Readability scores, explained

A short guide to what the Flesch Reading Ease and Flesch–Kincaid Grade Level scores actually measure, what numbers to aim for in different kinds of writing, and where automated scoring stops being useful.

What "readability" means in this context

Readability scores are estimates of how hard a piece of writing is to read, based purely on sentence length and word length. They were designed in the 1940s for the US Navy, originally to grade training manuals; they've since become the standard way of describing the linguistic difficulty of any piece of English prose.

What they measure is real but narrow. Short sentences with short words score as easy; long sentences with long words score as hard. They cannot measure whether the meaning is clear, whether the argument is good, or whether you've explained your acronyms. They are about surface complexity, not comprehensibility. A perfectly readable score can sit on top of badly organised, factually wrong text — and a low score can describe writing that is dense but still crystal clear.

Flesch Reading Ease

Devised by Rudolf Flesch in 1948. Higher is easier. The score is a single number, almost always between 0 and 100, calculated from average sentence length and average syllables per word.

For reference, this paragraph you're reading scores around 60–65 — "plain English". Most newspapers and consumer non-fiction sit in the same band.

Flesch–Kincaid Grade Level

Same input, different output. Developed in 1975 by Peter Kincaid for the US Navy, this one gives you the US school grade a reader needs to handle the text. A grade of 8.4 means "roughly an eighth-grader could read this".

The two scores correlate strongly but not perfectly: a grade level around 8 corresponds to a reading-ease score around 65; grade 12 ↔ score 50; grade 16 ↔ score 30.

What number should you aim for?

It depends entirely on who you're writing for. A common mistake is "aim higher" — but for most communication, lower is usually better. Aim for the lowest grade level that respects your subject's vocabulary.

The classic Plain English guidance — championed by Flesch himself, then by people like Ernest Gowers in the UK government — is to aim for a Flesch score of 60 or above for anything you want a general adult audience to read without effort. That's a reasonable default if you have no clearer guidance from your context.

How to nudge a score up

If your text is scoring lower than you'd like, both Flesch numbers respond to exactly two levers:

• Sentence length. Break long sentences. A 30-word sentence with one main clause and three subordinates can almost always become two 15-word sentences with no loss of meaning. Average sentence length is the bigger of the two factors in both formulas.
• Word length. Replace polysyllabic Latinate words with shorter Anglo-Saxon equivalents where you can. "Use" instead of "utilise"; "help" instead of "facilitate"; "show" instead of "demonstrate". The trick is to do this without losing precision — sometimes the long word is the right one.

Things that don't affect the score but matter for actual comprehension: defining your acronyms, putting the key point first, using examples for abstract claims, and structuring with headings and lists. A formula that ignores these will happily approve dense, unstructured paragraphs of short words.

Other scales you might see

• Gunning Fog Index. Like Flesch–Kincaid but weights complex words (3+ syllables) more heavily. Tends to score business and academic writing slightly higher (harder).
• SMOG (Simple Measure of Gobbledygook). Counts polysyllabic words across a sample of 30 sentences. Popular in healthcare communication and US legal compliance work because it correlates closely with comprehension in lower-literacy populations.
• Coleman–Liau. Uses character count instead of syllable count — easier to compute automatically because counting syllables in English is genuinely hard.
• Automated Readability Index (ARI). Also character-based. Designed for early computers that couldn't count syllables.
• Dale–Chall. Compares against a fixed list of ~3,000 words considered "familiar" to fourth-grade readers; any word outside the list counts as difficult. Very different output to the Flesch family — emphasises vocabulary over sentence length.

Where readability scoring fails

Both Flesch scores were calibrated on mid-20th-century American English prose. They generalise poorly to:

• Code, equations and structured data. A snippet of JSON or Python will score absurdly easy (very short "sentences", short "words") or absurdly hard, depending on tokenisation. The score is meaningless for non-prose content.
• Languages other than English. The syllable-counting and average word-length numbers are tuned to English. German, with its long compounds, scores artificially hard; Spanish, with its long average sentence length, also scores hard despite often being clearer to a fluent speaker.
• Specialist vocabulary that's correctly used. A medical paper that says "myocardial infarction" instead of "heart attack" will score worse, but for a cardiologist audience the precise term is the correct choice. Use the score as a check, not a target.
• Conversational writing. Dialogue and informal writing often score easier than they read. Lots of short sentences full of "you know" and "like" pull the number down.

How count. computes the score

The count. readability block uses both Flesch scores, calculated on the text exactly as you paste it. Sentence boundaries are detected on terminal punctuation (.?!), and syllables are estimated by a vowel-cluster heuristic that handles most regular English words. For very short inputs (under about 30 words) the score is too noisy to be useful; we hide it below that threshold.

Common questions

Is a higher Flesch Reading Ease always better?

No — it's only better if your audience would benefit from easier prose. For specialist or technical work, aiming for 90+ usually means dumbing down past the point of accuracy. The right answer is "as simple as the subject allows".

Why does my text score harder than a news article that feels harder?

Probably because of average sentence length. Newspaper prose uses lots of short paragraph-length sentences; an essay or technical post with one or two long sentences per paragraph scores higher even if the individual words are easier. Break the long sentences and the score will catch up.

Does Google use readability for ranking?

Not as a direct signal as of 2026. Google has said many times that they don't have a "readability score" input into their ranking. Indirect effects are real — easier-to-read pages tend to keep users engaged longer, and dwell time is part of the search signal — but you can't manipulate this by gaming the Flesch number alone.

Are these scores in the public domain?

Yes. The Flesch and Flesch–Kincaid formulas have been in the US public domain since they were developed under US government contracts. Any tool can compute them; the inputs are the text itself.

Why is the score different between tools?

Because syllable counting in English is approximate. The word "fire" — one syllable or two? "Every" — two or three? Different tools choose different heuristics, and the choice can shift a Flesch score by 5–10 points on the same text. The directional reading (easier vs. harder) is robust; the absolute number is approximate.