I have edited a lot of drafts against a lot of scorers. Hemingway open in one tab, the Flesch Reading Ease number ticking over in Word, Yoast throwing its traffic-light bullets in WordPress. For a long time I chased the obvious fixes first: swap the jargon, cut the passive voice, trim the adverbs. Those help. None of them moved the score the way one quieter change did.
The change was sentence length, and not just cutting it, but varying it. Once I started watching the shape of my sentences instead of only their content, the grade level dropped, the red highlights cleared, and the writing read better out loud. This guide is what I learned doing that over and over with the numbers in front of me. It covers why length is the biggest lever in every readability formula, where those formulas quietly mislead you, and the exact edits that raise the score without turning your prose into a pile of fragments.
Why length is the biggest lever in every score
Every readability formula worth knowing shares one input: the average number of words per sentence. Syllable counts and complex-word ratios matter too, but sentence length is the variable you control most directly, and the one that shifts the score fastest. Here is how the main formulas use it.
| Formula | How it uses sentence length | A shorter average | Web target |
|---|---|---|---|
| Flesch Reading Ease | Subtracts 1.015 points per average word | Raises the 0 to 100 score | 60 to 70 |
| Flesch-Kincaid Grade | Adds 0.39 grade levels per average word | Lowers the grade | Grade 7 to 8 |
| Gunning Fog | Sentence length is about half the formula | Lowers the fog index | Under 12 |
| SMOG | Counts complex words inside each sentence | Fewer long sentences lowers it | Under 12 |
| Automated Readability | Uses words per sentence directly | Lowers the grade | Grade 7 to 8 |
| Coleman-Liau | Uses sentences per 100 words | More sentences lowers the grade | Grade 7 to 9 |
The Flesch math in plain terms: drop your average sentence length by five words and Reading Ease climbs about five points, while the grade level falls close to two whole grades. That is the fastest legitimate way to move either number, and you have not touched a single word of vocabulary yet. |
The catch most guides skip: average, not variation
Here is the part almost no one mentions. Readability formulas score the average, not the spread. A passage of identical 20-word sentences and a passage that swings from 6 words to 34 can share the same average, and therefore the same Flesch score. The formula cannot hear the difference between them. A reader can.
| The formula measures the input. It does not measure the music. |
So if the arithmetic only cares about the average, why vary length at all? Three reasons, and each one helps either your score or your ranking.
The numbers that actually matter
How short is short enough, and how long is too long? The most cited answer comes from readability research popularized by the writing coach Ann Wylie, drawn from American Press Institute analysis of hundreds of newspapers.
~100% comprehension at 8 words or fewer per sentence | 90%+ still understood at an average of 14 words | <10% comprehension once sentences reach 43 words |
Treat the exact figures as directional rather than laboratory-precise. The shape holds across every study of the question: comprehension stays high into the low twenties, then drops sharply.
| Sentence length | How it reads |
|---|---|
| 8 words or fewer | Very easy |
| 11 words | Easy |
| 14 words | Fairly easy |
| 17 words | Standard |
| 21 words | Fairly difficult |
| 25 words | Difficult |
| 29 words or more | Very difficult |
Martin Cutts, author of the Oxford Guide to Plain English, gives the working rule most editors settle on: keep the average across the whole document between 15 and 20 words. Rudolf Flesch landed in the same range decades before him. The goal is an average in that band with a wide range of lengths underneath it, not every sentence pinned to the same count. The right band shifts a little with the job, though. Here is where I aim.
| Content type | Target average | Practical spread |
|---|---|---|
| General blog or web | 14 to 18 | mix from 5 to 25 |
| Landing page or ad copy | 8 to 14 | short and punchy, few long ones |
| 10 to 15 | one idea per sentence | |
| Product or help docs | 15 to 20 | split anything past 30 |
| Technical or academic | 18 to 22 | longer allowed, still break run-ons |
The 40-year-old paragraph that proves the point
The clearest demonstration of why variation matters is almost forty years old. In 100 Ways to Improve Your Writing (1985), Gary Provost wrote a paragraph that performs its own lesson. It opens flat and deliberate: “This sentence has five words.” He repeats the pattern until it drones, then breaks it, stretching into one long, building line that rises like a run of music before it lands.
Read it once and you feel the point in your ear, which is exactly where readability lives. The score is only a proxy for that feeling. Provost was making the case for rhythm long before a plugin could count your syllables, and nothing since has said it better.
