Text Readability Score Checker

• Text analyzer
• Word counter
• Case converter
• Text merger
• Text splitter
• Word frequency
• Keyword extractor
• Word to text
• PDF to text

A readability score answers a question every writer has and no spell-checker touches: is this draft written for the people who have to read it? Tune marketing copy to a general audience, pitch teaching material at a year group, or check a policy page against a plain-language requirement — without sending a word of it to a server.

Benefits

• Ten formulas: the five classics plus Coleman-Liau, Dale-Chall, Spache, Linsear Write and FORCAST, side by side.
• One number to act on: a consensus grade taken as the median of the seven grade-scale indices.
• Where, not just whether: each sentence scored on its own and tinted when it reads several grades above the rest.
• Style flags: passive constructions and adverbs marked inside the sentences that carry them.
• Your actual file: .txt, .md, .csv and .docx, not just what you have already copied.
• Free: every formula and every highlight.

The tool counts what the formulas need — sentences, words, syllables, letters, characters, familiar words — and applies each published definition exactly. The counting is where readability tools differ from one another, so it is worth knowing what this one does.

What is counted

• Sentences: split on . ! ? and their runs, with abbreviations, single-letter initials, decimals and ellipses recognised, and line breaks treated as boundaries so a heading is one sentence rather than part of the paragraph below it.
• Words: letters and digits, with apostrophes and hyphens joining a word rather than breaking it. Both are switchable.
• Syllables: a rule set with a 1,167-word exception list; all-capital initialisms are counted as spelled out, so ADHD is four syllables.
• Characters: letters and digits only. ARI and Coleman-Liau are defined that way, and counting punctuation as characters reads both indices high on quote-heavy copy.
• Complex words: three or more syllables, excluding proper nouns, hyphenated compounds and words made three-syllable only by an -es, -ed or -ing ending — which is Gunning's own definition, and the part most implementations drop.
• Nothing is clamped: Flesch Reading Ease is reported as computed, above 100 or below zero.

How to interpret readability scores

• Watch the trend across revisions rather than any single number in isolation. A draft moving from grade 14 to grade 11 is real; the difference between 11.2 and 11.4 is not.
• When the indices disagree with one another, the spread shown beside the consensus grade tells you by how much — a wide spread usually means the document mixes very short and very long sentences.
• Coleman-Liau uses no syllables at all. When it agrees with Flesch-Kincaid, the syllable counting is not what is driving the answer.
• Pair the numbers with a human read for tone, jargon and whether the text is actually correct. No formula has an opinion about meaning.

Every formula here was designed for a particular kind of text and a particular question. They are not ten opinions about the same thing, which is why the tool shows all of them rather than averaging them into one and hiding the disagreement.

What each index measures

• Flesch Reading Ease: sentence length and syllables per word, on a scale where higher is easier. The most widely quoted readability number there is.
• Flesch-Kincaid Grade Level: the same two inputs, expressed as a US school grade. Written into US government and military plain-language standards.
• Gunning Fog: sentence length plus the share of complex words. Sensitive to jargon in a way the Flesch measures are not.
• SMOG: polysyllable count against sentence count. Defined on a 30-sentence sample and widely used for health materials.
• Automated Readability Index: characters per word instead of syllables, so it never depends on a syllable estimate.
• Coleman-Liau: letters per hundred words and sentences per hundred words. Designed for machine scoring and, like ARI, syllable-free.
• New Dale-Chall: the share of words outside a 3,000-word list familiar to fourth-graders. Vocabulary rather than word length.
• Revised Spache: the same idea for early-grade text, against a smaller list and counting each unfamiliar word once.
• Linsear Write: easy words score one and hard words score three. Developed by the US Air Force for technical manuals.
• FORCAST: single-syllable words per 150, and the only index here that ignores sentence length — which makes it the one that still works on forms, headings and bullet lists.

Why the consensus grade is a median

Seven of the ten produce a US grade, and they disagree — sometimes by several grades on the same paragraph. The headline number is their median rather than their mean, because a mean lets one outlier drag the answer: SMOG on a four-sentence document, or FORCAST on a page of headings, can land grades away from everything else. A median ignores it. The spread between the highest and lowest index is shown next to the grade, so a document the indices cannot agree on says so.

Three of the ten formulas and the complex-word ratio behind a fourth are functions of one number: how many syllables are in each word. English spelling being what it is, that number is hard, and most online readability tools use the same four-line heuristic — count vowel groups, drop a silent e, call anything under four letters one syllable.

What the difference is worth

• The common heuristic: 94.0% of tokens correct, measured against the CMU Pronouncing Dictionary over two full-length books.
• This tool: 99.6% of tokens correct on the same two books, which the exception list was not built from.
• In Flesch points: on a 300 KB sample of dense non-fiction the heuristic is 3.47 points out; this is 0.11 out. Three and a half points is the width of half a band.
• Where it goes wrong: the heuristic under-counts hiatus — abbreviated is five syllables, not three; association is five, not four — and over-counts silent endings, calling acuteness four.

Sentence boundaries

The other place readability tools quietly disagree is where a sentence ends. Splitting on every period is simple and produces confident, badly wrong answers on ordinary text.

