March 24, 2026 · 2 min read
What Is OCR? Making Scanned PDFs Searchable in Your Browser
How optical character recognition turns photographs of text into searchable, copyable documents — and how to get the best accuracy from it.
Open a scanned PDF and press Ctrl+F. Search for a word you can see on the page. Zero results.
That's because a scan isn't text — it's a photograph of text. Your computer sees millions of colored pixels and has no idea that some of them form the word "Invoice". OCR (optical character recognition) is the technology that bridges that gap: software that looks at the pixels and recognizes the characters in them.
What OCR makes possible
Once a scanned document has been through OCR, it stops being a dead image:
- Search works. Ctrl+F finds "termination clause" in a 90-page scanned contract in milliseconds.
- Copy-paste works. Pull a paragraph into an email without retyping it.
- Accessibility works. Screen readers can finally read the document aloud.
- Indexing works. Desktop search and document management systems can find the file by its contents.
The clever trick: the invisible text layer
The best OCR output format is the searchable PDF, and it works in a way most people find surprising: nothing visible changes. Your scan stays pixel-for-pixel identical. The recognized text is added as an invisible layer, with each word positioned precisely over its printed image.
When you search, the invisible layer answers. When you select, you're selecting invisible text that sits exactly on top of the visible words. When you print, you print the original scan. It's the perfect retrofit — the document keeps its authentic appearance and gains a brain.
Our OCR PDF tool produces exactly this (or plain text, if you just want the words out). It runs the Tesseract recognition engine — the same open-source engine behind countless production document systems — compiled to WebAssembly so it executes inside your browser. Your document is never uploaded; the ~15 MB engine and language data download once and cache.
What accuracy to expect — honestly
OCR accuracy depends almost entirely on input quality:
| Input | Word-level accuracy |
|---|---|
| Clean 200–300 DPI scan, printed text | 95–99% |
| Decent phone photo, straight-on | 90–97% |
| Skewed/dim photo, small print | 70–90% |
| Handwriting | Not supported — don't expect results |
Three things you control:
- Language selection. Recognition models are language-specific. Running a German document through the English model wrecks accuracy on every umlaut. Pick the document's main language in the tool.
- Resolution. The tool's DPI setting controls how finely pages are rendered before recognition. 200 DPI suits most scans; bump to 300 for small print.
- Source quality. If you're scanning specifically for OCR: flat pages, straight alignment, good light. Every minute spent there pays back in accuracy.
OCR vs. text extraction — know which one you need
If your PDF was born digital — exported from Word, generated by an invoicing system — it already contains real text. You don't need OCR; you need simple extraction, which is instant and 100% accurate. That's our PDF to Text tool.
The 10-second test: try to select text in the PDF. Selectable → use PDF to Text. Not selectable → it's a scan, use OCR. (And if PDF to Text comes back empty, it will point you to OCR anyway.)
