Text Readability Analyzer

Readability Formulas: A Sober Guide

Readability formulas measure surface complexity of text, average word length, sentence length, syllable counts, and map those measurements to a comprehension score. They've been used in education, journalism, government communication, and insurance/medical disclosure for almost a century. They're useful, popular, and limited. Use them with appropriate skepticism.

The Big Five

Flesch Reading Ease (FRE)

Scale: 0-100 (higher = easier). Mapping:

Most plain-English writing targets 60-70. This paragraph tested with FRE: around 50, somewhat dense.

Flesch-Kincaid Grade Level (FKGL)

Output: US grade level (e.g., 8.5 = mid-8th grade). The most cited score in modern tools because the interpretation is intuitive.

SMOG Index

Polysyllables = words with 3+ syllables. Developed for healthcare comprehension. Tends to be slightly more conservative (suggests higher grade) than Flesch-Kincaid. Often used for medical pamphlets, consent forms, regulatory disclosure.

Automated Readability Index (ARI)

Output: US grade level. Character-based (no syllable counting). Quick to compute. Was originally designed for typewriter output to estimate grade level from key counts.

Coleman-Liau Index

Output: US grade level. Like ARI, character-based. Tends to give similar but slightly different scores from ARI; the two are usually within 1-2 grades of each other.

Why Five Formulas?

Each weighs inputs differently. If all five agree the text is grade 8, you can be confident. If they disagree (e.g., FKGL says 12, Coleman-Liau says 8), look at why, usually it's syllable-heavy short sentences (penalized by FKGL/SMOG) or short-syllable long sentences (penalized by ARI/Coleman-Liau).

The other inputs your editor sees:

• Words: total word count.
• Characters: total character count (sometimes with/without spaces).
• Sentences: counted by ending punctuation (., !, ?).
• Paragraphs: blank-line separated.
• Average words per sentence: a powerful single number. >20 is dense; <15 is light.
• Average syllables per word: ~1.5 = simple, ~2.0 = dense.
• Reading time: words / 250 (average reading speed in WPM).
• Speaking time: words / 150 (average speaking speed).

Syllable Counting (The Hard Part)

Counting syllables in English is famously hard. The standard heuristic:

1. Count vowel groups (sequences of vowels treated as one group).
2. Subtract one for each silent e at word end.
3. Subtract one if the word ends in le and preceded by a consonant (able).
4. Add one for words ending in -cious, -tion, etc. that the rules undercount.
5. Minimum: 1 syllable per word.

This gets ~95% of words right. Edge cases (compound words, names, technical terms) need a dictionary lookup for accuracy. Most online tools use the heuristic; this tool does too. Small inaccuracies generally don't change the grade level meaningfully.

Audience-Appropriate Reading Levels

The Plain Language Movement (now law in some US agencies) targets grade 8 for public communication. Many government and insurance materials fail this, readability tools highlight the gap.

Improving Readability

The big levers, in order of impact:

1. Shorten sentences. A 35-word sentence with three clauses isn't graceful; it's hard. Split it. Two 17-word sentences are easier than one 34-word.
2. Cut adverbs. very, really, basically, actually are filler. Removing them rarely changes meaning and lowers word count.
3. Prefer short words. use not utilize, help not facilitate, now not at this present time. Long words add syllables and feel pretentious.
4. Active voice. The committee approved the proposal not The proposal was approved by the committee. Active is shorter and clearer.
5. One idea per sentence. Compound-complex sentences hide structure. List bullets or sequential sentences make logic visible.
6. Avoid jargon, or define it. Each domain term is a comprehension barrier. Use them sparingly; define on first use; consider whether a layman term works.
7. Vary sentence length. All short sentences feel choppy; all long ones feel exhausting. Aim for a mix.

Hemingway Editor and Grammarly are good companion tools, they highlight specific sentences to fix, complementing the aggregate score from readability formulas.

The Formulas' Blind Spots

Vocabulary familiarity isn't captured. Quantum chromodynamics (5 syllables, 19 characters) scores hard. Banana (3 syllables, 6 characters) scores easier. But many readers find banana instantly comprehensible and quantum chromodynamics opaque despite similar-ish surface complexity. The formulas can't tell.

Structure isn't captured. Subject-verb-object active sentences are easier than equivalent passive ones at the same word count. The formulas treat them identically.

Coherence isn't captured. A page of grade-6 sentences in random order is incoherent; the formula scores it as readable.

Visual layout isn't captured. Bullet lists, headings, images, whitespace all aid comprehension. Plain paragraphs of the same text score identically.

Genre conventions aren't captured. A grade-15 academic abstract isn't a failure if the audience is academics; the formula doesn't know.

Domain familiarity isn't captured. Mitochondria produce ATP scores hard but is trivial to a biology student.

Don't optimize for the score blindly. Use it as a signal that prompts review: "this paragraph scored grade 14, is that intentional, or did I write convoluted sentences?"

How Tools Use Readability Scores

Microsoft Word and Google Docs include readability stats (Word in Editor settings, Docs via add-ons). Used to flag overlong sentences.

Yoast SEO / Rank Math check WordPress posts for FKGL/Flesch and flag dense paragraphs, SEO theory holds that easier-to-read posts rank better.

Hemingway Editor is dedicated to readability: highlights long sentences (yellow/red), passive voice (green), adverbs (blue), complex phrases (purple). Suggests rewrites.

Government plain-language tools (e.g., US PlainLanguage.gov resources) require FKGL ≤ 8 for public communication.

Healthcare and pharmaceutical compliance uses readability formulas (SMOG most common) to demonstrate patient-information materials are accessible.

Privacy

Analysis runs entirely client-side: text is split, counted, and scored in JavaScript. No part of your input is transmitted. Open DevTools Network during use: zero outbound requests. Important because the text being analyzed might be a draft of confidential material, internal memos, unreleased product descriptions, legal correspondence, executive emails, none of which should be uploaded to a third-party service for analysis. The math runs in your tab.