Sentence Counter
Count sentences without the classic miscounts: Dr., e.g., $3.50 and J. R. R. do not create false breaks. See totals, averages and your longest sentence, all locally in your browser.
How counting works (spec v1)
- Words are locale-aware word segments: punctuation and emoji are not words, numbers like 3.50 count as one, contractions (don't) count as one, hyphenated compounds (follow-up) count as two, and a URL or email address counts as one.
- Characters are what you perceive: a family emoji, a flag or an accented letter counts as 1. Code points, UTF-16 units and bytes are in Detailed statistics.
- Sentences use locale-aware boundaries with abbreviation handling, so Dr., e.g. and $3.50 do not split.
- Paragraphs are blocks separated by a blank line. Line counts are in Detailed statistics.
- Density uses the same words as the headline count, phrases never cross a sentence boundary, and filters change what is listed, never the percentages.
Reading time uses 238 words per minute, speaking time 130 words per minute.
Detailed statistics
*English-language estimate. For a full report use the Readability Checker.
X's own rules: every link counts as 23, emoji and CJK count as 2.
Hard limit 2,200 characters; only about the first 125 show before "more".
Guidance, not a limit: Google truncates by pixel width per device. Check the SERP preview.
No fixed limit exists; snippets are truncated to fit and often rewritten.
Keyword density checker
Density is each phrase's share of ALL words in the text (the same word count as the headline). Filters and the English stopword list only change which phrases are listed, never the percentages. Phrases never bridge two sentences.
Type or paste text above to see your top words and phrases.
Very long text: phrases are computed from the first 200,000 words.
Why most sentence counters get it wrong
The lazy implementation splits text at every period, exclamation mark and question mark. That fails on ordinary prose constantly: every "Dr." and "e.g." spawns a phantom sentence, "$3.50" splits in the middle of a price, and an ellipsis mints three sentences from one. On a normal business paragraph the naive method can overcount by 30 to 50 percent, which then corrupts every metric built on it: average sentence length looks artificially short and readability scores come out flattering when the text is actually dense.
Sentence count matters mostly as a lever on readability. If your average runs past 20 words, split sentences rather than trimming words: two clear 12-word sentences beat one 25-word sentence in every readability formula and for every human reader. Check the effect live in Detailed statistics, then run the full report in the Readability Checker.
Sentence counter FAQ
How does this tool decide where a sentence ends?
It uses locale-aware sentence segmentation, the same boundary analysis built into modern browsers, plus a correction layer for the classic trip-ups: abbreviations such as Dr., Mr., e.g., i.e. and U.S., single initials as in J. R. R. Tolkien, decimal numbers like $3.50, and domain names. Naive counters split on every period and report "Dr. Smith arrived." as two sentences; this one counts it as one.
Do ellipses and multiple punctuation marks create extra sentences?
No. "Wait for it... done." counts as one sentence, and strings of punctuation with no words in them count as zero. A fragment only counts as a sentence if it contains at least one word or number.
What is a good average sentence length?
For clear web writing, 15 to 20 words on average with plenty of variation. Long stretches above 25 words per sentence push text into hard-to-read territory, which shows up directly in the Flesch scores under Detailed statistics. Averages hide extremes though, so also glance at the longest-sentence stat: one 60-word monster can survive a healthy average.
Does it work for languages other than English?
Yes for boundary detection: the segmentation is locale-aware and handles scripts without Latin punctuation conventions. The abbreviation correction list is English-focused, so counts in other languages are solid but can differ on abbreviation-heavy text. Readability formulas are English-only estimates and are labeled as such.
Is anything sent to a server?
No. Sentence detection runs entirely in your browser, works offline once loaded, and the draft autosave is a visible on-device toggle you control.
More free tools: Word Counter · Paragraph Counter · Readability Checker · Reading Time Calculator · Character Counter