Sentence Counter

The counter starts from your browser's built-in sentence segmenter (Intl.Segmenter), which follows the Unicode sentence-boundary rules in UAX #29. Those rules already handle a lot: a full stop inside example.com, U.S. or $3.50 doesn't end a sentence, and neither does a full stop followed by a lowercase word.

They have no list of abbreviations, though. On its own, the segmenter ends a sentence after every Dr., Mrs. and Jan. followed by a capital or a number. Browsers ship ICU's abbreviation data but don't switch it on. The counter adds a second pass that joins those pieces back together, and does the same after a question or exclamation inside quotes when the next word is lowercase (“Is it?” she asked.).

The abbreviation list is deliberately short. It includes titles (Mr., Mrs., Ms., Dr., Prof., St., Sr., Sra.), reference words (Fig., No., Vol., p., pp., e.g., i.e., cf., vs.) and month abbreviations, plus common German, French and Spanish ones such as z. B., vgl. and Ud. It leaves out etc., Inc. and Ltd. because they often do end a sentence, as in the Acme row. Any single capital letter followed by a full stop is read as an initial, so Plan B. Then go. counts as one sentence. If that matters for your text, turn off "Abbreviations and initials".

Under the Unicode rules a line break always ends a sentence, whatever comes before it. That suits text you typed yourself, where a new line means a new heading, list item or paragraph. It doesn't suit text copied from a PDF or an email, where lines are wrapped at a fixed width and one sentence can run over three lines. In the table, ⏎ marks a line break.

"Join wrapped lines" still treats a blank line as a break, so paragraphs stay separate, and it switches the paragraph count to blocks separated by blank lines. "Leave out headings and list items" also skips the last sentence while you're still typing it, until you add the full stop. Skipped lines stay visible in the sentence list, greyed out.

Press Load example to put this paragraph in the counter:

The browser's segmenter alone finds 7 sentences, because it breaks after Dr., after Jan. and before her neighbours asked. The counter finds 4, with 57 words and an average of 14.3 words per sentence. That average looks fine, but the list shows one sentence of 36 words carrying most of the paragraph. Splitting it gives:

Average sentence length hides the problem. Check the longest sentence and the number of flagged sentences as well as the average.

There is no correct length, but there are well-used yardsticks. Rudolf Flesch graded average sentence length in The Art of Readable Writing (1949), and GOV.UK's content team quoted the same levels when it set 25 words as the point to check a sentence:

GOV.UK's rule is a prompt, not a ban: check any sentence over 25 words, split it if you can, and if you can't, make sure it's in plain language. The flag in the counter defaults to 25 for that reason. Pick 20 for web copy or 30 to 35 for academic writing. Mixing lengths matters too. A run of sentences that are all the same length reads flatly even when each one is short.

Microsoft Word shows a sentence count only in its Readability Statistics, after a spelling and grammar check, and like every tool it has its own rules for abbreviations, headings and list items. Expect small differences on text with many abbreviations or bullet points, and check this counter's sentence list to see exactly where it split.

Word counts here follow the word-processor convention of counting runs of characters between spaces. A hyphenated word, an email address or a URL is one word. Tokens with no letters or digits, such as a lone dash, aren't counted. The browser's word segmenter would split many of these, which is why the counter doesn't use it for spaced languages:

Characters are counted as you see them. An emoji such as 👨‍👩‍👧‍👦 is one character even though it is 11 UTF-16 code units, so the count can be lower than a program that uses JavaScript's length.

The Unicode rules cover most scripts: Chinese and Japanese full stops (。), Arabic question marks (؟) and Spanish opening marks (¿ ¡) are all handled. For Chinese, Japanese and Thai, which are written without spaces between words, the word count comes from the browser's dictionary-based word segmenter instead. Everything runs in your browser; your text isn't uploaded.

In an older browser the counter falls back to splitting after ., ! or ? followed by a space, and shows a notice under the tool. The abbreviation pass still runs, but the fallback splits after every full stop followed by a space, including before a lowercase word.