What OCR is, and how to make a scanned PDF searchable
By Shubham · 1 October 2026 · 3 min read
You've got a 60-page scanned agreement and you need to find the clause about the security deposit. You press Ctrl+F, type "deposit", and the PDF reader says: no results. The word is clearly there on page 14. You can see it. So why can't the computer?
Because to the computer, page 14 isn't text at all. It's a photograph.
Scans are pictures of text
When a scanner or phone camera captures a page, it records a grid of coloured dots. Some of those dots happen to form letter shapes, but the file contains no information about which letters they are. It's the same as a photo of a signboard. You can read it, your phone can't, at least not without help.
That's why scanned PDFs:
- Can't be searched
- Don't let you select and copy text
- Can't be read aloud by screen readers
- Convert to Word as a document full of pictures
What OCR does
OCR, optical character recognition, is the "help". Software looks at the image, finds the shapes that are letters, works out which letter each one is, and turns them into real text.
When you run OCR PDF, the result is usually a searchable PDF: the page looks exactly as it did, but an invisible layer of real text sits behind the image, lined up with the words. Search finds the word on page 14, you can highlight a sentence and copy it, and a screen reader can read the page.
If you only want the words and don't care about the layout, you can choose plain text output instead.
How accurate is it?
On a clean, straight, well-lit scan of printed text, modern OCR is very accurate, usually just a stray mistake per page. Accuracy drops with:
- Low resolution. Below about 200 DPI, small letters blur together. 300 DPI is the sweet spot.
- Skewed or curved pages. Book pages near the spine are a classic problem. OCR can straighten tilted pages, but curved lines are harder.
- Poor contrast. Faded photocopies, pencil, or text on coloured backgrounds.
- Stamps and signatures over text.
- Unusual fonts and very decorative type.
- Handwriting. Neat capitals sometimes work; normal joined handwriting mostly doesn't.
The characters most often confused are the ones that look alike: 0 and O, 1 and l and I, 5 and S, rn and m. Always proofread numbers, names and dates in anything important.
Choose the right language
OCR needs to know which alphabet to expect. Our tool supports English, Hindi, Spanish, French, German, Italian, Portuguese, Arabic, Chinese, Japanese and Korean. Pick the language of your document before starting. Running English recognition on a Hindi document produces gibberish, and the reverse is just as bad.
Getting the best results
- Scan at 300 DPI if you're using a scanner. With a phone, use good daylight and hold the camera flat. Our guide to scanning with your phone has more tips.
- Use greyscale or black and white for text documents. Clean contrast helps recognition.
- Straighten and crop pages first, or let the OCR tool deskew them.
- Choose the correct language.
- Proofread the important bits.
What you can do once a PDF has text
OCR unlocks a lot of other tools:
- Edit in Word: PDF to Word produces editable text instead of images.
- Redact reliably: search for every instance of a name or number and remove it with Redact PDF.
- Search across documents: keep years of receipts and statements searchable.
- Accessibility: people using screen readers can access the content.
Do I need OCR?
A quick test: open the PDF and try to highlight a single word. If you can, the PDF already has text and OCR isn't needed. If the whole page highlights as one block, or nothing highlights, it's a scan.
For a single photo or screenshot rather than a PDF, Image to Text pulls out the words directly.
OCR isn't magic, but on a decent scan it turns a dead image into a document you can actually work with.