How to count the words in a PDF
- Add the PDF. Tap Choose PDF or drop the file onto the page. Password-protected files ask for the password first.
- Wait for the page-by-page pass. The tool reads each page's text layer and drawing operators in turn. A progress bar shows the page it is on; a 300-page report takes a few seconds.
- Read the totals and the verdict. Five tiles show pages, words, characters, characters without spaces and images. Below them, the verdict says whether the file has a text layer, looks scanned, or is mostly images.
- Check individual pages. The per-page table shows the same counts for every page; pages with no text are greyed out. Files over 30 pages start with the table collapsed.
What it does
- Totals and a per-page table: words, characters, characters without spaces, images
- Unicode-aware word counting — works for accented, Cyrillic, Greek, CJK and Thai text, not just English
- Tells you whether the PDF has a text layer or looks scanned, and links the OCR tool when it does
- Counts embedded and inline images per page, so you can see which pages are pictures
- Opens password-protected PDFs when you know the password
- Runs entirely on your device — the only network request fetches the open-source PDF.js library
What do the counts mean?
A PDF is not a word-processor file. It stores drawing instructions — place this glyph here, that image there — and, if it was made from a digital document, a text layer that tells viewers which characters those glyphs are. This tool reads that text layer with PDF.js, rebuilds the lines from the positions of the text items, and counts what it finds. It also walks each page's operator list to count image-drawing operations, which is how it can tell a typed page from a scanned one.
| Count | What it measures | What it does not include |
|---|---|---|
| Pages | Pages in the page tree, as a viewer would show them | Hidden or deleted pages still present as unused objects |
| Words | Word-like segments of the text layer, found with the browser's Unicode segmenter; numbers count, punctuation does not | Text inside images, form-field values, comments, bookmarks |
| Characters | Unicode code points in the rebuilt text, including spaces and line breaks | Invisible or zero-width glyphs the viewer never shows |
| Characters without spaces | The same, minus every whitespace character — the figure many translation and typesetting quotes use | — |
| Images | Image-drawing operations: embedded images, inline images, image masks | Vector drawings, shading patterns, text drawn as outlines |
How many words is a page?
There is no fixed answer — it depends on font size, margins and spacing — but the ranges below are the usual rules of thumb for a Letter or A4 page, and they help you sanity-check a result. The scan row is the important one: a scanned page has no text layer, so it counts zero words no matter how much is printed on it.
| Kind of page | Typical words | Typical characters | Images | What the tool says |
|---|---|---|---|---|
| Business letter, 11–12 pt, single spaced | roughly 300–500 | roughly 2,000–3,000 | 0–1 (logo) | Has a text layer |
| Double-spaced manuscript or essay page | roughly 250–300 | roughly 1,500–2,000 | 0 | Has a text layer |
| Dense contract page, 9–10 pt, narrow margins | roughly 600–900 | roughly 4,000–6,000 | 0 | Has a text layer |
| Slide deck page | roughly 20–80 | roughly 150–600 | 1–5 | Has a text layer (or mostly images) |
| Scanned page without OCR | 0 | 0 | 1 (sometimes more if the scanner tiles) | Looks scanned — use the OCR tool |
| Scanned page with OCR layer | as the printed text, minus recognition errors | as printed | 1 | Has a text layer |
When is this useful?
- Translation and editing quotes are priced per word or per 1,000 characters without spaces; you need the figure before you ask for a price.
- Submission limits — grant applications, court filings, academic papers — are often stated in words or pages.
- Before OCR: a quick way to know whether a batch of PDFs is searchable already, and which pages are not.
- Before e-mailing: a file with hundreds of images per page is a scan and will be large; see the Compress PDF tool.
If you need the words themselves rather than a count, the PDF to text tool extracts them to a .txt file with the same line reconstruction.
What are the limits?
- It counts only the text layer. A scan without OCR counts zero; a scan with a bad OCR layer counts the OCR's mistakes.
- Line reconstruction is geometric. Multi-column layouts are read column by column when the columns are cleanly separated, but tables and text wrapped around images can interleave, which does not change the word count much but can split or join a word at the boundary.
- Form fields, annotations, comments and bookmarks are not part of the page text and are not counted.
- Password-protected files need the password. Files that PDF.js cannot parse fail with a message.
- Very large files are processed page by page and the page's resources are released before the next one, but a 1,000-page scan still takes a while on a phone.