Readability score from six formulas at once

It depends entirely on who has to read it, and the sensible targets are lower than most writers expect. General web copy and journalism aim at around grade 8 to 9, which is where a mainstream newspaper sits. Health information written for patients is usually held to grade 6 or below on the advice of medical bodies, because comprehension collapses when a leaflet outruns the person holding it. Academic and legal writing routinely runs at grade 14 and up, and that is defensible for its audience — a corporate policy at grade 16 aimed at all staff is not.

Because they were built at different times, for different readers, from different raw material. Flesch calibrated against reading tests from the 1920s and 1930s; Gunning worked with newspapers in 1952; McLaughlin built SMOG in 1969 against complete comprehension rather than the 50 to 75 per cent his predecessors accepted, which is why it usually reports the highest grade of the six. Coleman-Liau and the Automated Readability Index count letters instead of syllables and so treat short technical jargon as easy. A spread of two to three grades between them is completely normal; a spread of four or more usually means the text contains something the formulas cannot parse.

At least 100 words, and preferably 300 to 500. Flesch built and validated his formula on 100-word samples, and the convention in readability research is to take several such samples from different parts of a document and average them rather than scoring the whole thing in one pass. SMOG is the strictest: McLaughlin specified 30 sentences, taken as 10 from the beginning, 10 from the middle and 10 from the end, and its score is unreliable below that.

They distort it, sometimes badly. A heading with no full stop is treated as one sentence, which is fine; a bulleted list where the items have no terminators is read as a single enormous sentence, which is not, and it can push a whole document up several grades on its own. Tables and code produce runs of tokens the formulas were never designed to see. If a document is mostly structured content, score the prose sections separately — and treat a wide disagreement between the six as a hint that something like this is going on.

Flesch-Kincaid if you need one number and no reason to prefer another — it is the most widely recognised, is built into word processors and is written into US federal and state plain-language requirements. SMOG if the material is medical or safety-critical, because it targets full comprehension and health-literacy guidance is written around it. Gunning Fog for business and technical writing, which is what Robert Gunning built it for. Coleman-Liau or ARI when the text is full of proper nouns or technical terms whose syllable counts are unreliable.

Yes, trivially, and that is the honest limitation of every formula here. Replace half your commas with full stops and the grade drops a year or two while the prose gets choppier and no clearer. None of these formulas reads your text: they count sentence lengths and word lengths and infer difficulty from the correlation. They cannot see word order, unfamiliar concepts, missing context, ambiguity or a badly organised argument — all of which matter more to a real reader than the arithmetic does.

Not directly — no search engine publishes a grade-level threshold or scores pages on one. It matters indirectly, and substantially: a page a visitor can follow gets read, quoted and linked to, and a passage written in short declarative sentences is more likely to be lifted as an answer than one buried in a 60-word subordinate clause. Write for the grade your audience reads at because it makes the page work, not because a crawler is checking.