OCR explained: how a scanned PDF becomes searchable text

A scanned document — whether it came from an actual scanner or just a phone camera — is a photograph as far as a computer is concerned. Ctrl+F finds nothing, because there's no text there to find, only pixels that happen to look like letters to a human reading them.

What OCR is actually doing

Optical Character Recognition analyzes the shapes on the page and predicts which characters they represent, then adds that predicted text as an invisible, selectable layer positioned right on top of the original image. Visually, nothing changes — you're still looking at the same scan. Functionally, the page now has real text behind it that a browser or PDF reader can search, select, and copy.

Why OCR runs entirely in your browser here

OCR PDF uses Tesseract.js, a WebAssembly build of the open-source Tesseract OCR engine, running as a Web Worker so it doesn't freeze the page while it works. The recognition itself happens on your device — the scan doesn't get sent anywhere to be read.

What affects accuracy

• Scan quality — a crisp, well-lit, straight scan recognizes far more accurately than a blurry or heavily skewed phone photo.
• Handwriting — Tesseract, like most OCR engines, is built for printed text; handwriting recognition is a different, much less reliable problem.
• Unusual fonts or heavy stylization — decorative typefaces or low-contrast text reduce accuracy the same way they'd challenge a human skimming quickly.