How to OCR a scanned PDF on a Mac
Live Text versus OCR
This is the distinction worth taking away. Live Text recognises characters in an image at the moment you look at it, in the viewer, and hands you a selection. Nothing changes on disk. Close the file and the recognition is gone. OCR runs once and writes the recognised words into the PDF as an invisible layer positioned under the picture of the page.
Both let you copy a sentence. Only one makes the document searchable, indexable by Spotlight, readable by a screen reader, convertible to Word and findable in six months when you have forgotten what the file was called.
Live Text is genuinely useful — grabbing an address off a scan is faster than any alternative. It is simply not a substitute, and because it works so well it convinces people they already have a text layer.
The scan matters more than the tool
Recognition accuracy is set at capture time. 300 dpi is the floor for ordinary print; below 200 the engine starts guessing between similar shapes. A page photographed at an angle loses accuracy to keystoning. Shadow across a page from a desk lamp makes a band of text unreadable. Any OCR engine does well with a clean scan and all of them struggle with a bad one, so the time is better spent re-scanning than searching for better software.
Phone capture is fine if the page is flat, the light is even and the camera is parallel to the paper. Document scanning modes that square the page automatically help a lot.
What stays hard
Handwriting, in general. Dense tables, where the engine reads across cell boundaries. Two-column layouts, where reading order has to be inferred. Ligatures and old typefaces. Mixed scripts on one page. A result that is 98% right still means a wrong character every couple of lines, so anything consequential needs reading, not trusting.
Privacy, since it is usually a document that matters
Scanned PDFs tend to be the personal ones — contracts, medical letters, tax paperwork, ID documents. Online OCR sends all of it to someone else's server. Recognition runs perfectly well on a modern Mac without an upload, and for this class of document that is the sensible default.
Step by step
- Test what you have
Open the PDF and press ⌘F. If a word you can clearly see returns no results, there is no text layer. - Scan quality decides accuracy, not the software
300 dpi, flat page, even light. Nothing recovers a blurred or skewed capture. - Run OCR to produce a searchable PDF
The output should look identical and behave differently — selectable, searchable, indexable. - Verify with ⌘F, then with Spotlight
Search a word from the middle of the document. Then search it in Finder to confirm the index picked it up.
Our iOS app for exactly this. Runs on your device — no account, no tracking.