Seven edits that vary length on purpose
This is where the score actually changes. None of these takes long, and the first two carry most of the weight.
| 1. Fix your longest first | Scan for any sentence that fills two lines in your editor. That is where comprehension breaks and where a red highlight usually sits. Splitting those gives the biggest jump for the least effort. |
| 2. Split at the joint | Most long sentences are two sentences wearing a conjunction. Find the and, but, so, or because that joins two complete thoughts and swap it for a period. You gain clarity and lose nothing. |
| 3. Land a short one after two long ones | Once you have written two flowing sentences, drop a blunt one of three to six words. It creates emphasis, resets the reader's breath, and pulls the average down at the same time. |
| 4. Combine when it is choppy | Variation cuts both ways. When four short sentences march in a row, the writing sounds like a list. Merge two of them with a semicolon or a subordinate clause to bring the rhythm back. |
| 5. One idea per long sentence | A long sentence is fine when it carries a single idea. It fails when it carries three. Before you keep a long one, check that every clause serves the same point, then break off the stragglers. |
| 6. Vary the opening too | Yoast flags three or more sentences in a row that begin with the same word, and readers feel the repetition even when they cannot name it. Changing your openings compounds the effect of changing lengths. |
| 7. Read it aloud | Your lungs are a readability meter. Wherever you run out of air before the period, the sentence is too long. Put the period where the breath wants to fall. |
A before and after you can measure
Here is the same idea written two ways. Nothing about the meaning changes. Only the length does.
| BEFORE · ONE SENTENCE | AFTER · SIX SENTENCES |
|---|---|
| When you are trying to make a piece of online content easier to read, it helps to remember that the length of your sentences affects how easily readers follow your point, and for that reason you should keep an eye on how many words land in each sentence, so that you never overload the reader with too much at once, which is a mistake writers make far more often than they realize. | Sentence length shapes readability more than almost anything else. Watch the word count in each one. When a sentence runs long, the reader has to hold too much at once, and the point slips. That mistake is more common than writers think. Keep most sentences short. Save the long ones for when they earn it. |
| What moved. The before is a single 72-word sentence. Hemingway marks the whole thing red, and a Flesch calculator drops it far below college level. The after says the same thing in six sentences averaging about nine words, the longest 18 and the shortest 4. Every red highlight is gone, and Reading Ease climbs by around seventy points on the 100-point scale. The vocabulary barely changed. The length did all the work. |
Where writers go wrong with this
A handful of ways the technique backfires, most of which I have done myself.
• Making everything short. Chop every sentence down to eight words and the writing turns choppy and cold. It reads like an instruction manual. Variation is the goal, not minimalism.
• Chasing the bullet. A green Yoast light or a low Hemingway grade is a proxy, never the prize. If a rewrite scores higher and reads worse, keep the original and take the worse number.
• Short sentences, long words. Length is only half the formula. Trim a sentence to twelve words but leave four heavy, multi-syllable nouns inside it, and the syllable count keeps your score up. Plainer words finish the job.
• Fragments as a habit. A fragment lands once, for emphasis. Deploy it every third line and it stops working, because the reader stops noticing it.
• One target for everything. Blog numbers do not fit a legal clause or a scientific method section, where a longer sentence carries relationships the reader needs. Match the target to the audience.
When keeping a sentence long is the right call
Short is not automatically better, and the tools agree. Hemingway's own guidance is to save longer sentences for emphasis and rhythm rather than aiming for conciseness at any cost. A long sentence earns its place when it carries one unbroken chain of logic, a flowing description, or a legal or technical relationship that would snap if you cut it in half.
The test is never the word count on its own. It is whether the reader can hold the whole thing in mind at once. When they can, length is a tool. When they cannot, it is a wall.
The workflow I run on every draft
1. Draft without editing. Get the ideas down and ignore length completely on the first pass. Editing while drafting kills momentum.
2. Measure a baseline. Paste into Hemingway for the highlights and grade, and note one Flesch Reading Ease number from Word or a calculator. Now you know where you stand.
3. Clear the reds. Fix the very hard sentences first, then the yellows that genuinely clunk. Leave the ones that already read fine.
4. Hunt for monotony. Look for four or more sentences of similar length in a row, and break the pattern with a short one.
5. Vary the openings. Change any run of sentences that start with the same word. Small move, real effect.
6. Re-measure, then stop. When the average sits in your target band and the reds are gone, stop. Do not sand the life out of it chasing a rounder number.
The verdict, after a lot of drafts
After years of doing this with the scorers open, my honest take is that sentence-length variation is the single highest-return edit in the readability toolkit. It is the rare change that pleases the formula and the human at the same time, because the formula rewards the lower average while the person rewards the rhythm. Fix the jargon and the passive voice too. Those matter. But if you only have time for one pass, break your longest sentences, drop a short one after every couple of long ones, and read the whole thing out loud.
The number will climb on its own. Just keep in mind what the number is for. It stands in for a reader who wants to reach the end without working for it, and that reader was always the real audience, not the score.
