TL;DR: "Character count" isn't as literal as it sounds — an emoji can be one visible character but multiple code points, and what counts as a "word" changes across languages that don't use spaces. Picking the wrong counter gives you a number that doesn't match what a platform actually enforces. Pair this guide with Word Frequency, Remove Extra Spaces, AI Hidden Characters and Text Difference.
Character Counter and word counters solve different problems, and platforms enforce different limits: a tweet is measured in characters, an essay word count in words, and neither number predicts the other reliably.
Background: the Unicode text segmentation report (UAX #29) defines how text splits into "grapheme clusters" (what a person perceives as one character), and MDN's Intl.Segmenter reference covers the browser API that implements it correctly.
Why an emoji breaks naive character counting
A family emoji (👨👩👧👦) looks like one character but is actually four separate person emoji joined by invisible zero-width joiner (ZWJ) characters — a naive counter using JavaScript's .length reports 11 or more for something a person sees as one glyph, because .length counts UTF-16 code units, not visible characters.
A grapheme-aware counter, built on Intl.Segmenter, counts what a person would count: that same emoji as 1. This is the difference between a character count that matches a platform's real limit and one that silently over- or under-reports on any text with emoji, accented letters, or combined symbols.
Why word counting depends on the language
Word counters that split on whitespace work fine for English, Spanish or French — but Chinese, Japanese and Thai don't use spaces between words at all, so a whitespace-based counter returns a single "word" for an entire paragraph. Getting a meaningful word count in those languages requires actual dictionary-based or statistical segmentation, not a simpler split.
Even within space-separated languages, edge cases pile up: is "don't" one word or two? Is a hyphenated compound one word or two? Different tools make different calls, which is why the same paragraph can report different word counts depending on which counter you trust.
Matching the counter to what actually gets enforced
- Social media post limits (X/Twitter, SMS) → character count, and grapheme-aware if the text has emoji.
- Meta descriptions, title tags → character count, since search engines truncate by character/pixel width, not by word.
- Essays, articles, word-count minimums for content briefs → word count.
- CJK content (Chinese, Japanese, Thai) → character count is usually the meaningful metric, since "words" aren't well-defined without space breaks.
A quick sanity check
If a platform's limit is enforced in characters, paste the exact final text into Character Counter before publishing — not an estimate, the literal string, including any emoji or special punctuation, since those are exactly where naive counting tools diverge from what actually gets enforced. If you're editing down to a word target instead, run it through a word counter, then double check the character count separately if the platform (like a meta description) enforces both.
For text with hidden formatting issues that also skew counts — extra spaces, invisible Unicode marks — clean it first with Remove Extra Spaces or AI Hidden Characters, then count the cleaned version.