How to convert a PDF to Word and check the formatting

Choose editable words or faithful page images, then check the file you will actually use.

Free toolPDF to WordMake your PDF editable in Word. Review every page, correct text, read scans with local OCR and save a DOCX.Convert a PDF to Word

Start by checking whether the PDF contains text

Try selecting a sentence in your PDF reader. If you can copy readable words, use text conversion. If the page is a picture, it needs OCR.

A document can contain both. Check several pages, including any appendices or scanned signatures. A successful first page tells you little about the last.

What you need	Choose	Tradeoff
Rewrite a report or letter	Editable DOCX	Basic structure can survive. Precise spacing may change.
Edit a printed scan	OCR, then editable DOCX	Recognition can confuse letters and numbers. Proof-reading is essential.
Show the exact visible page in Word	Page pictures inside DOCX	The text is part of the picture and cannot be edited.
Reuse words in another application	TXT or a simple text conversion	Document design and pictures are removed.
Keep a PDF but search its scans	Searchable PDF	You keep PDF pages rather than gain a Word document.

Pick a method that fits the document

The best method depends on your input, your available software and where you want the document processed.

Use a local browser converter for a document you want to keep on your device

floi reads the PDF in your browser and builds a DOCX locally. Review the reconstructed words, choose pages and run local OCR for pages without text.

Its editable output favours usable paragraphs and simple tables. Page images preserves a designed page’s appearance. Browser memory sets the practical size limit, especially on a phone.

Open a text PDF in Microsoft Word

In desktop Word, choose File, Open and select the PDF. Accept the conversion prompt, review the document and save a DOCX.

Microsoft’s PDF opening guide explains that this works best for mostly text documents. It also explains why tables and other structures may need repair.

Use Google Docs when you already keep the file in Drive

Open the uploaded PDF with Google Docs, then save it as Microsoft Word if you need a DOCX. This sends the document to Google.

Google’s conversion guidance recommends clear, upright files of 2 MB or smaller. Its OCR guidance warns that tables, columns and some other structures may not transfer.

Try a dedicated converter for a difficult layout

Adobe Acrobat and other dedicated conversion engines offer another route when a complex document needs more reconstruction. Test a representative page before paying.

Adobe’s online converter receives an uploaded PDF. Check its current account requirements and limits. Keep the original so you can compare the result.

Formatting is reconstructed, not unlocked

A PDF does not have to describe a heading or a table cell. It can draw letters and rules at separate coordinates.

A converter estimates the relationships between those marks. That is why two files that look alike can convert differently.

Paragraphs: a short line could end a paragraph, introduce a list, or continue in another column.
Tables: aligned text can suggest cells, but merged cells and irregular rows need more interpretation.
Fonts: replacing a missing font changes character widths and can move line and page breaks.
Pictures: text over artwork can become overlapping objects or move into the document’s reading flow.

Keep an editable result when you need to change the content. Keep the PDF or a page image when the appearance matters more.

A sample that exposes the tradeoffs

floi’s three-page synthetic sample deliberately combines a report, reading columns and a scan. You can use the same file to compare methods.

Open the sample in the tool. Check which pages have embedded text.
Compare the report’s table and the two-column page with Original.
Save Editable Word before OCR, then recognise the scan and save again.
Open both DOCX files. Compare whether you can place a text cursor inside the scan’s words.

This exercise checks editability and reading order. It does not measure every PDF, prove OCR accuracy or guarantee how another word processor paginates the document.

What our synthetic documents produced

On 3 November 2026, we checked the sample and additional CV, invoice and landscape fixtures. These results describe those documents, including the formatting that changed.

Input	Observed editable output	What changed or needs checking
Sample report	One native Word table with 4 rows and 3 columns. Bold and italic runs remained. The raster picture was included.	Spacing changed. The picture moved into the reading flow.
Sample with two columns	The left column finished before the right started. The page number followed both.	The columns became one continuous reading flow.
Sample scan	No editable words before OCR. Local English OCR produced 82 words in 9 editable paragraphs.	The chart’s graphics were not reconstructed. Recognition still needs proof-reading.
Synthetic CV	The name and section text remained. All 3 bullets became native Word list items.	Spacing and line wrapping still depend on the word processor.
Synthetic invoice	A native table with 16 cells. The total remained `£3,125.50`.	Cell alignment was inferred from the text positions.
Landscape presentation	The words stayed editable and the Word section retained landscape orientation.	A slide’s precise design is better served by Page images.

Method: PDF.js 6.3.289 read the synthetic PDFs; checks inspected the DOCX text, tables, lists and sections. OCR ran in Chrome 156.0.8078.80 on an Apple M1 Max. Representative DOCX files were rendered with LibreOffice and visually inspected.

The source generator and repeatable checks are in the project’s research notes. This is a fixture test, not a conversion-accuracy score or a speed comparison with another service.

Check these details before sharing the DOCX

Inspect the saved file in the application your reader will use. A browser preview cannot establish the final pagination of every word processor.

Check	What to look for
Important text	Names, dates, negative signs, decimal points, currency symbols and leading zeros.
Reading order	The left column should finish before the right begins. Headers should not interrupt sentences.
Tables	Each figure belongs to the correct row and column. No merged label has shifted a value.
Pictures and scans	Every needed page is present. Check whether a page contains words or a non-editable picture.
Pagination	No clipped text, extra blank page, broken heading or detached table row.

If the conversion is not useful

Go back to the source document when you can. Exporting the original Word, layout or spreadsheet file is usually more reliable than reconstructing a PDF.

Choose Clean text if formatting is getting in the way of rewriting.
For one difficult table, rebuilding the cells may take less time than fixing a large conversion.
For a scan, improve rotation and legibility before OCR. Recognition cannot recover detail missing from the image.
To fill answers into a form, use a PDF form filler and keep its original layout.

PDF to TextExtract and clean the words without creating a Word document.Searchable PDFKeep the scan and add a searchable text layer.Fix PDF text extractionUnderstand broken lines, missing characters and reading order.