Flesch-Kincaid, Gunning Fog and SMOG look like different tests. Open the formulas and they share one dominant input: how many words you put between full stops. Change your average sentence length and every readability score you have moves with it.
Why This One Number Carries So Much Weight
Readability formulas were built for a practical problem. The US Navy needed to know whether enlisted personnel could understand technical manuals, and insurers needed to know whether policyholders could understand contracts. Both needed a cheap, mechanical estimate rather than a comprehension study.
The formulas that survived all combine two proxies: sentence length, standing in for syntactic complexity, and word length or syllable count, standing in for vocabulary difficulty. Flesch-Kincaid weights average sentence length at 0.39 and average syllables per word at 11.8. Gunning Fog uses sentence length plus the percentage of words with three or more syllables. SMOG uses only polysyllabic words, counted across a fixed number of sentences.
None of them understand meaning. A sentence of nonsense scores the same as a sentence of insight if the two have the same shape. That is the honest limit of what these scores measure, and it is why they are useful as a warning system rather than a target.
Sentence Length Targets by Audience
Averages, not maximums. A document where every sentence hits the average reads worse than one that varies around it.
| Audience | Average words per sentence | Rough grade level |
|---|---|---|
| Children's material | 8-11 | Grade 3-5 |
| Popular news writing | 15-20 | Grade 7-9 |
| General web content | 14-18 | Grade 8-10 |
| Business and marketing | 15-20 | Grade 9-11 |
| Technical documentation | 15-22 | Grade 10-12 |
| Academic writing | 22-30 | Grade 13+ |
| Legal drafting | 25-45 | Grade 15+ |
Variation Matters More Than the Average
An average is a poor description of prose. Ten sentences of exactly 17 words average 17. So do sentences of 5, 28, 11, 34, 8, 22, 14, 31, 6 and 11. The first reads like a metronome. The second reads like someone talking.
Gary Provost demonstrated this in a passage that has been quoted for forty years, building from short sentences into a long one and back, so the rhythm of the paragraph enacts the point it is making. The lesson holds: readers register monotony as difficulty even when every individual sentence is easy.
So when the counter shows your average, treat it as one of two numbers you care about. The other is the spread. If your longest sentence and your shortest are within a few words of each other across a whole article, the writing will feel flat regardless of what the readability score says.
The practical move is to find your longest sentence and split it, then find two short adjacent sentences and join them. Both edits leave the average roughly where it was while making the paragraph noticeably better.
How this counter detects a sentence
A full stop, question mark or exclamation mark followed by whitespace or the end of the text. Requiring that whitespace is what stops "example.com" from splitting. Before that, common abbreviations, decimal numbers and ellipses are masked so they cannot register as breaks, which is why it handles the first three cases below that trip a naive split.
Where Sentence Counting Goes Wrong
The full stop is overloaded. It ends sentences, but it also does four other jobs. This counter handles the first three below; the rest defeat every tool.
Abbreviations (handled)
"Dr. Smith arrived at 9 a.m. on Tuesday." reads as one sentence to a human and three to a naive splitter. A list of common titles and Latin forms, including e.g. and i.e., is masked before splitting so they do not count.
Decimals and version numbers (handled)
"Version 2.1.4 shipped." contains two full stops that end nothing. Numbers with internal separators are masked first, so prices, percentages and measurements do not inflate the total.
Ellipses (handled)
Three dots may end a sentence or trail off mid-thought. Runs of three or more are collapsed rather than treated as terminators, which is the safer of the two wrong answers.
Quotations
A full stop inside closing quotes can belong to the quoted sentence, the containing sentence, or both, depending on whether you follow US or UK punctuation conventions.
Lists and headings
Bullet points often carry no terminal punctuation at all, so a document that is mostly lists reports far fewer sentences than it appears to contain, which drags the average length up.
Count sentences in your draft
Sentence totals alongside word, character and paragraph counts.
Using Readability Scores Without Being Ruled by Them
Writing to hit a score produces bad prose in a predictable way. Formulas reward short sentences and short words, so optimising against them pushes you toward choppy text full of monosyllables, and comprehension actually falls.
The useful application is diagnostic. Run the numbers on a finished draft, and if the average sentence length is far above your target, look for the cause rather than mechanically chopping sentences. Usually it is one of three things: nested clauses that should be separate sentences, lists written as prose that should be bullets, or unnecessary qualification.
It also helps to check the score of something you know works. Run a page of writing you admire in the same genre. If your draft sits five grade levels above it, that gap is meaningful. If it sits within a level or two, the score is telling you nothing you need to act on.
Frequently Asked Questions
Related Tools
Keep Reading
Word Count Targets: What Editors, Platforms and Google Expect
Two counters can disagree by five percent on the same paragraph. Here is how counting actually works, and the targets worth writing to.
Reading Time Estimates: Where 238 WPM Comes From and When It Lies
The number behind every "5 min read" label, why technical content reads at half that rate, and how to time a script for narration.
Why Character Counts Disagree: Emoji, Accents and SMS Segments
One emoji can cost seven characters, and a single curly apostrophe can triple your SMS bill. The three definitions of a character, and when each applies.
Title and Meta Description Length: Pixels, Not Characters
Google truncates by rendered width, so a title of capitals runs out thirty characters early. Plus why it rewrites most descriptions anyway.