Making a Scanned PDF Searchable: What OCR Does
A scanned PDF is a set of pictures of pages. The text on the page is pixels, not characters, so you cannot search it, select it, or copy it. Optical character recognition turns those pixels back into characters and writes them into the document as a hidden text layer behind the picture.
What OCR actually does
OCR takes the page image, finds regions that look like text, and tries to recognise the characters in each region. It outputs the recognised characters and their positions, and the PDF stores that text invisibly on top of the image. The result is a page you can search and select while still displaying the original picture. The image is not replaced — only a text layer is added.
When a scanned PDF becomes searchable
A PDF becomes searchable when a text layer is added to it. That can happen at scan time, if the scanner or scanning app runs OCR as it saves, or afterwards by running OCR on an existing scanned PDF. If you search a scanned PDF and get no results, it has no text layer and needs OCR. If a scan has a partial text layer — say, OCR ran on the first ten pages of a thirty-page batch — only those pages will be searchable.
Why accuracy varies
OCR accuracy depends on the quality of the source. A clean scan at high resolution of printed text in a common font recognises well. A low-resolution scan, a skewed page, handwriting, a photograph of a screen, or text in a decorative font recognises poorly. Compression artifacts from a previous save add noise that the recogniser may read as characters.
Character-level errors are the usual failure: O and 0, l and 1, rn and m. These are exactly the confusions that break a search for a number or an identifier. A single misread digit in an account number is invisible at a glance and fatal to a lookup.
The accuracy caveats that matter
OCR is a recognition step, not a guarantee. Treat its output as a draft to verify, not as authoritative text. For anything you will rely on — an invoice total, an account number, a legal reference — check the OCR result against the image. For search, remember that a misrecognised word is simply not findable; searching for the correct spelling will miss it. If you are scanning for later search, scanning at higher resolution is the single most effective accuracy improvement, more than any OCR setting.
What searchable output is good for
A good OCR text layer makes a document navigable: you can find the page with a specific clause, copy a reference number, and let a screen reader read the page. That last one matters for accessibility — a scanned document with no text layer is unreadable to assistive technology. OCR does not change the page image, so the document looks exactly as it did; it only makes the content usable.