This produces real, editable Word text — not a layout-perfect replica. A PDF stores fixed positions on a page; Word stores flowing text. Independent testing of paid and free converters in 2026 reports 75–96% formatting accuracy even at the best of them, and every one struggles on multi-column layouts and borderless tables. This tool makes the same honest trade — strong on ordinary reports and prose, weaker on complex layouts — and shows you exactly which font sizes it read as headings so a wrong guess takes a dropdown, not a rewrite.

No upload, no account, no daily limit. The PDF is read by your browser and the Word document is built by your browser — not a privacy policy, a fact you can confirm in the network panel. Competitors capping the free tier at one to three conversions a day do so because they pay for server time; there is no server here.

PDF to Word

Local

Choose a PDF, or drop one here Nothing is uploaded — the file is read in this tab

Reading uses PDF.js; writing uses the docx library. Both are fetched from a CDN only when needed — your document is never part of that request.

What carries over, and what does not

A PDF stores fixed glyph positions; a Word document stores flowing paragraphs with styles. Reconstructing one from the other means inferring structure that was never actually stored:

Element	Result	Why
Paragraph text	Reliable	Positioned glyphs rejoin cleanly into lines and paragraphs by their coordinates
Headings	Inferred	Guessed from font size, shown in the editable map above — genuine Word heading styles, usable for a table of contents
Bold and italic	Inferred	Read from the font's family name and carried into real Word bold/italic formatting — a heuristic, not a certainty: a PDF whose font name gives no hint of weight despite looking bold on the page will be missed
Bullet and numbered lists	Usually	Detected from the bullet character or leading number, output as genuine Word list formatting
Tables	Not attempted	A PDF table is ruled lines with text positioned between them, not structured data — rebuilding the grid means guessing, and the guess breaks on merged cells. Use PDF to Excel if the source is a real table
Multiple columns	Poor	Reading order is not recorded in a PDF, so column two often interleaves with column one
Scanned pages	Nothing	Images of text with no text layer — run PDF OCR first, or use PDF to Text, which has the same OCR fallback built in
Fonts, margins, exact spacing	Not preserved	Word uses your document's default styles; this is a text-and-structure conversion, not a visual clone

How the free tools actually compare

Checked directly against each product's own published free-tier terms and independent 2026 testing, not marketing pages. Free-tier specifics change often and vary by account status, login state and market, so treat exact numbers below as directional rather than a fixed contract — checked August 2026:

Tool	Free limit	Where your file goes
This tool	No fixed cap	Never leaves your browser
iLovePDF	Daily task cap, file size cap around 15–25MB	Uploaded to their server
Smallpdf	Daily task cap	Uploaded to their server
Adobe Acrobat Online	Varies by login state and account — not a flat daily number	Uploaded to Adobe's server

Independent testing published in 2026 put formatting accuracy at 75–96% across these tools, Adobe scoring highest on complex layouts and every tool losing ground on multi-column pages and borderless tables — a limit of reconstructing flowing text from fixed positions, not simply a free-tier restriction.

Getting the best result

Check the heading map before downloading, not after. The map lists every font size found and what it became. A document with a large pull-quote or a caption set at heading size will confuse the guess — a thirty-second glance and a dropdown change fixes it before it is baked into the file.

Expecting a two-column layout to come out in order. PDF does not store reading order across columns, so a two-column paper or newsletter usually interleaves line by line. Check the middle of the document, not just the first paragraph.

Trying this on a scan. A scan is an image, not text — there is nothing here to extract. Run it through PDF OCR first, which adds a text layer, then convert.

Frequently asked questions

Is my PDF uploaded anywhere?

No. The file is read using your browser's own File API and the Word document is assembled the same way — neither step involves a network request carrying your document, which you can confirm in the network panel of developer tools. The only requests are for the two libraries themselves, fetched once each and containing nothing of yours.

Is there a file size or page limit?

There is no cap imposed by this tool. Hosted converters ration free use because each conversion costs them server time; nothing runs on a server here, so there is nothing to ration in that sense — but every page's text is held in your browser's own memory while converting, so a genuinely enormous document is bounded by your device, not by a rule this tool enforces. For a typical report or manuscript, that ceiling is far beyond anything you would hit in practice.

Will the layout match the original exactly?

No, and no free or paid tool checked in 2026 achieves this reliably either — independent testing put the best of them at 96% "formatting accuracy," with complex layouts pulling every tool down further. This one converts text and structure — headings, bold, italic, lists — into genuine Word formatting rather than trying to clone the exact page appearance.

Why are my headings wrong?

A PDF does not record which text is a heading — it is inferred from font size, and a document with an unusually large pull-quote or caption can confuse that guess. The heading map lists every size found and what it will become; change any assignment before downloading and the result updates.

Can it read tables?

Not reliably, and this tool does not attempt it — a PDF table is ruled lines with text positioned between them, not structured data, and guessing the grid breaks on merged cells and wrapped text. If your source has a real table, PDF to Excel is built specifically for extracting that grid.

Why did a scanned PDF produce nothing?

Because a scan is an image of a page with no text layer to read. Run it through PDF OCR first, which adds a searchable text layer without changing how it looks, then convert the result here.

Does it handle Arabic, Chinese or other scripts?

The characters themselves, yes — a PDF stores a map from its internal codes to actual characters, and that map is read correctly regardless of script: Latin, Cyrillic, Arabic, Hebrew, Chinese, Japanese and Korean all extract with the right letters. Reading order for right-to-left scripts is a separate, harder question: this tool uses PDF.js's own text-direction data to order right-to-left lines correctly in the common case and marks them as right-to-left in the Word document, but PDF.js's own issue tracker documents PDFs where the underlying extraction order does not match human reading order at all. Treat right-to-left output as needing a check in Word before you rely on it, the same caution the status message gives when it is detected.

What about password-protected PDFs?

An encrypted PDF cannot be read without its password, and this tool says so rather than failing silently. Remove the protection in a PDF reader first, then convert.

Does the output stay editable in Word?

Yes — headings, bold, italic and lists are genuine Word formatting, not an image or a fixed layout. Open the result in Word, Google Docs or LibreOffice and edit it like any other document.

Related tools

See the full list of Document Tools, or try:

PDF to Markdown — the same extraction, output as plain-text Markdown you can edit before converting further
PDF to Excel — when the source is a real table, not prose
PDF OCR — add a text layer to a scan before converting it here
Markdown to Word — the same output format, starting from Markdown instead of a PDF
PDF to Text — plain text with built-in OCR fallback for scanned pages

PDF to Word Converter: Free, Real DOCX, No Upload