Type this into two different word counters: "The candidate—a former mayor—won reelection." One says 7 words. The other says 5. Neither is broken. They just disagree about what a word is when an em dash glues two of them together, and that one design choice ripples through every count you'll ever run.
Most of the time the numbers match and you never notice. The gap shows up exactly when you're up against a hard limit — a 280-character tweet, a 155-character meta description, a 650-word essay cap — and being three words off actually matters.
What a word counter actually counts
A word count is almost always the number of whitespace-separated chunks in your text. The tool splits on spaces, tabs, and line breaks, throws away the empty pieces, and counts what's left. That's it. "Hello world" is 2. "Hello world" with four spaces between is still 2, because empty chunks get dropped.
Character counts are simpler but come in two flavors: with spaces and without. "Hello world" is 11 characters with spaces, 10 without. Punctuation always counts — the period ending this sentence is a character. When a tool shows one character number without saying which, assume it includes spaces, since that's what platforms like X and SMS count against you.
The disagreements live in the edge cases:
- Hyphenated words: "mother-in-law" is one whitespace chunk, so almost every counter calls it 1 word — even though you could argue it's 3.
- Em dashes with no surrounding spaces: "mayor—won" reads as one chunk, so it undercounts.
- Contractions and numbers: "don't" is 1 word, "3.14" is 1 word, "9 a.m." is 2.
- Markdown and HTML: paste
**bold**and some tools count the asterisks as part of the word.
The limits you'll actually bump into
X/Twitter: 280 characters. Two catches. Any link counts as exactly 23 characters no matter how long the URL is, because X wraps it in a t.co shortener. And CJK characters (Chinese, Japanese, Korean) count as 2 each, so a Japanese tweet hits the wall around 140 characters. A generic counter that just measures string length will tell you you're fine when X says you're over.
Meta descriptions: about 155 characters. Google truncates the snippet in search results at roughly 920 pixels on desktop, which works out to ~155–160 characters for average text, and less on mobile. It's pixel-based, not a hard cap, so wide letters eat the budget faster. Aim for 150 and you'll rarely get cut off. Title tags follow the same logic at ~60 characters.
SMS: 160 characters — until it's 70. A plain-text message using the standard GSM-7 alphabet fits 160 characters in one segment. Drop in a single emoji or a curly "smart quote" and the whole message switches to Unicode (UCS-2) encoding, where the limit collapses to 70 characters per segment. That's why one stray emoji can turn a one-part text into a two-part text and double your send cost.
Essays: the target is the target. The Common App caps the personal essay at 650 words. Most "500-word" assignments tolerate about ±10%. Here word count is the metric graders paste your text into, so match their tool's behavior, not your word processor's.
Why two counters give different numbers
The splitting rule is the whole story. A counter that splits only on whitespace treats "mayor—won" as one word. A counter that splits on any non-letter boundary — a regex like \b\w+\b — treats the same string as two, and also splits "don't" into "don" and "t." Microsoft Word uses its own rules again, which is why the count in Word rarely matches a browser tool to the digit.
Line breaks, trailing spaces, and pasted formatting add noise. Copy from a PDF and you often drag in non-breaking spaces and soft hyphens that a counter may or may not treat as separators. None of these tools is wrong; they've each drawn the boundary in a slightly different place. The fix is to pick one tool and use it consistently for a given limit — ideally the one your destination platform uses.
Word Counter shows words, characters with and without spaces, sentences, and paragraphs at once, and it runs entirely in your browser — nothing you paste is uploaded or stored, which matters when the text is a draft contract or an unpublished essay.
Reading time, and why it's a range
Reading-time estimates come from one calculation: word count divided by words per minute. Adults read silent prose at roughly 200–250 words per minute; a large 2019 meta-analysis put non-fiction reading around 238 wpm. So a 1,200-word article is about 5 minutes at 240 wpm, or 6 minutes at 200.
The range is honest, not lazy. Dense technical writing, code blocks, and tables drag the real number down; a familiar topic speeds it up. Reading aloud is much slower — closer to 130–150 wpm — so a script that reads in 4 minutes silently takes about 7 to narrate. Use the estimate to set expectations ("6 min read"), not as a stopwatch.
When a counter isn't the right tool
A word counter measures length. It says nothing about whether the writing is good, and chasing a number is a fast way to pad. If you're 40 words over a limit, cut — don't reach for a tool to confirm what you already know.
It also can't judge platform rules it doesn't model. If you're writing bilingual tweets, budgeting a serious SMS campaign, or working in a CMS that counts meta descriptions by pixel, verify against the real platform before you trust any generic count. And for readability — sentence length, grade level, passive voice — a plain counter is the wrong instrument; those need a dedicated readability checker.
So decide which limit you're writing to first, then count with the tool that mirrors it: 280 with URL-and-CJK weighting for X, characters-with-spaces for SMS, ~150 for a meta description. Ignore the other numbers. A count only means something once you know what's doing the counting.