Readability
Readability is how much effort a piece of text takes to read, estimated in practice by formulas that count sentence length and word length rather than by anything that understands the writing.
Last reviewed 15 August 2026 · The HumanFlow team
In plain English
Long sentences are harder to follow than short ones, and long words are harder than short ones. Readability formulas take those two observations, weight them, and produce a single number.
That is genuinely all they do. No formula in common use knows what your sentences mean, whether they are in a sensible order, or whether the argument holds together.
A worked example
Two sentences with identical readability arithmetic — same word count, near-identical syllable counts — and wildly different difficulty.
A. The cat sat on the mat and the dog ran to the park in the rain. B. The bond spread will gap out if the curve inverts and the desk cuts risk.
Both are seventeen short words in one sentence, so every length-based formula scores them as equally easy. A is readable by a seven-year-old.
B is opaque unless you work in fixed income. The difficulty lives entirely in the vocabulary being specialised, and the formulas cannot see specialised — only short.
The reverse failure is just as common. A long sentence, carefully built with parallel clauses, can be effortless to read while scoring as difficult.
Why it matters for AI detection
Readability is the most misused number in writing tools, because it is easy to compute, easy to display, and easy to mistake for a quality score. It is not one. It is a proxy for effort, and only for the part of effort that length explains.
It matters here for a specific reason: readability and detection are constantly conflated. A detector responds to how predictable your word choices were, which is a different quantity from how long your words are. Plain, short writing can be highly unpredictable, and dense academic prose can be almost perfectly predictable.
Writing to hit a target grade level is the failure mode. Chopping sentences to lower the number produces prose that scores better and reads worse, and the number cannot tell you that has happened.
Commonly confused with
- Perplexity
- Readability estimates effort from length. Perplexity measures how expected each word was to a language model. They are unrelated quantities that both happen to produce a number about your prose.
- Clarity
- Clarity is whether a reader understands you, which depends on ordering, evidence and vocabulary. Readability is an estimate of one contributing factor, measured without reading.
Read next
Part of the AI detection glossary.