine. Why?
Almost always the page, not the process. A blurred photograph, a page shot at an angle, or a shadow across the print gives the recogniser shapes it can only guess at — and it guesses wrong, page by page. Re-scan that one page straight, flat and sharp, replace it, and run the recognition again. Pages that were photographed well rarely need a second pass.
How OCR works is general document technology. Tool facts — languages, quality settings, in-browser processing — are from PDFJadoo's own OCR PDF page, checked 10 November 2026.
Keep reading
Deskew PDF
Crooked pages OCR badly. Straighten the scan first, recognise second.
PDF to Text
Once the text layer exists, pull the words out as a plain text file.
Scan Documents, no app
Where the picture-pages come from — and how to shoot them straight in the first place.
Edit a PDF resume
A searchable scan is the first step to a resume you can actually work with.