How to count words and characters accurately (and why tools disagree)

A word count is a count of runs of letters and digits separated by spaces. That sounds unambiguous until you meet 1,000.50, well-known, https://example.com or a Japanese sentence with no spaces at all — and every counter resolves those cases its own way. The totals you get from two tools on the same text are usually within about 1% of each other, which is fine for a blog post and not fine for a hard 500-word cap.

The quickest route here: paste the text into Word Counter. It shows words, characters with and without spaces, sentences, paragraphs, lines, reading time and your most repeated words, and it updates as you type. The text stays in the browser, so a confidential draft never leaves your device.

Where to count, and what each place includes

Where	Good for	Watch out for
A browser counter like this one	Any text you can paste, live totals, character totals with and without spaces	You have to paste; formatting is discarded
Microsoft Word (desktop)	The document you are actually submitting	Text boxes, footnotes and endnotes are counted only if you tick the box
Google Docs	Shared drafts, counting a selection	Headers, footers and footnotes are excluded
The platform's own counter	X, LinkedIn, an SMS gateway	Each platform counts by its own rules, not yours

Microsoft Word. The running total sits in the status bar at the bottom of the window. Click it, or use Review → Word Count, to open the full box with characters, paragraphs and lines. Select a passage first and the status bar shows "X of Y words" — the selection and the whole document. The Word Count box has a checkbox for including text boxes, footnotes and endnotes; whether your limit includes them is a question for whoever set it. Word for the web gives an approximate count only: it skips text in text boxes, headers, footers and SmartArt graphics, so open the file in the desktop app before submitting anything to the last hundred words.

Google Docs. Open Tools → Word count. Google's documentation is explicit about the scope: unless you select part of the document first, the count "applies to everything except headers, footers, and footnotes". The same box has a Display word count while typing option that pins a small live total to the bottom-left corner.

What counts as one word here

This counter treats a word as a run of letters or digits, and an apostrophe, hyphen, underscore or dot between two such runs does not split it. So don't and well-known are one word each. Chinese, Japanese and Korean characters are counted individually, which is the usual convention for those scripts.

Take this sentence: Don't forget the well-known 1,000.50 figure for 2019–2021. This tool reports 10 words, because the comma inside 1,000.50 and the en dash inside 2019–2021 are not word-joining characters, so each of those splits into two. A word processor that treats both as single tokens reports 8 for the same sentence. Neither is wrong; they are answering slightly different questions. That is the whole story behind "my counts don't match".

If the limit is strict, stay 2–3% under it. On a 1,500-word essay that is 30–45 words of headroom, which costs you nothing and covers any counter your marker happens to use.

What counts as one character

Character Counter counts what a reader sees as one character — a grapheme — rather than what a computer stores. That distinction matters more than it sounds:

• é is one character on screen. In UTF-8 it is two bytes (C3 A9), and it can also be stored as two code points (e plus a combining acute accent). All of those still count as 1 here.
• 👍🏽 is one character on screen, two code points, and four UTF-16 code units.
• 🇫🇷 is also one on screen and two code points, because a flag is a pair of regional indicator letters.

The counter also gives you the total without spaces, which is what most translation and copywriting rates are quoted against.

The limits you are actually writing against

• SMS. One message holds 160 characters in the basic GSM 7-bit alphabet, because 140 bytes hold 160 septets. Add an emoji or a character outside that alphabet and the whole message switches to UCS-2, where one message holds 70 characters. Once a message is split into a concatenated series, part of each segment carries the joining header, leaving 153 characters per segment in GSM 7-bit and 67 in UCS-2.
• X. The limit is 280 weighted characters. Latin letters, punctuation and common symbols weigh 1; emoji, CJK characters and other Unicode weigh 2. Every link is replaced by a t.co address and counts as 23, however long the original was.
• Meta descriptions. Google publishes no character limit; snippets are shortened to fit the available width, so the cut-off depends on the letters you used. Roughly 120–160 characters survives intact in most results.

From words to minutes

For a blog post, a script or a conference talk, Reading Time turns the word count into a duration: 238 words per minute for silent reading, 150 for speaking aloud. A 1,500-word article is about 6 minutes 18 seconds to read and 10 minutes to say out loud. Treat those as planning figures — a dense technical page is slower, and your own speaking pace may not be 150.

Why a count comes out wrong

Double spaces and trailing whitespace. They do not change the word count, but they inflate the character count, and they are invisible. Text Cleaner collapses runs of spaces and strips trailing ones before you measure.

Text copied out of a PDF. A PDF stores positioned lines, not paragraphs, so a copied page arrives as dozens of short lines. The word count survives, but the paragraph and line counts are meaningless until you rejoin the lines — see How to clean up text copied from a PDF. For a whole document, PDF to Text extracts every page in one go.

Non-breaking spaces. Copied from a web page or a word processor, they look like ordinary spaces but are a different character. They still separate words, so the word count is right, and they still count as characters.

Counting a selection by accident. In both Word and Google Docs, a live selection changes what the counter reports. Click into the text to clear it before reading the total.

Reference lists and footnotes. Most disagreements about length are not about the algorithm but about whether the bibliography, the abstract and the footnotes sit inside the limit. Ask, then count the right region.

Questions people ask

How many pages is 1,500 words? With the common academic layout — 12 pt, double-spaced, 1-inch margins — a page holds roughly 250 words, so about six pages. Single-spaced it is nearer 500 words a page, so about three. Your font and margins decide the real answer.

Why is my count slightly different from Word? Edge cases: numbers with separators, URLs, email addresses and dashes without spaces. On ordinary prose the gap stays under 1%.

How do I count the words in a PDF? Extract the text with PDF to Text and paste it in. If the extraction comes back empty, the PDF is a scan — an image of text — and needs OCR first.

Is my text uploaded anywhere? No. Counting happens in your browser, and the text disappears when you close the page.

How do I count sentences rather than words? Sentence Counter splits on terminal punctuation and skips the common abbreviations (Mr., e.g., etc.) that would otherwise create phantom sentences.

Next steps

• Count characters with and without spaces: Character Counter
• Estimate reading and speaking time: Reading Time
• Tidy a draft before measuring it: Text Cleaner, and the guide How to clean up text copied from a PDF