A readability score is a single number computed from the words, sentences, and syllables in a passage, and it is meant to predict how difficult that passage is to read. To check a readability score, you paste your English text into a tool that counts those inputs and then evaluates one or more published formulas against them. Most tools report only the final number, but the better ones also show the raw counts behind it, because readability scores are simple arithmetic functions of those counts. Two tools that return different scores for the same text almost always disagree about counting rather than about the formula. A good workflow is to run six published formulas at once, look at the raw counts together, and use the disagreements between formulas as a sanity check rather than a problem. This article walks through what the numbers mean, how to run them in one pass with the Readability Score Checker, and how to read the six scores together without being misled by any single one.
What a Readability Score Actually Measures
A readability score is a linear combination of length signals. Every published formula starts by counting words, sentences, and syllables (or characters), and then plugs those counts into an equation with coefficients that were fitted against reading-comprehension data in the mid-twentieth century. There is no natural-language model involved in the basic version; it is arithmetic on text statistics, with constants chosen decades ago and rarely revisited.
The most familiar output is Flesch Reading Ease, which returns a 0 to 100 style score where higher means easier. It is bounded in the sense that ordinary prose sits between 0 and 100, but the formula is genuinely open-ended: very short, very simple sentences can score above 100, and dense academic prose can score below zero. Because of that, the interpretation bands are applied as thresholds rather than as hard ceilings, and a score of 95 means the same thing whether the formula ever actually reaches 95 or not.
The other five common formulas — Flesch-Kincaid Grade Level, Gunning Fog, SMOG, Coleman-Liau, and the Automated Readability Index — return a United States school grade level. Grade-level outputs are easier to interpret at a glance, but they only make sense for running prose, not for lists, code blocks, tables, or heavily numeric text. A grade level of 8 on a poem or a recipe is not a meaningful number.
The Six Formulas and Why They Disagree
Readability formulas disagree by design, and the disagreements are informative. They disagree because they weigh different signals. The table below summarizes what each formula in the standard set actually measures.
| Formula | Output | What it weighs |
|---|---|---|
| Flesch Reading Ease | 0–100 style score (higher = easier) | Sentence length and syllables per word |
| Flesch-Kincaid Grade Level | US grade level | Sentence length and syllables per word |
| Gunning Fog | US grade level | Sentence length and three-syllable complex words |
| SMOG | US grade level | Polysyllabic words, calibrated on 30 sentences |
| Coleman-Liau | US grade level | Characters per 100 words and sentences |
| Automated Readability Index (ARI) | US grade level | Characters per word and sentences |
Flesch-Kincaid and SMOG weigh syllables. Gunning Fog weighs long words per sentence. Coleman-Liau and ARI count characters instead of syllables, which makes them useful as a cross-check when a syllable estimate looks suspect, because they share no counting logic with the syllable-based formulas.
Coleman-Liau and ARI can legitimately return a negative grade on very short, very simple sentences. That is the published arithmetic rather than an error, because both formulas subtract a small constant that can outweigh the input when the input is tiny. A score of negative one on a six-word sentence is the formula doing exactly what it was specified to do.
How to Score a Passage Step by Step
To check a readability score with the Readability Score Checker, the workflow is three actions and one reading step.
- Paste the English passage you want to score into the text box.
- Select Analyze to compute all six formulas at once.
- Read the Flesch Reading Ease band first, then compare the grade-level scores and check the raw counts underneath.
The raw counts underneath the scores are words, sentences, syllables, letters, characters, three-syllable complex words, and polysyllables. Those seven counts are the entire input to the six formulas. If a score looks wrong, the cause is almost always in one of those counts rather than in the formula, which is why the tool surfaces them rather than hiding them.
Sentence detection is heuristic, not exact. Decimals, initials, and a short abbreviation list are shielded so that a price or a title does not end a sentence early. If your text contains many acronyms or currency strings, glance at the sentence count underneath before trusting the grade levels, because an under-counted sentence count inflates every grade-level formula that has it in the denominator.
