Count the words in a PDF
Count words, characters and pages in a PDF with a per-page breakdown, reading time and a per-word fee estimate. Nothing leaves your device.
- Runs in your browser
- Files are never uploaded
- Free, no account
Your manuscript never leaves this tab — here is how to check
This page has no upload step and no server behind it. Two ways to confirm that yourself, neither of which requires trusting us:
- Turn off Wi-Fi or put the device in aeroplane mode after this page has loaded, then run the tool. It still works.
- Open your browser’s developer tools, switch to the Network tab, clear it, and process a file. No request appears.
The reason this matters: in 2025 the PDF service pdf.net was found exposing roughly 3.5 million user-uploaded documents, and formatter sites JSONFormatter.org and CodeBeautify leaked over 80,000 saved snippets containing live cloud credentials. Those are architectural failures — a file that is never sent cannot be exposed that way.
Why counting words in a PDF is harder than it sounds
A word processor knows what a word is, because you typed it. A PDF does not. It stores positioned runs of glyphs — “draw these three characters at x=112.4, y=703.8 in this font at this size” — and a single visual word is frequently split across several runs because of kerning, ligatures or a change of font.
Recovering words means reading every glyph run on a page, measuring the horizontal gap between consecutive runs, and deciding whether that gap is a space or just kerning. This tool uses the run height as the yardstick: a gap wider than about 16% of the line height becomes a space, and a vertical jump of more than half a line height becomes a line break. It is the same heuristic a PDF reader uses when you select text with the mouse.
Expect a small disagreement with Word
Counts from a PDF and counts from the source document will differ by a percent or two, in both directions, for reasons that are inherent rather than fixable:
- Headers, footers and page numbers are text on the page and get counted.
- A word hyphenated across a line break may be counted as two.
- Text inside images — a chart label, a scanned stamp — is not counted at all.
- Footnote markers, table cells and captions all count as words here.
- Soft hyphens, ligatures and non-breaking spaces are normalised differently by different writers.
Use the figure for quoting, planning and pricing. For a legally binding count — a thesis limit, a court filing — count in the source document.
Counting CJK
Chinese, Japanese and Korean do not separate words with spaces, so whitespace counting returns near-zero for a document that plainly has thousands of words in it. The convention in those languages, and in translation pricing for them, is to count characters.
Every kana, CJK ideograph, compatibility ideograph, half-width katakana and Hangul syllable is counted as one unit. Latin text in the same document is still counted by whitespace, so a bilingual document totals sensibly rather than being wrong in one direction or the other.
Standard pages and per-word rates
Translation and editing are priced by the word almost everywhere, but the unit of a “page” varies by market:
| Market | Standard page |
|---|---|
| UK / US / general | 250 words |
| Germany (Normzeile) | 55 characters per line, 30 lines — 1,650 characters |
| Poland (strona rozliczeniowa) | 1,800 characters including spaces |
| Italy (cartella) | 1,500 characters including spaces |
| Russia | 1,800 characters including spaces |
The rate tab shows both the 250-word and the 1,800-character page counts alongside the fee at your own per-word rate, with the arithmetic written out so you can paste it straight into a quote.
Reading and speaking time
The default of 238 words per minute is the mean silent reading rate for adults reading English non-fiction, from Brysbaert’s 2019 meta-analysis of 190 studies covering nearly eighteen thousand participants. It is a mean, not a target: the same analysis puts the range for ordinary adults roughly between 175 and 300.
Reading aloud is much slower. The 130 words per minute used for the spoken estimate is a comfortable presentation pace — audiobook narration sits around 150, and anything above 180 starts to sound rushed. If you are timing a conference talk, use the spoken figure and add time for slides.
Scanned PDFs, honestly
If the tool reports almost no words, the document has no text layer: it is a photograph of a page. Reading it would require optical character recognition, which needs a trained model — several megabytes that we would have to fetch from a server, and doing that would break the one promise this site is built on. So we do not offer OCR, and we say so rather than returning zero and letting you assume the file is empty.
The quickest way to confirm it yourself: open the PDF in any reader and try to select a word with the mouse. If nothing highlights, there is no text there to count.
Questions
Why does the count differ slightly from Word?
A PDF has no notion of a word — it stores positioned glyph runs, and the word boundaries have to be inferred from the gaps between them. Hyphenation across line breaks, ligatures, headers, footers and page numbers all shift the total by a percent or two. Use the figure for quoting and planning, not for a legally binding count.
How are Chinese, Japanese and Korean counted?
By character, not by whitespace, which is the convention those languages use. Each CJK ideograph or kana counts as one unit and any surrounding Latin text is still counted as whitespace-separated words, so a mixed document totals correctly.
What is a standard page?
A billing unit used by translation agencies: 250 words in the English-speaking world, or 1,800 characters including spaces in Germany, Poland and much of central Europe. Both figures are shown so you can quote in whichever convention your client uses.
It says my PDF has almost no words — why?
Because it is a scan: a photograph of a document with no text layer beneath it. Reading it would need OCR, which requires a model we would have to fetch from a server, and that would break the promise that nothing leaves your device. Try selecting a word in a PDF reader — if nothing highlights, there is no text to count.
Where does 238 words per minute come from?
It is the mean silent reading rate for adults reading English non-fiction reported by Brysbaert’s 2019 meta-analysis of 190 studies. Dense technical or legal text runs closer to 150, and reading aloud sits around 130, which is the figure used for the spoken estimate.