How to make a scanned PDF searchable

A scan looks like a document but behaves like a photograph — you cannot search it, select it or copy from it. Making it searchable puts words invisibly under the picture so the page looks the same and reads like real text.

Last reviewed October 2026. If a page like this is ever rewritten, the date moves with it.

How to tell which kind of PDF you have

Open it in any PDF reader and try to select a word. If a word highlights, the document already has text and there is nothing to add — its words can already be copied exactly. That covers every digital PDF (exported from a word processor, a browser or this site).

If the cursor selects nothing, or highlights the whole page as one block, the page is a picture and recognition is the right tool. A flatbed scan, a phone photo saved as PDF and a fax all behave this way.

What searchable adds — picture plus words

The picture stays exactly as it was — crest, letterhead, handwriting around the edges — and a set of words is placed invisibly underneath it at the right positions. Searching finds those words, selecting copies them, and a screen reader can read them.

Only the picture and the words are added; the page is not rewritten into paragraphs, columns are not rearranged, and the file does not grow much — the text layer is tiny compared to the images.

Accuracy is about the picture

Recognition reads the shape of each letter, so a clean, straight, well-lit scan comes back close to perfect and a dim, crooked phone photo does not. The same engine already reads printed Sinhala, Tamil and English photographs elsewhere on this site, and a scanned PDF is just a set of such photographs.

Handwriting in any language is not reliably readable — that is a different and much harder problem, unaffected by any setting on this page.

Questions

Is my scan uploaded?

No. Each page is rendered and recognised in your browser tab and never sent anywhere — you can verify in the network tab or disconnect after the page loads and watch it still work once the models are cached.

My PDF already has text. Will this break it?

No. The picture stays. A page that already had text keeps it and gains an unnecessary layer underneath; a scanned page gains the ability to be searched.

How long does it take?

The first run downloads the engine and language models (about 6 MB for English, more for Sinhala/Tamil) — once, cached afterwards. Then a few seconds per page is typical.

Can I edit the recognised words?

The text layer is for searching and copying. To get editable text, copy the recognised words from the box shown and paste them into Text to PDF, which builds a fresh, editable document.

All tools