How to Extract Text from a PDF

A PDF is built for printing, not editing. Instead of paragraphs, headings and tables, it stores where each piece of text sits on the page and which font draws it. The line breaks are simply where one line of print stopped. A running header is just more text that happens to sit at the top of every page. A table is a set of lines with text placed between them.

So any tool that pulls text out of a PDF has to rebuild the structure from positions, font sizes and spacing. Copy and paste barely tries, which is why it produces the mess it does. A good extractor works much harder at it - here's what that looks like on real examples.

1. Line breaks in the middle of sentences, and split words

Every printed line ends with a hard break, and words hyphenated at the edge of the page come out in two pieces. Pasted into another document, the text won't reflow.

Shipping  performance  improved  across  all  regions  this
quarter, with average delivery times falling from 4.2 to 3.6
days. The new routing software, introduced in July, accoun‐
ted for most of the gain, while warehouse consolidation in
the Midwest reduced handling costs by an estimated elev‐
en percent.

Shipping performance improved across all regions this quarter, with average delivery times falling from 4.2 to 3.6 days. The new routing software, introduced in July, accounted for most of the gain, while warehouse consolidation in the Midwest reduced handling costs by an estimated eleven percent.

Lines are joined back into whole paragraphs, split words are rejoined, and the extra spaces from justified text are removed. Real hyphens - like "order-tracking" or a code such as "SB-2248" - are left alone.

2. Headers, footers and page numbers mixed into the text

Running headers and page numbers are just text in a PDF, so they land in the middle of your paragraphs wherever a page ends.

is for more charging points at the busiest depots, where vans queue
at the end of each shift.
Northwind Logistics - Annual Report 2026
Page 1
The company retired forty older delivery vans during the year and
replaced them with electric models, cutting fuel spending by a third.

Driver feedback has been positive overall. The most common request is for more charging points at the busiest depots, where vans queue at the end of each shift.

The company retired forty older delivery vans during the year and replaced them with electric models, cutting fuel spending by a third.

Text in the header and footer areas of each page is left out, and a sentence cut in half by a page break is joined back together.

3. Tables that fall apart

A PDF table is lines with text placed between them. Copy it and you usually get one cell per line, with no way to tell where a row starts.

Region
Q2 (days)
Q3 (days)
Change
Northeast
4.0
3.4
-0.6
Midwest
4.5
...

| Region | Q2 (days) | Q3 (days) | Change |
| --- | --- | --- | --- |
| Northeast | 4.0 | 3.4 | -0.6 |
| Midwest | 4.5 | 3.7 | -0.8 |
| South | 4.1 | 3.8 | -0.3 |
| West | 4.3 | 3.5 | -0.8 |

The table is rebuilt row by row. The text formatter gives it to you as a Markdown table; in Format Magic's main app it becomes a real, styled Word table. Very complex tables, especially ones with merged cells, are worth a quick look after extracting.

4. Columns read in the wrong order

On a two-column page, the text you want runs down the left column and then down the right. A naive extraction can jump back and forth between the columns, mixing paragraphs together.

Format Magic works out where the columns are and reads each one top to bottom, including pages with a full-width title above the columns and layouts with a sidebar. Very irregular layouts - magazine-style pages or text wrapped around pictures - can still come out in an odd order, so give those a quick read.

• Scanned PDFs. A scan is a picture of a page, with no text inside it at all. Quick test: try to select a single word in your PDF viewer. If you can't, it's a scan, and it needs OCR (text recognition) first - many PDF apps, including Adobe Acrobat, have a "recognize text" option. Format Magic doesn't run OCR.
• Text drawn as pictures. Occasionally a PDF draws small pieces of text as images - heading numbers are a common case. There's no text there to extract, so those pieces go missing. A quick look at a few numbered headings will tell you.
• Text that copies out as gibberish. Some PDFs look perfect but use fonts that don't record which letter each shape is, so the extracted text is nonsense. If that happens, the original document (or OCR) is the way around it.

• Use the original document if you have it. The Word file a PDF was made from has the real headings, lists and tables, so nothing needs rebuilding. From Google Docs, use File > Download > Microsoft Word (.docx). From Pages, use File > Export To > Word.
• A PDF made from Word or Google Docs extracts well. These have a proper text layer, so most of the time you'll get clean text with no extra work.
• Converting the PDF to Word first isn't a reliable fix. It can help with an unusual layout, but it often adds text boxes and frames. It's worth a try if a direct extraction struggles - not as a first step.
• Spot-check the tricky parts. A quick look at a few numbered headings and any complex tables is usually enough to catch anything that didn't come through.
• Scans need OCR first. Run text recognition on a scanned PDF, then extract the text from the result.

Getting the text out is often only step one. If the end goal is a proposal, report or handout that looks professional, Format Magic can take you the rest of the way: upload the PDF to the main app, pick a design, and download a real, editable Word document - headings, lists and tables rebuilt in proper Word styles.

You wrote it, we format it. Your wording isn't rewritten; the AI handles structure and design.