Automated Readability Index Calculator

ARI: 4.71·(chars/words) + 0.5·(words/sentences) − 21.43

E.A. Smith and R.J. Senter built the Automated Readability Index in 1967, originally so the US Air Force could grade technical manuals being typed on electric typewriters. The trick that set it apart was counting characters rather than syllables. That choice made it the first readability formula a machine could run without needing a phoneme dictionary. The formula itself is ARI = 4.71·(chars/words) + 0.5·(words/sentences) − 21.43, and you round the result up to a US grade level somewhere between 1 and 14. Say you feed it 500 chars, 100 words and 5 sentences: 4.71·5 + 0.5·20 − 21.43 ≈ 12.1, so 12th grade. Because it never tries to count syllables, it sidesteps the kind of errors that trip up Flesch and SMOG on web text.

Applications

It shows up in automated content analysis (CMSs, content pipelines), live scoring inside editors like Grammarly and Hemingway, bulk grading of technical docs and chatbot replies, and A/B tests on landing-page copy. It's also handy for accessibility audits on text where syllable counting falls apart, think URLs, code snippets and abbreviations.

FAQ

Why characters instead of syllables? Splitting words into syllables needs a phonetic dictionary, and it falls over on proper nouns, made-up words and abbreviations. Counting characters always gives the same answer and doesn't care what language you're in.

Do punctuation and spaces count? Most implementations look at letters and digits only. Spaces, punctuation marks and line breaks are left out of the character tally.

ARI vs Flesch-Kincaid? Both report a US grade level. ARI usually lands 1–2 grades higher, since long words pack in more characters and that weighs heavily. It also holds up better on text full of abbreviations and numbers.

Can I use ARI for Portuguese? You can. The character counting doesn't depend on language, though keep in mind the constants were tuned for English. Expect the grade scale to be in the right ballpark but a touch inflated for PT-BR, where words run longer on average.