Readability Score

Three readability formulas computed on your text at once, with sentence-level statistics and a list of the sentences that run past thirty words. Nothing here depends on a language model — the scores are arithmetic on syllables, words and sentences, which means they are reproducible and you can check them.

Three Formulas Because One Is Not Enough

Readability formulas disagree with each other on the same text, and the disagreement is informative rather than a defect. Each was designed for a different purpose and each fails in a different direction, so reading three together tells you more than reading one carefully.

• Flesch Reading Ease returns a number where higher is easier, calibrated so that seventies and eighties sit in the comfortable range for general readers and anything under thirty is genuinely hard. It responds to sentence length and to syllables per word, which means it reacts to vocabulary as much as to sentence construction.
• Flesch-Kincaid Grade returns a school year. It is the more familiar number to most people, and it is derived from the same two inputs as Reading Ease in the opposite direction, so it will always agree with it — the pair is one measurement expressed two ways rather than two independent opinions.
• Gunning Fog uses a different measure of difficulty: the share of words with three or more syllables, counting only those that are not proper nouns, familiar jargon or made of simpler words combined. Because it keys on long words rather than on syllables per word across the board, it punishes a technical vocabulary that Reading Ease only mildly notices.
• Sentence statistics accompany the scores — counts, average length, and the distribution — which is what lets you connect a bad score to a cause.
• A dedicated list flags every sentence over thirty words. This is the most actionable output on the page, because a long sentence has a specific fix and a low readability score does not.
• A sample-text button loads a passage so you can see how the output looks before pasting your own work into it.

Turning a Score Into an Edit

1. Paste your text and read the long-sentence list before the scores. If five sentences are over thirty words, the scores are a symptom and the list is the diagnosis.
2. Check whether the three formulas agree. When Gunning Fog is much worse than the two Flesch-based numbers, the problem is long words rather than long sentences, and the fix is vocabulary. When all three are bad, the sentences are the cause.
3. Cut the flagged long sentences first, and cut rather than split where you can. Splitting a forty-word sentence into two twenty-word sentences improves every score while changing nothing about the writing; deleting a clause that was restating the previous one improves the writing as well.
4. Re-run after each editing pass. The number moving is the confirmation that you edited the thing the formula measures, which is the only way to avoid optimising the score instead of the text.
5. Ignore the passive-voice flags as a list to follow narrowly. The detection is a pattern match on auxiliary verbs followed by participle-looking endings, so it marks genuine passives and also constructions where the word after the verb merely ends in those letters.
6. Use a consistent passage for comparison when tracking a document over time — scores are only comparable between texts of similar length and purpose.

Reading Ease, Grade Level and Fog Side by Side

Alternative	Notes
Flesch Reading Ease	Best read as a difficulty dial with no target. It is useful for tracking a single document through edits and for comparing two drafts of the same piece. It is not useful as an absolute judgement, because a legal document should score badly and a children's book should score in the nineties.
Flesch-Kincaid Grade	The same computation expressed as a school year, which makes it easy to state a target — most business writing lands most comfortably between grade eight and eleven. Because it shares its inputs with Reading Ease, consulting both adds no information; pick whichever framing your audience understands.
Gunning Fog	The one to consult when the criticism is that a piece sounds academic. It counts three-syllable-plus words and is therefore sensitive to nominalisation — organisation, implementation, utilisation — which is the single most common habit that makes corporate prose hard to read. A Fog score far above the Flesch-Kincaid grade is a specific, fixable finding.
Why they disagree at all	They were built for different readers and different eras. Reading Ease and its grade-level companion were derived for adult literacy assessment and later repurposed as style guides; Fog was built to measure how much difficulty a newspaper reader would face with a military manual. None of the three was designed to grade prose quality, and a high score is not a compliment.

A Paragraph Before and After

Take a typical two-sentence corporate paragraph of about sixty words, with a nominalised subject and two subordinate clauses.

• Before editing, the sentence-length statistics show an average well above twenty words and the long-sentence list holds both sentences. All three formulas return poor figures, with Fog worst of the three because of words like implementation and optimisation.
• The first edit is splitting the first sentence where the second clause begins. This reduces average sentence length and moves one entry off the long list, and all three scores improve by a similar amount — the sign that sentence length was the shared cause.
• The second edit replaces a nominalisation with the verb it was hiding — using decide instead of decision-making, for instance. The Flesch numbers barely move, because the syllable count of the replacement is close, but the Fog score improves noticeably, which is exactly the divergence the three-formula approach is for.
• The passive-voice list before editing flags both auxiliary constructions. After editing it flags none, which is a coincidence rather than a target — the rewrite removed the passives because restructuring the sentences made the actors explicit, not because the flags were chased.
• Note that the word count rose slightly across these edits. Splitting a sentence adds a subject and a verb, and a readability edit that shortens the text is usually deleting content rather than improving clarity.

How the Numbers Are Computed

Syllable counting is the foundation of all three formulas and it is a heuristic rather than a dictionary lookup. Words of three letters or fewer count as one syllable; trailing word endings that silence the final vowel are stripped before counting; a leading y is removed; and the remaining vowel groups are counted. That gets ordinary English mostly right and is occasionally wrong on uncommon words and proper nouns, which is why the exact figure can differ by a point or two from a tool that uses a pronunciation dictionary. The formulas themselves are the published ones, applied to those counts.

• Complex-word detection requires three or more syllables and additionally excludes words with a pattern of consonant, vowel group, consonant and a silent final e — which stops words ending in a silent e from inflating the Fog score.
• Passive detection looks for a form of the verb to be followed by a word with a participle-like ending. It catches genuine passives and also flags constructions where the following word merely ends that way, so treat the count as an upper bound.
• No provider is required and none is used: this page has no AI call in it at all. The analysis runs on the counts alone.
• Nothing is transmitted, and the page works with the network disconnected. The sample text button loads a passage compiled into the page rather than fetching one.

Questions About the Scores

Which of the three numbers should I actually aim at?

Whichever one matches the thing you are being criticised for. If the feedback is that sentences are too long, watch the sentence statistics and the long-sentence list and let the formulas follow. If the feedback is that the writing sounds academic, watch Gunning Fog, because it is the one sensitive to long words rather than sentence length. Aiming at a number for its own sake produces text that scores well and reads strangely, since the formulas count syllables and cannot tell whether a sentence means anything.

Why are my passive-voice flags wrong?

Because detection is a pattern: a form of the verb to be followed by a word ending in a participle-like suffix. That matches genuine passives such as "the report was finished" and also matches constructions where the following word happens to end the same way, such as a noun or an adjective. Use the flags as an upper bound and read each one rather than treating the count as a score to reduce.

The same text gives slightly different numbers on another tool. Why?

Almost always the syllable count. This tool uses a rule-based counter — short words count as one, certain endings are stripped, initial y is dropped, vowel groups are then counted — and a tool built on a pronunciation dictionary will disagree on uncommon words, abbreviations and proper nouns. The formulas are standard; the input to them is an estimate. Differences of a point or two are expected and do not indicate that either tool is wrong.

Is my document sent anywhere to be scored?

No. Syllable counting, sentence splitting and the three formulas are all arithmetic performed in the page, and this tool contains no provider integration — there is no AI panel to configure and no request to make. You can paste unpublished work into it and verify in the network panel that nothing left the browser.

Last verified: October 3, 2026

Launch Readability Score

All Tools · All Tools · About · How We Test · Contact · Privacy

Maintained by Huang Bin · Last updated October 3, 2026