Live counts as you type, plus readability, keyword density, platform limits and a writing check that highlights long sentences, passive voice and filler words. Everything stays in your browser.
Drop in your draft, open a .txt / .md file or press Sample. The text is kept in this browser only, so a refresh does not lose it.
Words, characters with and without spaces, sentences, paragraphs, reading time (238 wpm) and speaking time (140 wpm).
Type a word or character limit in Word goal. The bar turns amber above 90% and red when you are over.
The writing check marks long sentences, passive voice, -ly adverbs and filler words. Click a legend item to hide that kind.
The sentence rhythm chart shows every sentence as a bar; click one to jump to it. The limits panel checks X, LinkedIn, SMS and meta tags.
Every word counter gives you a number. The useful question is why your number differs from someone else's. Microsoft Word, Google Docs, this page and a marking rubric can look at the same essay and report different totals, because each draws the line in a different place around hyphens, numbers, footnotes and captions. This page shows exactly where the lines are drawn here, so you know what you are working with when a limit is tight.
Alongside the counts you get readability scores, keyword density, a sentence rhythm chart, live platform character limits and a writing check that highlights long sentences, passive voice, adverbs and filler words in your text. Everything runs in this browser tab.
| Metric | Rule used here |
|---|---|
| Words | Runs of letters and digits. Apostrophes and internal hyphens stay joined, so don't and state-of-the-art are one word each. Each Chinese, Japanese or Korean character counts as one word. |
| Characters | Every character including spaces, punctuation and line breaks. An emoji counts as one. |
| Characters without spaces | The same, minus spaces, tabs and line breaks. |
| Sentences | Split on . ! ? and line breaks. Any run of text with at least one word counts, even without a full stop. |
| Paragraphs | Separated by a blank line, which is how most word processors export text. |
| Reading time | Words ÷ 238 per minute, the average silent reading speed for adults reading non-fiction (Brysbaert, 2019). |
| Speaking time | Words ÷ 140 per minute, a comfortable presentation pace. |
Word can include footnotes, endnotes and text boxes in its count (there is a checkbox for it), and it counts a lone symbol between spaces, such as &, as a word; this page does not. Spacing around dashes changes things too. A difference of 1–3% between two counters is normal. If a hard limit is close, check the final figure in the software your marker or editor will use.
Both are one word, here and in Word. Hyphens inside a word keep it together, so hyphenated compounds do not inflate the count. A spaced hyphen or a dash between spaces does split words: well - known is two, well-known is one. The apostrophe is treated as part of the word, so don't, it's and O'Brien count once. Writing contractions out in full ("do not") raises your count, which helps when you are just under a minimum.
With spaces counts every keystroke, including spaces, tabs and line breaks. Without spaces counts only the visible characters. Translation and copywriting rates are usually quoted per character with spaces; some forms and older publishing systems count without. Social platforms count with spaces.
| Where | Limit | Note |
|---|---|---|
| X (Twitter) post | 280 | Spaces count. Every link counts as 23 characters, whatever its length. |
| Bluesky / Threads | 300 / 500 | |
| LinkedIn post / headline | 3,000 / 220 | Long posts are folded behind "…more" after the first few lines, so put the hook first. |
| Instagram caption | 2,200 | |
| YouTube title / description | 100 / 5,000 | |
| SEO title tag / meta description | ~60 / ~155 | Google cuts by pixel width, not characters, so these are rules of thumb. |
| SMS | 160 or 70 | See below. |
The bars turn amber above 90% of a limit and red past it. Platforms change their limits now and then, so check the app itself if you are within a few characters.
A single SMS holds 160 characters in the standard GSM-7 alphabet. One character more and the message is split into parts of 153 characters, because each part carries a small header that tells the phone how to join them. If the text contains a character outside GSM-7 — most emoji, curly quotes “ ”, many accented letters — the whole message switches to UCS-2: 70 characters for one SMS, 67 per part after that. Symbols such as € [ ] { } ~ ^ | are in GSM-7 but take two units each. The SMS row in the limits panel does this maths for you and shows the encoding and the number of messages. You are usually billed per part, so a 161-character text costs the same as two.