Syllable counting is also a rule-based estimate rather than an exact algorithm, because English has no authoritative spelling-to-syllable mapping. The counter splits hyphenated words, handles consonant-plus-le endings, decides whether a final "es" or "ed" adds a syllable, drops silent final "e", and counts the remaining vowel groups, with a floor of one syllable per word. It is tested against a hand-checked list and is reliable on ordinary prose, but it under-counts some words where vowels sit next to each other.
How to Read the Numbers Without Being Misled
Start with the Flesch Reading Ease band. The band labels follow the plain-language convention published by the MSKTC plain-language tool, not Flesch's own original wording, and they are applied as open-ended thresholds because the score is genuinely unbounded. A score that lands in the "fairly easy" band on a corporate website means the text is roughly at the level of a general-interest magazine, not that every reader will find it effortless.
Then look at the five grade-level scores. If they cluster within a grade or two of each other, the text has a stable reading level and any one of them is a reasonable summary. If they spread across four or five grades, the formulas are reacting to different features — one is picking up long words, another is reacting to character density — and that spread is the signal, not the average. The average hides the disagreement.
Two specific things to watch on the page:
- If SMOG diverges sharply from Flesch-Kincaid, the difference is almost always in polysyllable counting. Check the polysyllable count underneath, since both formulas are syllable-based and ought to track each other.
- If Coleman-Liau or ARI returns a negative grade, the passage is short and simple enough that the formulas' constants are dominating the input. Use the Flesch-Kincaid or Flesch Reading Ease number instead, and confirm with the raw sentence count.
Use the scores as a direction, not a verdict. They are good at telling you that a paragraph got denser after an edit, and poor at telling you whether a specific reader will understand it.
When the Score Stops Being Useful
Grade-level formulas were validated on running prose. On lists, code blocks, tables, or heavily numeric text, the results should be treated as indicative only, because the counts are computed but the formulas were not designed for that shape of text. A bullet list of twenty acronyms will produce a word count, a sentence count of one, and a grade level in the double digits, all of which are mechanically correct and editorially meaningless.
SMOG has an extra caveat. McLaughlin defined SMOG for a sample of exactly thirty sentences. Below that, the score is an extrapolation from a smaller sample, and because the formula takes a square root, the scaling also amplifies counting noise. The Readability Score Checker shows the SMOG result on shorter text but notes that it is extrapolated; treat the number as a direction, not a measurement, until you reach thirty sentences.
Lists of words, captions, dialogue fragments, and fragments of poetry all produce counts that the formulas will dutifully evaluate. The numbers you get back are valid arithmetic on those counts, but they were never calibrated against those kinds of input. If your text is not paragraph-shaped prose, treat every score as approximate and trust the counts underneath more than the scores themselves.
Privacy and What the Tool Does Not Do
Everything is computed in the browser, so the text is never uploaded, stored, or attached to an account. That matters for unpublished drafts, client material, and any text you do not want leaving your machine.
The tool does not score meaning, clarity, or readability in the editorial sense. It does not detect passive voice, filler words, or over-long sentences; it reports counts and the formulas built from them. For a defensible numeric score on plain prose, the six formulas here are the standard reference, and comparing them is the intended workflow.
The constants used in every formula were checked against a primary or authoritative source. The Flesch-Kincaid and ARI coefficients come from the 1975 Navy technical report that derived them; the SMOG constants come from McLaughlin's original 1969 paper; the Flesch Reading Ease coefficients and its interpretation bands come from a government plain-language document. Where published tables contradict each other, the disagreement was resolved with two independent sources rather than a majority vote: the Flesch coefficient is 84.6 and not the 83.6 that an OCR scan appears to show, and the Coleman-Liau coefficient is 0.0588 and not the 0.0855 that circulates in at least one journal table. Treating those constants as authoritative is what keeps the six scores comparable across tools.
If you're weighing options, How to Read Difficult Texts With a Line Diff covers this in detail.
If you're weighing options, How to Convert PPT to Text and Read It Slide by Slide covers this in detail.