Free Readability Checker
Paste your text or enter a URL to score it on six readability formulas at once, with the sentence and word counts behind them. You get all six rather than one, because they disagree with each other, and how much they disagree tells you something the individual scores do not.
⚠ Under 30 sentences — SMOG was designed on a 30-sentence sample, so treat that row cautiously.
A narrow gap means the text is consistently structured. A wide gap means the formulas disagree about how hard this is.
Longest sentences
Hardest words
Text mode runs entirely in your browser. Nothing you paste is sent anywhere.
Readability at a Glance
What a readability score actually measures
Every formula on this page measures two things: how long your sentences are, and how long your words are. That is all. Some count syllables, some count letters, one counts words of three syllables or more, but none of them reads anything.
They cannot tell whether a sentence makes sense, whether a term is jargon, whether the argument follows, or whether the reader already knows the subject. A paragraph of confident nonsense in short sentences scores well. A clear explanation that happens to use a four-syllable technical term scores worse.
That is not a reason to ignore the scores. Sentence length and word length genuinely do predict reading difficulty across large amounts of text. It is a reason to treat the number as a prompt to look, not as a verdict.
Word choice is also worth checking on its own terms — see how often you repeat the same ones with the keyword density checker.
Why six formulas, and why they disagree
They count different things. Flesch Reading Ease returns a score where higher is easier, and Flesch-Kincaid rescales the same two inputs into a US school grade. Gunning Fog counts words of three or more syllables. SMOG was built for health materials and is the most conservative of the six. Coleman-Liau and ARI count characters instead of syllables, because characters are easier to count mechanically.
Run identical text through all six and they land in different places. On plain writing they cluster: the sample in this tool spreads 7.3 to 10.7, a gap of a little over three school grades. On dense writing they scatter, 31.7 to 44.4, nearly thirteen grades apart.
The gap between the highest and lowest is worth as much attention as any single figure. A narrow spread means the text is consistently structured. A wide spread usually means long sentences and long words are combining in ways the formulas weight very differently, and that in itself is a sign the writing is hard going.
Method
Readability tools disagree partly because the formulas differ and partly because the implementations do. The specific decisions behind this one, published rather than buried in a tooltip:
Dale-Chall, Spache, Linsear Write and the Fry graph are left out. The first two score against fixed word lists that are dated, American-English specific and unclear on licence, and the other two add nothing the six here do not already cover.
There is no exact syllable algorithm for English, so two reasonable implementations differ by a few percent and that moves every grade score. Stating the method is what makes this tool checkable against another.
Flesch Reading Ease is not a percentage
The scale is conventionally described as 0 to 100, and it is neither bounded nor a percentage. Very simple text can score above 100. Dense text goes below zero: the dense sample in this tool returns minus 136.7, which is not an error. It means the sentence is 38 words long and averages over three and a half syllables a word.
If a tool shows you a Flesch score as a progress bar out of 100, it is presenting the number wrongly.
What score should you aim for
The one your readers can handle, which is not a number anybody else can give you.
The commonly quoted advice of 60 to 70 Flesch Reading Ease, or around US grade 8, is a reasonable default for general consumer writing. Government and health communications are frequently written to grade 6 deliberately. Technical documentation for engineers written to grade 6 would be padded and patronising.
The useful questions are who reads this, what do they already know, and are they reading by choice. A page that answers those honestly will land on its own target, and the tool will tell you whether you hit it. One genuinely universal point: nobody has ever complained that something was too easy to read.
Writing that machines quote accurately
Assistants and answer engines extract passages to quote. Short, self-contained sentences survive extraction. Long sentences carrying three clauses and a parenthetical do not: they get truncated, and a truncated sentence can say something you did not.
This is not a reason to write to a grade level for machines. It is a reason to notice that the habits which make writing readable — one idea per sentence, plain words where plain words exist, answers before elaboration — also make it quotable. Once the sentences hold up, the meta description generator helps carry that same clarity into the snippet Google shows.
How to improve a score, and how not to
Split your longest sentences first. It moves every formula at once and it is the largest effect for the least work. The tool lists them for you.
Replace long words only where a shorter one means the same thing. Use instead of utilise, help instead of facilitate. Do not swap a precise technical term for a vague short one to move a number.
Put the answer first. It shortens the sentence that carries it and it helps every reader.
Cut qualifiers. Very, really, quite, somewhat, arguably. They lengthen sentences and weaken claims at the same time.
Do not chop sentences into fragments to game the average. The score improves, the writing gets worse, and the reader notices even if the formula does not.
Do not write to a target you have not justified. If you cannot say why grade 8 is right for this page, the number is decoration. These same habits carry over to product copy — the product description generator applies them to listings specifically.
Common mistakes
Treating one formula as the answer. They disagree by design.
Reading Flesch Reading Ease as a percentage. It is unbounded and can go negative.
Applying a consumer target to specialist writing. Grade 6 on documentation for surgeons is not a goal, it is a mistake.
Comparing scores between tools. Syllable counting differs between implementations, which is why this page publishes its method.
Scoring a whole page including navigation and code. That measures your template, not your writing.
Optimising the score instead of the writing. The formula is a proxy, and improving a proxy without improving the text is a vanity metric.
Writing that people finish and machines quote correctly is not a formula output, but the formula is a fast way to find the sentences getting in the way. See how the Zaprev team approaches content and GEO.
Frequently asked questions
What is a good readability score?
It depends entirely on who is reading. Around grade 8, or 60 to 70 on Flesch Reading Ease, is a sensible default for a general audience. Specialist readers handle considerably denser text, and health and government material is often written deliberately simpler.
Why do readability tools give different scores for the same text?
Two reasons. They use different formulas, which weight sentence length and word length differently. And they count syllables using different rules, because English has no exact syllable algorithm. Both effects are small individually and they compound.
What is the difference between Flesch Reading Ease and Flesch-Kincaid?
They use the same two inputs and present them differently. Reading Ease gives a score where higher is easier. Flesch-Kincaid converts that into a US school grade, where higher is harder. The two always move in opposite directions on the same text.
Does readability affect SEO?
Not directly, since there is no ranking factor for reading level. It affects the things that do matter: whether people stay on the page, whether they understand the answer, and whether a passage can be lifted cleanly by something quoting you.
Can a Flesch score be negative?
Yes. The scale is not bounded and it is not a percentage. Very long sentences built from very long words produce large negative numbers, which simply means the text is far outside the range the formula was designed around.
Should I write to a grade level for AI search?
No, but the habits overlap. Assistants extract passages, and short self-contained sentences survive that better than long ones. Write clearly for people and the machine-facing benefit follows without aiming at a number.
Can I check a page that is not published yet?
Yes. Paste the draft into the text mode, which runs entirely in your browser and sends nothing anywhere. Use URL mode only for pages that are already live.
How much text do I need for an accurate score?
A hundred words is a reasonable floor and a few hundred is better. These formulas average sentence length and word length, so a very short sample lets one long sentence dominate. SMOG is the most sensitive of the six because it was designed around a thirty-sentence sample.
Does it work for languages other than English?
No. Every formula here was built and calibrated in English, and syllable counting rules that work in English do not transfer. Text in another language will still produce numbers, which is exactly why the tool stops instead of showing them.