PDF Page and Word Count
Count the pages, words and characters in a PDF, page by page. It reads the real text layer rather than guessing from file size, tells you which pages carry no text at all, and warns you when a document turns out to be a scan with no readable text in it. Nothing is uploaded.
- Nothing uploaded
- Works with no signal
- No sign up
- Free to use
Nothing loaded yet.
Results
Load a PDF to see the counts.
Per page
Every page, its counts and what kind of page it is.
| Page | Words | Characters | Without spaces | Unique | Avg length | Kind |
|---|
Word counts from a PDF are always an approximation of what a word processor would say, because a PDF stores positioned glyphs rather than sentences. The article below explains where the differences come from.
Your PDF never leaves this device.It is opened and read by your browser, the counting happens here in the page, and nothing is written anywhere. No file is sent to a server, so nothing is stored on one. Open the network tab in your browser tools and load a document: there is no upload request to see.
Counts from the text layer, not a guess
The words are read from the document itself, the same text you would get by selecting and copying. A page with no text is reported rather than counted as empty prose.
How to use this tool
Drop a PDF on the box at the top. Every page is opened and its text layer pulled out, which takes a moment on a long document because each page is read separately.
The summary gives you the totals, the bars show how the words are spread across the pages, and the table gives every page in detail. If you need the numbers elsewhere, the per page counts download as a CSV.
The two settings only change how the results are reported. Reading speed scales the time estimate, and the sparse threshold decides how few words a page needs before it is called sparse rather than a text page.
Why a PDF word count is never exact
Open the same document in Word and here, and the numbers will differ. That is not a bug in either one, and understanding why tells you which number to trust for what.
A word processor knows what a word is, because you typed it. A PDF does not. It stores glyphs at coordinates: the letter c at this position, the letter a next to it, and so on. Where one word ends and the next begins has to be inferred from the gaps. Most of the time that inference is right, and occasionally it is not, particularly in justified text with stretched spacing or in tables where columns sit close together.
There is also a scope difference. This tool counts everything with text on the page, which means running headers, footers, page numbers, table contents and figure captions all contribute. A word processor typically excludes headers and footers from its main count. On a fifty page report with a header on every page, that alone is a few hundred words of difference before anything else.
So use this for the practical questions: roughly how long is this, how is the content distributed, which pages are dense, is there any text here at all. For a contractual word limit, count in the program the document was written in.
The hyphen problem, and what the tool does about it
Justified text breaks long words across lines with a hyphen. Extracted naively, compaction split across two lines becomes two tokens, compac and tion, and the count goes up by one for every hyphenated break in the document.
So fragments are rejoined before counting: a piece ending in a hyphen followed by one starting with a lowercase letter is treated as a split word and glued back together. That single rule fixes almost all of it.
The reason it checks the case of the next letter is that some hyphens are real. Anglo-Saxon, state-of-the-art and twenty-four all contain hyphens that belong to the word, and a capitalised continuation is a strong signal the hyphen is part of the name rather than a line break. It is not perfect, because a genuine hyphenated compound in lowercase broken across a line is indistinguishable from a line break hyphen, but that case is rare and the error is one word.
A worked check you can repeat
Take the sentence: The quick brown fox jumps over the lazy dog. That is nine words, 43 characters counting the eight spaces, and 35 without them. Eight of the nine words are unique, because the appears twice, and the average word length is 3.9 letters, being 35 letters over 9 words.
Now a document. Two pages of 120 words each, one blank page, and one page carrying only the number 7. That is 241 words over four pages, which averages 60.3 a page. One page is empty, one is sparse because a single number is fewer than the ten word threshold, and two are text pages. At 230 words a minute the reading time is about a minute.
Note that the unique word count for that document is 2, not the sum of the per page uniques. Unique means unique across the whole document, so repeating the same word on every page still counts once.
All of this was checked against an independent implementation before publishing: sixty hand written cases covering punctuation, numbers, accented and non Latin scripts, hyphen repair and page classification, then twenty five thousand generated cases confirming that unique never exceeds total, that character counts without spaces never exceed the full count, and that rejoining line fragments can never create a word that was not there.
Mistakes that catch people out
Assuming a zero count means the file is broken. It almost always means the file is a scan. The page looks like text to you because your eyes read pictures of letters. Nothing in software can count those without optical character recognition first.
Comparing against a word processor and trusting the wrong one. They are answering slightly different questions. For a submission with a word limit, the authoring program is the one that counts.
Reading the reading time as a schedule. It is words divided by a speed. A dense specification with tables and clause references takes far longer than the estimate, which is why the slower setting is there.
Treating an empty page as a problem. A full page drawing legitimately has no text. On a drawing set, most pages being empty is the correct answer rather than a fault.
Expecting the order of text to match reading order. Extraction follows the order the glyphs were written into the file, which for multi column layouts is often not the order you would read them in. That does not affect the counts, but it does affect anything you paste elsewhere.
Common questions
Is my PDF uploaded anywhere?
No. The file is opened and read by your own browser and the counting happens on this page. Nothing crosses the network, so there is nothing on a server to store or delete.
Why does the count differ from Word?
A PDF does not store sentences. It stores glyphs at positions, and the reader infers where words begin and end from the spacing. Headers, footers, page numbers, table cells and figure captions all get counted here because they are text on the page, while a word processor may treat some of them separately.
Why does it say my document has no words?
Because it is almost certainly a scan. The words are pictures of letters rather than letters, so there is nothing to count, search or select. Running the file through optical character recognition adds a real text layer.
What does a sparse page mean?
A page that yielded fewer words than the threshold you set, ten by default. That is usually a full page drawing or photograph with just a caption, a section divider, or a scanned page where only a stray page number was recognised.
Are words split across two lines counted twice?
No. A fragment ending in a hyphen followed by one starting with a lowercase letter is rejoined before counting, so compac- and tion become one word. A capitalised continuation such as Anglo- Saxon is left alone, because that hyphen is part of the word.
How accurate is the reading time?
It is the word count divided by the reading speed you pick, nothing more. Dense technical material with tables and figures takes far longer than the number suggests, which is why the slower setting exists.
Related PDF tools and reference
Reference: ISO 32000, the standard that defines how text is stored in a PDF.