Sentence Counter
Paste a draft below to see how many sentences it contains, along with its word and character totals. Divide the words by the sentences and you have average sentence length — the single most actionable number in readability. Everything is computed locally in your browser.
Average Sentence Length as a Working Number
A sentence count on its own is inert. Paired with a word count it becomes average sentence length, and that figure predicts how hard your prose is to read better than almost anything else you can measure cheaply.
The guidance converges on a narrow band. The US federal plain-language guidelines ask for an average of about 20 words or fewer. The GOV.UK style guide sets a hard ceiling of 25 words on any individual sentence. The Plain English Campaign recommends an average of 15 to 20. Across those sources the working answer for a general audience is 15 to 20 words on average.
That band is not a law of nature. Academic and legal writing addressed to specialist readers routinely averages 25 to 30 words and is not thereby bad; the readers have the domain knowledge to hold longer structures in working memory. Instructions, safety notices and interface copy should sit well below the band — closer to 10. The band is the default you deviate from deliberately, not a target to hit.
Where the number earns its keep is as a diagnostic on your own draft. Run a paragraph you suspect is heavy and check the average. If it comes back at 34, you have found the problem without needing to reread anything.
Where Sentence Length Enters the Readability Formulas
Every widely used readability formula takes sentence length as a direct input. Flesch Reading Ease is the most familiar:
206.835 − (1.015 × words per sentence) − (84.6 × syllables per word)
Two variables. One is your vocabulary, which is slow and disruptive to change. The other is your sentence length, which you can change with a full stop.
Worked through: take a 300-word passage written as 15 sentences. Average sentence length is 20, contributing a penalty of 20.3 points. Split it into 20 sentences instead — same words, same vocabulary, five extra full stops — and the average falls to 15, a penalty of 15.2. The score rises by about 5 points for no editorial work beyond punctuation. Run the same change through Flesch-Kincaid Grade Level:
(0.39 × words per sentence) + (11.8 × syllables per word) − 15.59
and the reading grade drops by nearly two years. Gunning Fog, which adds average sentence length to the percentage of three-syllable-plus words and multiplies by 0.4, responds just as sharply. The Automated Readability Index carries sentence length at a coefficient of 0.5.
Scores map to bands roughly as follows:
| Flesch Reading Ease | Reads as | Rough level |
|---|---|---|
| 90–100 | Very easy | 5th grade |
| 80–90 | Easy | 6th grade |
| 70–80 | Fairly easy | 7th grade |
| 60–70 | Plain English | 8th–9th grade |
| 50–60 | Fairly difficult | 10th–12th grade |
| 30–50 | Difficult | College |
| 0–30 | Very difficult | Graduate |
These are not purely advisory. A number of US states set a minimum Flesch Reading Ease score — commonly in the 40s — that consumer insurance policies must clear before they can be issued in the state. Sentence length is the lever underwriters reach for first, because rewriting the vocabulary of a policy means renegotiating its meaning.
Variance Beats the Average
The mean hides the thing that actually makes prose readable. A draft averaging 17 words where every sentence is 17 words long is monotonous in a way readers feel before they can name — it produces a flat, metronomic rhythm that stops carrying emphasis.
Gary Provost made the point better than anyone in 100 Ways to Improve Your Writing, with a demonstration passage that begins in identical five-word sentences and then breaks into varied lengths so the reader hears the difference in the act of reading it. The mechanism is simple: a short sentence lands hard because the sentences around it are long. Length is a contrast effect. Spend it all evenly and you have nothing left to spend.
Practically, that means reading the count alongside the shape of the text rather than instead of it. Two useful moves:
- After a long, qualified, multi-clause sentence, write a short one. Four words. It functions as punctuation at the paragraph level.
- Watch for three consecutive sentences of similar length. That is usually where a reader's attention starts sliding, and it is fixed by merging two of them or splitting one.
Semicolons deserve a mention here because they distort the count in a revealing way. A semicolon joins independent clauses into a single sentence, so semicolon-heavy prose returns a low sentence count and a high average length. If your average comes back at 30 and the text does not feel heavy as you read it, count the semicolons before you start cutting — you may simply be looking at sentences that want to be separate.
The Punctuation That Fools Sentence Counters
Splitting on periods, question marks and exclamation marks gets you most of the way and then fails in specific, predictable places. Here is exactly what this tool does and does not resolve.
Handled correctly:
- Ellipses. Any run of two or more periods is collapsed before splitting, so "Wait... what?" counts as one sentence, not three.
- Decimal numbers. Decimals are masked before splitting. "The rate rose to 4.5. That surprised nobody." counts as two sentences, where a naive counter reports three.
- Stacked terminators. "Really?!" and "Stop!!!" each count once, because runs of terminal punctuation are treated as a single boundary.
- A missing final period. Text that simply stops without punctuation still counts as a sentence.
Not handled — adjust for these yourself:
- Abbreviations. "Dr.", "e.g.", "i.e.", "etc.", "Inc.", "Ph.D." and "U.S." each add a spurious boundary. A passage dense with them counts high by roughly one per abbreviation. This is the standard failure mode of period-splitting counters; solving it properly requires an abbreviation dictionary, and dictionaries are language-specific and always incomplete.
- URLs and domains. "Visit usefulspot.com today." splits at the dot in the domain and counts as two.
- Three-part version numbers. Plain decimals are protected, but "Python 3.11.4 shipped." splits after the second dot.
- Quoted dialogue. "'Stop!' she said." counts as two, because the exclamation mark inside the quotation is a real terminator as far as the splitter can tell.
None of this is hidden, and knowing the seams makes the number usable: if you are counting a technical document full of abbreviations and version strings, read the result as an upper bound.
Reading the Count Back Into a Draft
Once you have the number, a few interpretations are reliable enough to act on directly.
A very high sentence count relative to word count — an average under 10 — reads as choppy or, in longer pieces, as a bulleted list that lost its bullets. The fix is usually subordination: join the sentences that stand in a causal or contrastive relationship with "because", "although", "which".
A low count with a high average is the more common failure, and the fix is nearly always to find the conjunction and cut it. Most 40-word sentences are two sentences with an "and" welded in the middle.
A stable average across a long document suggests you have settled into one register. Section openings and closings usually benefit from being shorter than the material between them.
For length against essay and article targets, the word counter has the thresholds. If you are writing to a hard platform limit rather than a readability goal, the character counter covers SMS segments, post limits and search-result truncation.