You open a scanned PDF, press Ctrl+F, type a word you know is on the page, and the reader finds nothing. You try to select a sentence to copy it, and nothing selects. The document looks perfectly readable, yet the computer treats it as a picture. That is because a scanned PDF often contains images of pages, not text. Optical character recognition, usually shortened to OCR, bridges that gap. OCR PDF runs OCR in your browser and adds an invisible text layer to a scanned PDF, so find, select, and copy start working. This foundational guide explains what happens, how to use it, and how to check the result.
Why a scanned PDF has no text to search
A scanner does not read words; it photographs a page. The resulting PDF stores that photograph as an image, with no information about which marks on the page are letters, which are numbers, and where one word ends and the next begins. A PDF reader displaying such a file shows the image faithfully, but its search function has nothing to search, because there is no text layer underneath the image. Copying fails for the same reason: there are no characters to copy, only pixels.
This is different from a PDF created from a word processor or a web page, where the text was born digital and stored as characters. Those PDFs are searchable from the start. Scanned PDFs only become searchable when something reads the image and writes down what it sees, which is exactly what OCR does. Understanding that distinction prevents a common misunderstanding: a scanned PDF is not broken, and your reader is not faulty. It simply never had a text layer, and OCR is the step that adds one.
What OCR adds, and what it does not change
OCR examines each scanned page, recognises the shapes of letters and numbers, and builds a layer of text positioned to match the image. That layer is invisible: when you open the processed PDF, you still see the original scanned page, exactly as it looked before. The difference is underneath. Press Ctrl+F and the reader searches the new text layer. Drag across a paragraph and the words select. Copy a sentence and paste it elsewhere, and the characters arrive as text.
Crucially, OCR does not replace or redraw the scanned image. If a page was crooked, faint, or stained before OCR, it will look crooked, faint, or stained afterwards. The text layer is an addition, not a cleanup. That is a strength in one sense, because the original appearance is preserved as evidence, but it also means OCR cannot rescue a page that was unreadable to begin with. A human can often guess a faded word from context; OCR must recognise its shape, and a shape too faint to recognise will be guessed, skipped, or misread.
Worked example: a fictional scanned letter, before and after
Consider a fictional example. Alex Morgan, an invented volunteer, has a scanned two-page letter from the fictional Example Residents' Association about a meeting on 14 May 2026. Before OCR, Alex opens the PDF, searches for "meeting", and gets no results, even though the word appears in the first paragraph. Copying the date to paste into a calendar fails too, because there is no text to copy, only an image of the date.
Alex runs the file through OCR PDF. Afterwards, the pages look identical, but searching for "meeting" now jumps to the first paragraph, and copying the sentence containing the date pastes clean text. Alex then proofreads the copied text against the scanned image, because OCR accuracy depends on scan quality and the result needs checking before it is quoted or filed. One word in the second paragraph was misread, a "5" recognised as an "S" in a reference code, and Alex corrects it in the notes rather than trusting the copied text blindly. The before-and-after is striking, but the proofreading step is what makes the after version trustworthy.
How to check an OCR result properly
Do not trust an OCR result on the basis of one successful search. Search for several distinctive words from different pages, including numbers, names, and dates, which are the details most likely to be misread. Copy a paragraph and read it against the scanned image, watching for swapped characters, missing punctuation, and words split across lines. Check the beginning and end of each page, where scanning artefacts and cropped margins cause the most errors. If you plan to quote from the document, verify every quotation against the image, not just against the text layer.
If errors cluster on particular pages, those pages probably had a quality problem: low contrast, skew, a shadow, or small print. Rescanning those pages at a higher resolution, flatter and better lit, and running OCR again usually improves the result more than running OCR repeatedly on the same poor images. OCR is a reading of the image you give it; a better image gives a better reading.
For advice on comparing two versions of a document after OCR, see how to compare documents.
Search for phrases, not just single words
Once a text layer exists, use it the way a reader would, not just the way a test would. Searching for a single distinctive word proves the layer works, but searching for a short phrase, such as a full clause or a person's full name, tests whether word spacing and line breaks survived recognition correctly. Phrases that cross a line break are a useful check, because OCR sometimes joins or splits words at line ends in ways a single-word search would miss. If a phrase search fails where you expect it to succeed, open that page and compare the image with the text layer directly. That comparison usually shows whether the problem is a single misread character or a wider layout issue on that page, and it tells you whether a rescan or a careful manual note is the right response.
Honest limitations
OCR accuracy depends on scan quality, and that dependence must be stated plainly. Clean, high-contrast, straight scans of printed text OCR well. Faded print, handwriting, decorative fonts, complex tables, and photographs of pages taken at an angle OCR poorly, and handwriting in particular is usually recognised badly, if at all. OCR output always needs proofreading before it is quoted, filed as a record, or used for anything where an error would matter. OCR also does not make a scanned document editable in the sense of a word-processor file: the text layer supports searching and copying, but the page itself remains an image, and correcting the text layer does not retypeset the page. Finally, OCR cannot recover text that is not visible in the scan, such as words cut off at a margin or hidden under a shadow.
FAQs
Why can't I search or copy text in my scanned PDF?
Because the PDF contains images of pages, not text. A scanner photographs a page rather than reading it, so the reader has no characters to search or copy until OCR adds a text layer.
What is an invisible text layer?
It is a layer of recognised text positioned underneath the scanned image on each page. You see the original image, while search, selection, and copying use the text beneath it.
How accurate is OCR?
It depends on the scan. Clean printed text on a straight, high-contrast scan is usually recognised well. Poor scans, small print, and handwriting are recognised badly, so always proofread OCR output before relying on it.
Will OCR change how my scanned pages look?
No. The visible pages stay the same. OCR adds a searchable text layer without redrawing or cleaning up the scanned image itself.
If a scanned document needs pages reordered after OCR, Organize PDF handles that visually, and JPG to PDF builds a PDF from photographs when a document starts as images.
human review pending client designation