How to Copy Text From Scanned PDF Pages Into Word

To copy text from a scanned PDF into Word, run OCR on the file first, because a scan holds pictures of pages and no characters. Then select the passage in the recognized copy, paste it into Word as text only, join the lines that broke at the scan's margins, and read the result against the page.

The copying itself is the short part. The work is in what arrives in Word: a line break where every printed line ended, a font that belongs to nobody, a footnote dropped into the middle of a sentence. Each of those has a specific cause and a specific repair, and none of them needs the passage to be retyped.

What has to happen before text in a scanned PDF can be copied?

Recognition has to happen first. A scanned PDF stores each page as a picture, so the clipboard has no characters to take. OCR reads the picture and produces characters, either as a hidden layer behind the page image or as a separate text file. Only after that step does dragging across a line select words.

Two things can stand in for "the recognized copy", and they behave differently when you copy from them.

If you are not sure which kind of file is in front of you, drag across one word. A box drawn over the page like a crop tool means a picture. A highlighted word means text. The post on why you can't copy text from a PDF explains the image-only case and the two other causes that look like it.

A text layer that came with the file

A scanner or copier may have added a text layer on its own, without anyone asking it to. Test that layer before you rely on it. Copy one line that contains a name and a number, paste it into a blank document and compare it with the page. If the pasted line reads correctly, you can copy from the file as it stands. If letters are swapped or words run together, recognize the file again and copy from the new result.

Where recognition can come from

Any OCR tool that gives you characters will do for this job. In Schema Law you upload the PDF from your browser, the pages are read into text and tables, and you export a searchable PDF, a Word file or plain text. The list of export formats shows all eight. For copying a passage, the searchable PDF and the plain text file are the two that matter.

Copying a passage from a recognized scan into a Word document

Open the recognized copy, find the passage, select from its first word to its last, copy, and paste into Word with the text-only option. Then join the broken lines, restore any emphasis the original had, and compare the pasted words with the page image before the passage goes anywhere.

Here is the sequence in full, with an example. Tomas is drafting a demand letter and needs the notice clause from a scanned distribution agreement between two fictional companies, Pellham Tool Supply and Voss Marine Hardware. The clause is on page 14 of a 31-page scan.

1. Open the recognized file and go to the page. Tomas opens the searchable PDF and jumps to page 14. He does not search for the clause by its heading alone, because a misread heading would hide it.
2. Select by clicking, not by dragging. He clicks just before the first word of the clause, scrolls to its end, holds Shift and clicks after the last word. A Shift-click selection, in a reader that supports it, stops exactly where you put it. A drag that strays a few pixels can pick up the next clause or the footer.
3. Copy. Ctrl+C on Windows, Command+C on a Mac.
4. Paste into a scratch document first. He pastes into an empty Word document, not into the letter. Cleanup is easier to do, and easier to see, away from the draft.
5. Choose text only. After pasting he picks the Keep Text Only paste option, so the clause takes the scratch document's own font and size.
6. Join the lines. The clause arrives as eleven short lines. He removes the breaks so it becomes one paragraph, by hand for a passage this short.
7. Put back what the paste removed. The original has the defined term "Notice Address" in bold. He re-applies bold by hand, looking at the page.
8. Read it against the scan. He checks the address, the number of days and the section reference character by character.
9. Move the clean paragraph into the letter. Only now does he copy the paragraph from the scratch document into the draft, and add the page reference.

Step eight is where a wrong digit is caught. The method for it, including how to read numbers and names, is in the post on quoting from scans under how to verify a quotation against the page image.

The notice clause you just cleaned often states the address, the number of days and what follows a repudiation. Read any condition subsequent through to the end of the sentence. Copy a waiver only when the scan shows that word, because the demand letter will quote this passage as the agreement's own text.

If the passage will not select cleanly

If the selection jumps, you click at the start of a paragraph and the highlight grabs a line from another part of the page. That means the text layer's order differs from the visual order. Do not fight it. Copy the whole page instead, paste it into the scratch document, and cut out the passage there, where you can see every line that came along.

Why does text pasted from a scanned PDF break at the end of every line?

Text pasted from a scanned PDF breaks at every line end because the text layer records each printed line as its own unit. The page had no concept of a paragraph that rewraps. When you copy, each line arrives with a paragraph mark after it, and Word treats every one as the end of a paragraph.

Turn on Word's formatting marks to see it. The Show/Hide button on the Home tab, marked with a pilcrow (¶), displays a ¶ wherever a paragraph ends. In a freshly pasted clause you will see one at the end of every line. Some sources produce a bent arrow instead, which is a manual line break. Both have the same effect on the page and the same cure.

Removing the breaks from one short passage

For four or five lines, do it by hand. Click at the end of a line, press Delete to remove the mark, and type a space if the two words have run together. Work down the passage. It is quicker than setting up a replacement, and you read every line as you go.

Removing the breaks from a long passage

For a page or more, use Find and Replace. Microsoft's article on finding and replacing text in Word describes a Special menu in Advanced Find & Replace for finding and replacing paragraph marks, page breaks and other special characters. In the search box, a paragraph mark is written ^p. The trap is that real paragraph endings use the same mark, so the replacement goes in three passes:

1. Protect the real paragraph endings. Where the pasted text has an empty line between paragraphs, two marks sit together. Replace ^p^p with a placeholder that appears nowhere in the passage, such as @@@.
2. Join the lines. Replace every remaining ^p with a single space.
3. Restore the paragraphs. Replace @@@ with ^p.

