Optical character recognition — pulling editable text out of an image, screenshot, or scanned document — is genuinely useful but has real, honest limits worth understanding before you rely on it. These tools use a real OCR engine (Tesseract) running entirely in your browser, and accuracy depends heavily on what you're feeding it: clear, well-lit printed text works well; handwriting, blurry photos, and unusual fonts work much less reliably.
A distinction worth knowing: a PDF's text layer (if it has one) can be extracted instantly and perfectly, with no OCR needed at all — that's what our PDF to Text tool does. OCR is only necessary for scanned documents or images that don't have an underlying text layer, which is a fundamentally harder, slower, and less accurate process, since the engine has to actually recognize characters rather than read data that's already there.
For structured documents — tables, receipts, invoices — these tools attempt automatic field detection, but this is pattern matching (looking for "total" near a dollar amount, for instance), not real document understanding. Every result is editable specifically because it should be checked, not trusted blindly, especially for anything financial.