How to OCR a Scanned PDF

Add the file, select the document's main language and start processing. If the document mixes two languages, the combined option is safer; for single-language documents, choosing that language improves accuracy.

The quickest way to confirm the result worked is to search a PDF reader for a word you can see on the page. If it is found, the text layer exists. The page itself looks unchanged, because the recognised text sits invisibly behind the image.

Scan quality matters more than the software. A page that is straight, captured at sufficient resolution and lit evenly without shadows will always outperform a better tool fed a poor image. No setting rescues a very low-resolution scan.

Choosing the right document language has a direct effect, since language-specific characters are the first to be misread under the wrong model. For documents photographed with a phone, correcting the perspective and cleaning the page before OCR gives noticeably better results than running recognition on the raw photo.

OCR is a convenience, not a certification. Compare names, dates, amounts, account numbers and identification numbers against the page image. The usual confusions are predictable: zero and capital O, one and lowercase l, five and S, eight and B.

For documents full of figures, recalculating a total and comparing it with the printed total is far faster than reading cell by cell. For low-quality scans there is no setting that replaces human review.

Once a text layer exists, the document stops being just searchable and becomes usable as input for other tools. Converting to Word or Excel, translating into another language, comparing two versions and redacting text all require a text layer to work at all.

For scans destined for an archive, running OCR and then converting to PDF/A leaves a copy that still opens years later and can still be searched, which is considerably more valuable than storing page images alone.