Select the passage before you start and answer "No" when Word offers to search the rest of the document. A replacement that runs through a whole brief will join every paragraph in it.

If the pasted text has no empty line between paragraphs, pass one has nothing to find. In that case mark the real paragraph endings yourself first: go through the passage with the scan beside it and type @@@ at the end of each true paragraph, then run passes two and three.

What the replacement leaves behind

Check four things after joining lines.

• Split words. A word hyphenated at the end of a printed line, such as "termi-" and "nation", is now "termi- nation". Search the passage for a hyphen followed by a space and repair each one, keeping genuine hyphens like "thirty-day".
• Double spaces. A line that already ended with a space now has two. Replace two spaces with one.
• Numbered subparagraphs. Items (a), (b) and (c) may have been joined into one run if no empty line separated them. Break them apart again at each label.
• Lost emphasis. Joining lines does not remove bold or italics, but the text-only paste already did. Compare with the page.

Which Word paste option suits text copied from a scan?

Keep Text Only suits text copied from a scan. The formatting that travels with a text layer describes how the scanned page was drawn, not how your document is styled, so there is nothing in it worth keeping. Text-only paste drops it and gives the passage the font, size and spacing of the paragraph you paste into.

Microsoft's guide to controlling the formatting when you paste text describes the choices Word offers. Its descriptions, set against what each one does to a scanned passage:

Why the odd fonts appear

A text layer has to name some font and some size for its invisible characters, and those values are chosen to fit the layer over the picture. They are not a reading of the typeface on the paper. Paste with source formatting and you can get a paragraph in which one line is 10.5 point and the next 11, or a typeface your firm's template never uses. A reviewer who sees mixed fonts in a quoted clause will wonder what else was changed.

Setting the default

If you paste from scans regularly, change Word's default so you do not have to choose each time. The same Microsoft guide gives the path: File, then Options, then Advanced, and under "Cut, copy, and paste" a setting named "Pasting from other programs". Set it to Keep Text Only. Pastes from your PDF reader and your browser will then arrive plain, and you can still pick another option from the Paste Options button for a single paste.

One caution about tables

Text-only paste turns a table into lines of loose words, because the option discards tables by definition. If the passage you need includes a table, copying is the wrong route for that part. Export the page to Word or Excel so the table arrives as a table.

Selecting across columns, footnotes and page breaks in a scanned PDF

Selections in a scanned PDF follow the order in which the text layer was written, which may not be the order a reader follows. Across two columns, a footnote or a page break, the pasted passage can arrive interleaved or with page furniture in the middle. Select in smaller pieces and reassemble them in Word.

Two columns

On a two-column page, a drag from the top of the left column to the bottom of the right may select straight across both, giving you line one of the left column followed by line one of the right. Test with two lines before you copy fifty. If the order is wrong, copy each column separately and paste them one after the other. Where the whole page is recognized out of order, the post on whether OCR can handle two-column documents explains what to try.

Footnotes

A footnote sits at the bottom of the page, so a selection that runs to the last line of body text can take the footnote with it. In the pasted result, the footnote's words follow the body text with nothing to mark the change. Look for two signs: a stray superscript number that pasted as an ordinary digit ("termination3 of"), and a sudden change of subject near the end of the paste. Decide whether your quotation needs the footnote. If it does, copy it separately and present it as a footnote. If not, delete it and remove the orphaned digit.

Page breaks

A sentence that runs from the foot of one page to the top of the next has a footer, a page number and a running header between its two halves. All three come along when you select across the break.

Suppose page 14 of the Pellham agreement ends "shall be delivered to the" and page 15 begins "Notice Address by courier". Pasted in one go, the middle of the sentence reads "delivered to the Pellham / Voss Distribution Agreement 14 CONFIDENTIAL Notice Address by courier". Copy the two halves separately, or paste the lot and delete the furniture, then confirm against both pages that nothing else went with it.

Line numbers and margin text

Pleading paper and transcripts print a number beside each line. Those numbers are text on the page, and they paste as text. You can strip them after pasting, but note them first if you will cite by line. Margin stamps and handwritten notes may also be recognized and land wherever the layer placed them.

Copying a few lines versus exporting the whole scan to Word

Copy and paste suits a passage you will quote or reuse once: a clause, a paragraph, a page. Export the scan to Word when you need most of the document, when it has tables or numbered clauses you will edit, or when you catch yourself repairing line breaks for the third time in one file.

Use these questions to choose.

Count the selections before you start. Three separate clauses from one lease means three selections, three rounds of line joining and three checks, with a real chance of copying from the wrong page on one of them. One export gives you all three in a single file that you verify once.

There is also a difference in what you are left holding. A pasted passage is a fragment with no memory of where it came from, so add the page reference as you paste. An exported Word file is a full reading of the document that you can name, store beside the scan and return to.

If export is the better route, the post on OCR PDF to Word goes through the choices that shape the Word file before the conversion starts. For a first test on your own pages, Schema Law's free plan covers one document a day of up to three pages with all eight export formats, and the pricing page lists the limits for longer files.