A scanned PDF is a photograph pretending to be a document. You cannot search it, select a sentence, copy a paragraph, or have a screen reader read it aloud — because as far as software is concerned there is no text in it at all, just pixels arranged to look like words.
OCR fixes that. Upload your scan above and you get back a searchable PDF: the page looks exactly as it did, but Ctrl+F now works.
This is worth understanding, because it explains both why the result looks perfect and why you still need to check it.
OCR does not redraw your document. The scanned image stays exactly as it was — same pixels, same appearance, same coffee stain in the corner. What gets added is an invisible text layer sitting behind the image, with each recognised word positioned precisely where it appears in the picture.
When you search, your PDF reader looks at that hidden layer and highlights the corresponding area of the image. When you select a line, you are selecting invisible text and seeing the picture underneath light up.
Two consequences follow directly:
That second point is the whole reason to verify your result rather than assume it worked.
Open the finished PDF and press Ctrl+F (Cmd+F on a Mac). Search for a distinctive word you can see on the page — a surname, a product name, a specific figure. If it highlights, the text layer is working.
Then try selecting a sentence with your cursor and pasting it somewhere. That shows you what the text layer actually contains, which is the only way to gauge accuracy. A quick paste of one paragraph tells you more than any accuracy percentage.
If nothing highlights and nothing selects, OCR did not succeed on that page — usually because of scan quality, orientation or language, all covered below.
Accuracy depends far more on the file you upload than on the software. Four factors account for nearly everything.
Resolution. Around 300 DPI is the standard target for text. Below roughly 200, letter shapes lose the detail needed to tell similar characters apart. Above 600 you gain nothing but file size and processing time.
Orientation. Recognition expects horizontal lines of text. A sideways page can fail almost completely, and even a few degrees of skew hurts the line detection that decides where one line ends and the next begins. Run Rotate PDF first — before OCR, not after.
Contrast and evenness. Crisp black on white is ideal. Faded thermal receipts, grey-on-grey forms, photocopies of photocopies, shadows across the page and watermarks crossing the text all reduce accuracy sharply.
Language. OCR uses a trained model per language. Running an English model over French text will mangle accented characters, and over Arabic, Chinese or Cyrillic it will produce nothing usable at all. If your document is not in English, make sure the correct language is selected.
Recognition is pattern matching, so errors follow predictable shapes. The genuinely ambiguous pairs are:
0 and O · 1 and l and I · 5 and S · 8 and B · 2 and Z · rn read as m · cl read as d
In prose these matter little. A misspelled word in a searchable archive is a minor annoyance, and you will usually still find the document by searching for something else on the page.
In numbers they matter a great deal, because a wrong digit still looks like a valid figure. An account number, an invoice total, a case reference, a dosage or a date can be silently incorrect. Never treat OCR output of financial, medical or legal data as verified — check it against the original before relying on it.
These are different jobs and people often reach for the wrong one.
OCR PDF keeps your document looking exactly as scanned and adds searchability underneath. Choose it for archives, records you must keep in original form, legal and compliance material, and anything where the visual appearance — signatures, letterheads, official stamps — is part of the value.
PDF to Word rebuilds the content as an editable document. Choose it when you need to revise the text, restructure it, or reuse it elsewhere. The trade-off is that the layout is reconstructed rather than preserved, so it will not look identical.
For a table of figures you intend to calculate with, PDF to Excel is the better route — though the same warning about misread digits applies with full force.
Searchability is the obvious benefit. Three others are less obvious and often more important.
Accessibility. A scanned PDF is completely invisible to a screen reader — there is nothing to read aloud. For a blind or partially sighted reader, an un-OCR'd document simply does not exist. Adding a text layer makes it accessible, and in many jurisdictions that is a legal obligation for public and educational bodies rather than a courtesy.
Finding documents again. Once a scan contains real text, your operating system's search can index it. A folder of a thousand scanned invoices becomes searchable by supplier name instead of a filename lottery.
Archiving standards. Long-term preservation formats and many records-management policies expect documents to be searchable. OCR is usually the step that gets a scanned archive into compliance.
Straighten a scan first with Rotate PDF, or trim scanner borders with Crop PDF. Afterwards, PDF to Word makes the content editable, PDF to Excel suits tables, and Compress PDF shrinks large scans for email. Browse all PDF tools.