How do I make a scanned PDF searchable so I can find the text?
A scan is a photograph of a page. Optical character recognition adds an invisible text layer behind the image, which is what makes it searchable, selectable and readable by a screen reader.
- Difficulty
- intermediate
- Time
- 30 min
- Read
- 4 min
- Safety
- caution
Short answer
Run optical character recognition over it. The scan itself is just a picture, so there is no text to find. OCR reads the picture, works out the words, and stores them as an invisible layer positioned behind the image — the page still looks identical, and now you can search it, select text and copy from it. Many scanner apps and phone scanning apps do this automatically; check the setting before rescanning a filing cabinet.
Understanding the two-layer structure explains everything people find odd about searchable scans: why the page looks the same, why the selectable text is sometimes slightly wrong, and why the file gets a little larger.
Safety
Step by step
- Confirm it is actually a scan.Try to select a word. If nothing highlights, or the whole page highlights as one block, it is an image. A PDF produced by exporting from a word processor already has real text and needs nothing.
- Check whether your scanner already does it.Most scanner software and phone scanning apps have a searchable PDF or OCR option, often switched off by default. Turning it on before a big scanning job saves reprocessing everything afterwards.
- Choose where the processing happens.On your own machine for anything sensitive: a PDF editor, the scanner's own software, the built-in tools on macOS, or free command-line OCR software. An online service only for material you would be content to hand to a stranger.
- Scan at a sensible resolution.Around 300 dots per inch for ordinary printed text is the standard for reliable recognition. Much lower and characters break down; much higher produces huge files with little accuracy gain. Small print and poor originals benefit from more.
- Scan in greyscale, not colour, for text.Unless the colour carries meaning. Greyscale scans are smaller, cleaner and recognise more accurately. Pure black-and-white can lose faint text on poor originals.
- Straighten and clean before recognising.Skewed pages, shadows from a phone camera and speckled photocopies all reduce accuracy markedly. Most tools offer deskew and despeckle; use them, and rescan a page that came out badly rather than accepting a poor result.
- Set the language.Recognition uses a language model, and using the wrong one produces plausible nonsense. Set it explicitly, particularly for documents with unusual names or terminology.
- Check the result by searching.Search the finished file for a word you can see on the page. If it is not found, recognition failed or ran on the wrong pages. Expect a small error rate on any scan — it is good enough for finding documents, not for reproducing them verbatim without checking.
- Name the file properly anyway.Searchable content is a safety net, not a filing system. A file named by date, sender and subject is found in seconds; a searchable file named scan0043 is found only if you remember a phrase inside it.
If it doesn't work
OCR produces gibberish
Cause: Low resolution, a skewed or shadowed scan, an unusual typeface, or the wrong language setting — Fix: Rescan at 300 dpi in greyscale, flat and evenly lit, and set the language. Phone scans in poor light are the usual cause.
File size ballooned after OCR
Cause: The tool re-encoded the images at higher quality — Fix: Look for an option to keep the original images or to compress them. Some tools also offer to replace the image with the recognised text, which is much smaller and loses the original appearance entirely — a poor choice for anything official.
Handwriting is not recognised
Cause: Standard OCR handles printed text; handwriting needs a different technology — Fix: Some modern tools do recognise clear handwriting with mixed results. For a handwritten archive, expect to rely on file names and dates instead.
Only the first page was recognised
Cause: The tool processed the current page rather than the document — Fix: Look for a whole-document or batch option. Scanner software often defaults to per-page.
Text is selectable but copies in the wrong order
Cause: Multi-column layouts confuse the reading order — Fix: Better tools detect columns. For a complex layout, accept that searching works and copying does not.
Questions people ask
Does OCR change how the page looks?
No. The recognised text is placed invisibly behind the original image, positioned over the words it corresponds to. That is why you can select text and see the highlight land in the right place.
Is it accurate enough for legal or financial records?
For finding documents, yes. For relying on the text without reading it, no — always treat the image as the authoritative version, because recognition errors are subtle and confident.
Can I do this on a phone?
Yes. The scanning features built into modern phone camera and notes apps produce searchable PDFs, and are genuinely good for occasional pages. For a filing cabinet, a document scanner with a feeder is a different level of speed.
Is there free software that does it well?
Yes — there are mature free OCR engines, and several free PDF tools that use them. macOS also recognises text in images system-wide, which covers many casual needs without any extra software.