Readability checker
Scores your text for reading ease, US grade level and Gunning Fog, and says plainly what those 1940s–70s formulas do and don't measure — they count sentence and syllable length, not whether the argument holds.
Runs entirely in your browser — nothing is uploaded, and there is no account or limit.
Last reviewed 15 August 2026 · The HumanFlow team
Verdict
Paste some text above to see how hard it is to read, and at roughly what school grade.
These are heuristics, not measurements. Flesch (1948), Flesch–Kincaid (1975) and Gunning Fog (1952) all count sentence length and syllables — none of them reads your argument, your evidence or your structure. Two passages can score identically and be worlds apart in quality. Use the band, ignore the decimal.
Readability is not an AI signal on its own — detectors react to how much your sentence lengths vary, which is a different thing. See your rhythm →
What each score actually measures
All three formulas on this page work from the same two ingredients: how long your sentences are, and how long your words are. Neither of those is a measure of quality, which is why a page of fluent nonsense and a page of careful argument can score identically. Knowing what each one counts tells you when to listen to it.
Flesch Reading Ease (1948)
Rudolf Flesch’s original formula, scored 0–100, where higher is easier. It weighs syllables per word about three times as heavily as words per sentence, so it punishes long words harder than long sentences. Written for the US Navy and newspaper copy, it remains the most widely quoted readability figure and the one built into most word processors.
Flesch–Kincaid Grade Level (1975)
A rescaling of the same two inputs onto US school grades, commissioned by the Navy to check whether technical manuals could be read by enlisted personnel. A score of 9.0 means the text should be readable by an average ninth-grader. It is the same measurement as Reading Ease wearing different units, so the two never really disagree — if one moves, so does the other.
Gunning Fog Index (1952)
Robert Gunning’s index, also expressed as school years, but it swaps syllable counting for a blunter rule: any word of three or more syllables is “complex”. That makes it quick to calculate by hand, which is why it caught on in business writing, and it means technical vocabulary is penalised whether or not it was the right word.
What none of them can see
This is the part most readability tools leave out, so it is worth stating plainly. These formulas were designed in the 1940s, 50s and 70s, before anything resembling a modern corpus was available, and they count only surface features. They cannot see whether your argument follows, whether your evidence supports your claim, whether your paragraphs are ordered sensibly, or whether a sentence means anything at all. Text that is deliberately scrambled word-by-word scores the same as the original.
They also have a specific failure mode worth knowing if English is not your first language: the formulas reward short words and short sentences, which are exactly the features of careful, simple prose written by someone still building vocabulary. A perfectly good essay can score as “easy” for reasons that have nothing to do with its quality — and the same plain, evenly-paced writing is what AI detectors misread as machine-generated. Neither number is telling you what it appears to be telling you.
Use the band, not the decimal. If the score says “very difficult” and you did not intend density, find your three longest sentences and split them — that single edit moves every formula on this page more than any amount of word-swapping.
The two formulas quoted most often are worth understanding rather than trusting: Flesch Reading Ease runs 0–100 with higher meaning easier, while Flesch–Kincaid Grade Level outputs a school year and runs the other way, so quoting one as the other reverses the meaning. Both are worked through step by step on those pages, and readability itself covers what the whole family of formulas can and cannot see. If you want the score to move honestly, the largest single lever is removing nominalizations — the syllable term in both formulas is weighted heavily enough that turning a buried noun back into a verb shifts the number more than cutting a sentence in half.
Common questions
- What is a good Flesch Reading Ease score?
- It depends entirely on who is reading. 60–70 is the range most general non-fiction sits in and is a sensible target for a blog post or a report. Academic writing routinely lands between 30 and 50 and is not wrong for doing so — a paper on econometrics cannot be written at the reading level of a news article without losing the argument. The score is only a problem when it is low for reasons that have nothing to do with the subject.
- What grade level should my essay be?
- For undergraduate work, grade 12–15 is normal. The number is not a target to hit, though. Writing deliberately down to a lower grade produces short, choppy sentences that read as simplistic, and writing up to a higher one usually means padding clauses onto sentences that were fine. Write the idea clearly and let the grade land where it lands.
- Does a readability score affect AI detection?
- Not directly, and this is worth being precise about because the two get conflated constantly. AI detectors respond to how predictable your word choices are and how much your sentence lengths vary — not to how long your words are. A piece can be highly readable and read as machine-written, or dense and read as human. If you want to see the pattern detectors actually react to, the sentence rhythm analyzer shows sentence-length variation directly.
- Why do different readability checkers give different scores?
- Because every one of them has to guess at syllables, and they guess differently. There is no way to count syllables exactly without a pronunciation dictionary, so each tool uses its own heuristic and they disagree on words like "poet", "queue" and "business". Sentence splitting varies too — abbreviations and decimal points fool some tools into counting extra sentences, which changes every formula. Expect a spread of a grade level or so between any two checkers.
- Is a lower Gunning Fog score always better?
- No. Gunning Fog counts words of three or more syllables as "complex", which means accurate technical vocabulary is penalised exactly like padding is. "Photosynthesis" and "utilisation" cost the same three-syllable penalty, but one is the correct word and the other is "use" wearing a coat. Cut the second, keep the first, and ignore what that does to the number.
- Do these formulas work on any language?
- No. Flesch, Flesch–Kincaid and Gunning Fog were all built and validated on English, and their syllable-per-word constants encode English word structure. Running German or Finnish text through them produces a number, but not a meaningful one. Adapted variants exist for several languages and use different coefficients.
Need more than a utility?
HumanFlow rewrites AI drafts to read naturally and scores them for AI, with the signals behind the number — free up to 1,500 words per scan.
Try the humanizer free