Word Counter & Text Statistics

What counts as a word

There is no single universal definition, which is why two counters can disagree by a few percent on the same document. This tool uses a Unicode-aware rule: a word is one or more letters or digits, optionally continued by apostrophes and hyphens. That means:

• don’t is one word, not two.
• state-of-the-art is one word, matching Microsoft Word.
• 2026 counts as a word, as it does in Google Docs.
• Accented and non-Latin scripts count correctly, because the pattern uses Unicode property escapes rather than [a-z].
• Emoji and standalone punctuation are not words.

Character counts include every code point, spaces included, which is the figure social networks and CMS field limits use. The “no spaces” tile removes all whitespace, which is what most typography and translation quotes are based on.

Character limits worth knowing

Choosing a limit from the dropdown turns on a live meter. These are the ones that matter most in practice:

How reading and speaking time are estimated

Reading time is words divided by a words-per-minute rate. The default of 238 wpm comes from Marc Brysbaert’s 2019 meta-analysis of 190 reading studies, which found a mean of 238 wpm for silent reading of non-fiction English prose. It is a far better default than the 200 or 250 usually quoted without a source.

Speaking time uses 130 wpm, near the middle of the comfortable presentation range of 120–150 wpm. Audiobook narration sits around 150–160, conversational speech nearer 180.

Reading the word frequency table

The frequency table ignores stop words — the grammatical filler such as the, of and is that dominates any raw count without telling you anything about the subject. What remains is a rough topical fingerprint of the text.

Use it as a diagnostic, not a target. If a page about espresso machines has “machine” at 4% and the actual product category at 0.2%, the emphasis is wrong. But there is no optimal density figure to hit: Google’s ranking systems have used contextual language models since BERT rolled out in 2019, and repeating a phrase to reach a percentage is at best neutral and at worst spam.

For n-gram analysis, readability scoring and a full report, use the dedicated keyword density and readability analyzer.

Useful length targets for writers

• Blog post: 1,000–2,000 words is the usual sweet spot for informational queries; length should follow the question, not a quota.
• Product description: 150–300 words.
• Email subject line: under 50 characters to survive mobile truncation.
• Paragraph: 40–80 words on desktop, shorter on mobile.
• Sentence: average 15–20 words; anything over 25 sustained hurts comprehension.
• Academic abstract: 150–250 words for most journals.

Frequently Asked Questions