How to Use This Tool
Paste your writing. You get the distribution of sentence lengths, not a single number that hides it.
Why the average is the wrong number
An essay of six twenty-word sentences and an essay running 7, 34, 9, 31, 12, 27 have exactly the same average. One is monotonous and one has rhythm, and no reader would confuse them. The average cannot tell them apart because averaging is precisely the operation that throws that information away.
The number to look at is spread — the standard deviation of the sentence lengths. Roughly:
- Under 4 — the sentences are nearly all the same length. This is the most common problem in exam essays, and it usually comes from writing every sentence to the same template.
- 4 to 12 — normal, varied prose.
- Over 15 — strong contrast. Deliberate in good writing, but check the long ones are still grammatical, because that is where run-on sentences live.
These bands are a description of what varied prose looks like, not a marking criterion. No exam board measures standard deviation. They are here because “vary your sentence length” is advice nobody can act on without knowing what their current spread is.
Runs matter more than the total
Spread is a whole-text number, so a text can have a healthy spread overall and still contain a paragraph where five sentences in a row are all 18–22 words. That run is what a reader notices. The tool flags any stretch of three or more consecutive sentences within a few words of each other.
About the sentence splitter
Splitting on every full stop is wrong, and it is what most word tools do. e.g., Dr., U.S., Fig. 3 and 3.5 all contain a full stop that does not end a sentence. A tool that reports fourteen sentences where there are eleven has corrupted the data before any analysis begins.
This splitter holds a list of common abbreviations, ignores full stops inside numbers, and requires a following capital letter. It is not perfect — an unusual abbreviation will still split — but the sentence list is shown in full so you can see exactly what it did rather than trusting the count.
