What readability scores actually measure

Every readability formula in common use measures the same two things: how long your sentences are, and how complicated your words are. That is the whole idea. They differ only in how they define “complicated” and how they weight the two numbers against each other.

This matters more than it sounds, because it tells you exactly what a readability score can and cannot do for your essay.

The two inputs

Sentence length is words divided by sentences. Easy to count, and genuinely predictive — long sentences hold more in working memory before they resolve.

Word complexity is measured one of two ways, and this is the only real difference between the formulas:

• By syllables. Flesch, Flesch-Kincaid and SMOG count syllables per word, or the proportion of words with three or more syllables.
• By letters. Coleman-Liau and ARI count characters per word instead. This was a deliberate choice: when these were designed in the 1970s, letters could be counted mechanically and syllables could not.

Neither is more correct. They are two proxies for the same underlying thing.

The six formulas

Flesch Reading Ease

The only one that runs 0–100 with higher meaning easier. Everything else outputs a US school grade level.

The score bands conventionally read as: 90–100 very easy, 60–70 plain English, 30–50 difficult, 0–30 very difficult. Academic writing in the humanities typically lands in the 30s and 40s. That is not automatically a problem — see the last section.

Flesch-Kincaid Grade Level

The same two inputs, rescaled to a US grade.

A result of 12 means “a US twelfth-grader could read this”, roughly a UK Year 13 student or a first-year undergraduate.

Gunning Fog

Uses complex words — three or more syllables — rather than average syllables. This makes it more sensitive to a few long words than Flesch is.

One consequence worth knowing: inflected endings should not count. “Managing” is not a genuinely harder word than “manage”, so a careful implementation strips common suffixes before deciding a word is complex. Not all of them do, which is one reason two tools can give you different Fog scores for identical text.

SMOG

SMOG is the one with a genuine precondition: it was derived on samples of 30 sentences or more, and below that the square root makes it unstable. A tool that reports SMOG on a single paragraph is reporting a number, not a measurement. Ours computes it but treats anything under 30 sentences as indicative only, and you should too.

Coleman-Liau

No syllables at all. Because letter-counting is unambiguous and syllable-counting is a heuristic, Coleman-Liau is the most reproducible of the six — different tools will agree on it far more closely than they agree on Fog.

Automated Readability Index

Also letters-based, and the most aggressive of the six about long words — that 4.71 coefficient is doing a lot of work.

Why they disagree, and by how much

Run the same paragraph through all six and you will routinely see a spread of three or four grade levels. This is normal and it is not a bug in any of them.

The reason is structural. Syllable-based formulas punish Latinate academic vocabulary — “consequently”, “methodology”, “interpretation”. Letter-based formulas punish long words regardless of how they sound. A paragraph full of short-but-technical words (“gene”, “allele”, “codon”) scores as easy on every one of them, which is obviously wrong.

Because of that spread, an average across several formulas is more stable than any single one. That is what our readability checker reports as its headline grade, alongside all six so you can see the disagreement rather than have it hidden from you.

The blind spot

Here is the thing every readability tool should tell you and most do not.

These formulas cannot read. They count. Take a well-written paragraph, shuffle the sentences into a random order, and the score does not move at all — same words, same sentences, same syllables. Replace every noun with an unrelated noun and, again, the score barely moves.

So a good score does not mean your writing is clear. It means your writing is short and plain, which is usually necessary for clarity and never sufficient.

It also means the formulas are trivially gameable, and gaming them makes writing worse. You can improve any score by chopping sentences in half at arbitrary points. The number goes up; the prose gets choppier and the argument gets harder to follow.

So what should you actually do with the score?

Use it as a flag, not a grade.

The number itself is close to meaningless in isolation — no marker has ever given marks for a Flesch score. What is genuinely useful is the distribution: which specific sentences are dragging the average up. A 60-word sentence in the middle of an otherwise tight paragraph is nearly always a sentence that is doing two jobs and should be two sentences.

That is why it is worth looking at the individual sentences a checker flags rather than the headline number. Fixing three overlong sentences will improve both the score and the essay. Chasing the score directly will improve only the score.

A reasonable target for an undergraduate essay is a grade level somewhere in the 12–16 range. Below 10 usually means you are underselling a complex argument; above 18 usually means at least some of your sentences have got away from you. But treat that as a sanity check, not a target — a philosophy essay legitimately reads harder than a lab report, and neither is wrong.

One more caveat about word counts

Readability formulas depend on the sentence count, and sentence detection is harder than it looks. Abbreviations (“Dr.”, “e.g.”, “et al.”) end in full stops without ending sentences. A tool that gets this wrong will silently overcount sentences, shorten the apparent average, and report your writing as easier than it is.

If you are also checking length against a word limit, be aware that different tools count differently — hyphenated compounds and numbers especially. Our character and word counter shows both figures and states which convention it uses, because “2,000 words” means different things to different counters and your marker’s is the one that counts.