• "Dr. Smith met Mrs. Jones at 4 p.m. on Jan. 3 to discuss the U.S. economy." is one sentence. Split naively it is eight, and the average sentence length falls from 18 words to 2.25.
• "The item costs $3.50." is one sentence, not two — a period between digits is a decimal point.
• A heading or a bullet list has no terminator at all. Treated strictly it produces zero sentences, and a tool that divides by zero sentences reports a headline as college-graduate reading level.
• Line breaks end sentences by default, which is what makes headings, bullets and poetry measurable. The behaviour is a switch in the rail if your text uses hard-wrapped paragraphs.

A score summarises a document; it cannot point at the sentence that needs rewriting. Every sentence here gets its own reading grade, calculated from its own length and its own words, and is tinted according to how far above the document it sits.

The bands

• Amber: four or more grades above the document's level. Usually one clause too many.
• Red: six or more grades above. Almost always worth splitting in two.
• Fixed targets: switch the comparison to grade 8, 10, 12 or 14 when you are writing to a stated requirement rather than to your own register.
• Relative by default: a legal brief where every sentence is grade 16 is not usefully described as "every sentence is a problem".

Style flags

Two patterns account for most sentences that scan as heavy without being long, and both are marked inside the sentence that carries them.

• Passive voice: a form of "to be" followed by a past participle, including irregular ones — "was written", "has been adopted", "is taken". Passive is not an error; a page where most sentences are passive is usually harder than it needs to be.
• Adverbs: -ly words, with the ones that only look like adverbs excluded, so "family", "supply" and "July" are not counted.
• Both markers can be switched off if you only want the reading-grade tint.

Any time the question is "who can read this?" rather than "is this spelled correctly?" — blog posts, help articles, internal comms, classroom handouts, consent forms and policy pages written to a stated level.

Ideal use cases

• Web copy: simpler pages reach more people, and the highlighted sentences are the ones costing you readers.
• Education: check a handout against the year group it is written for, with Spache for early-grade material and Dale-Chall above it.
• Health and public information: SMOG is the index this field uses, and the tool tells you when your sample is long enough for it to mean anything.
• Editing: measure a draft before and after a pass and see which change actually moved the number.
• Comparison: queue two versions of the same message with Pro and read them in one table.
• Localization QA: a translated page that scores four grades harder than its source usually has a sentence-length problem, not a vocabulary one.

These indices were developed for English prose between the 1940s and the 1970s, mostly on textbooks and technical manuals. That history shows up in what they handle well and badly.

Key points

• Every index here is a function of surface features — lengths and counts. None of them reads the text, so none of them can tell whether it is coherent, accurate or well organised.
• Code, tables and reference lists mixed into prose distort word and syllable statistics badly. Measure the prose on its own.
• SMOG is defined on a 30-sentence sample; below that the square root in the formula turns a couple of long words into a whole grade. The tool marks the score as provisional rather than hiding it.
• Spache is designed for text at or below fourth grade and stops discriminating above it, which the tool also marks.
• Grade level is not comprehension. A grade-6 sentence about an unfamiliar subject is still hard.
• Different tools produce different numbers for the same text, and the difference is almost always tokenization rather than arithmetic.

Treat the scores as signals about structure, not as verdicts about quality.

Quality tips

• Measure plain text. Strip navigation, captions and code blocks first — they are not what your reader is reading.
• Compare like with like: the same document, the same settings, before and after an edit.
• Fix the red sentences before worrying about the overall number. They usually move it on their own.
• Very short samples produce noisy metrics on every index. A few hundred words is a reasonable floor.
• For medical, legal or regulatory copy, follow whatever methodology your organisation specifies — including which formula it names.

When not to rely on it

• Non-English or mixed-script documents. The formulas assume English orthography, and the syllable rules assume English spelling.
• Regulatory filings that specify a different methodology, sample size or word list.
• Poetry, dialogue-only scripts, song lyrics and anything where sentence length is a stylistic choice rather than a difficulty signal.
• As a certification. No score here is evidence of compliance with an accessibility or plain-language standard.

Every measurement this tool makes is free. Pro pays for scale, not for numbers — a readability score that only paying users can see would not be worth coming back for.

• Free: all ten formulas, the consensus grade, every statistic, sentence highlighting, passive and adverb marking, and report export as .txt or .csv, on documents up to 10 MB.
• Pro: a queue of up to 20 documents with a comparison table across them, all reports in one archive, no ads, and no imposed size limit beyond what your device can handle.
• Never: an upload. Everything runs on your device on both plans, including with the network switched off once the page has loaded.

See what Pro includes

English-oriented formulas, a rule-based syllable estimate, and a style detector that works on surface patterns rather than grammar. It needs JavaScript, and it works in any current browser on desktop, tablet and phone. Documents are read on your device; nothing is uploaded.

Recommended editing workflow

• Measure the draft first and note the consensus grade and the spread, so you know what you are working against.
• Rewrite the red sentences, then the amber ones. Splitting one long sentence usually moves several indices at once.
• Look at the passive count next: it rarely changes the grade much, but it changes how the page reads.
• Re-measure with the same settings, and check the number moved for the reason you expected.
• Confirm the final text with the analyzer and word counter before publishing.

For a full text-quality pass, combine readability with the text analyzer, the word counter and the word-frequency tool. If your draft is in a Word file or a PDF, convert it first with Word to text or PDF to text.

A megabyte of prose — about 180,000 words — is measured in under a tenth of a second on your device, off the main thread, so the page stays responsive while it works.