It is an estimate. 238 words per minute is the average silent reading speed for adults on non-fiction, from a large meta-analysis of reading studies (Brysbaert, 2019). Real speed varies a lot: technical or unfamiliar material can drop below 150 wpm, light fiction can go above 300. Use it as a ballpark for a blog post or a talk, not as a promise. Under More detail you also see a handwriting estimate at 20 wpm, handy for timed exams.
The formula is 206.835 − 1.015 × (words ÷ sentences) − 84.6 × (syllables ÷ words). Higher is easier. 60–70 is the usual target for general-audience writing. Above 80 reads as simple and direct; below 30 is dense academic or legal prose. The formula only looks at sentence length and syllables per word, so it cannot tell whether your writing is good: a page of short, punchy nonsense scores beautifully.
The Flesch–Kincaid grade level (0.39 × words/sentence + 11.8 × syllables/word − 15.59) maps the text onto US school years: 8.0 means a typical 13–14 year old can follow it. Much public-facing writing aims for grade 8 or lower; the UK government's GOV.UK style guide goes further and writes for a reading age of 9. Very short, simple text can score below zero, which just means "easier than the scale goes"; it is shown as 0.
By counting vowel groups and applying rules for silent endings, which is the standard approach. It is approximate by design: English spelling is not phonetic, so words like science, area and fire can be off by one. Over a paragraph these errors mostly cancel out, which is why the readability scores stay useful.
The chart under the editor draws every sentence as a bar: green up to 20 words, amber from 21 to 30 and red above 30, with the average as a cyan line. Long sentences are harder to follow, especially on a phone, but the real warning sign is a wall of bars that are all the same height. The variety label uses the standard deviation of the sentence lengths; "monotone" means most sentences are about equally long, and mixing in a short one usually helps. Click a bar to select that sentence in the editor.
The writing check flags a form of to be followed by a past participle as passive: "was written", "is being reviewed", "has been decided". Passive voice is not wrong. It is the right choice when the actor is unknown or does not matter — "the samples were incubated for 48 hours" is normal scientific style. Treat the highlight as a prompt to check, not an order to rewrite.
Words ending in -ly are marked because they often prop up a weak verb: "ran quickly" is often better as "sprinted". Plenty of them carry real meaning ("legally", "annually"), and obvious non-adverbs such as "only", "family" and "reply" are skipped. Filler words are intensifiers and hedges such as very, really, quite, just, actually, basically, literally, simply, definitely. In speech they are natural; in writing they usually dilute the sentence. "This is really very important" says less than "this matters".
Keyword density is how often a word or phrase appears as a percentage of all words. Search engines stopped rewarding repetition long ago, so use it as a sanity check: if your main term sits at 5% the text probably reads awkwardly, and if it appears twice in 2,000 words the page may be about something else. Around 0.5–2% usually reads naturally. Common words such as "the", "and" and "of" are hidden by default because they are always on top; switch the filter off to spot your own tics, such as overusing "however" or "just".
Set your limit in Word goal to see how many words are left. Check what your school or journal counts: the title, abstract, footnotes, tables, captions and reference list are treated differently from one place to the next. Many universities exclude the reference list and include in-text citations such as (Smith, 2024). This page counts everything you paste, so paste only the part that counts. A 10% margin either side of a limit is common, but never assume it.
Each CJK character counts as one word, the same convention Microsoft Word uses. These scripts do not put spaces between words, so counting space-separated runs would report a whole paragraph as one word. Latin text in the same document is still counted normally. Emoji count as one character each; some platforms count complex emoji (flags, skin tones, families) as two or more, so check in the app if you are right on a limit.
You can drop in .txt and .md files or paste from anything. For a .docx or .pdf, open it and copy the text: you then see exactly what you are counting, without headers, footers and page numbers that an extractor might pull in. There is no size limit. Long documents stay fast; for a text of several hundred thousand words, switch the writing check off and counting stays instant.
All counting happens in this browser tab and the page sends no text anywhere. Your draft is saved in this browser's local storage so it survives a refresh; press Clear to remove it. Once the page has loaded it also works offline. The site uses cookie-free analytics that record page views, never what you type.