Automated Readability Index (ARI): formula and worked example

The Automated Readability Index, or ARI, has the most honest name of any readability formula. It was designed to be computed automatically. E. A. Smith and R. J. Senter developed it for the US Air Force in 1967, at a time when counting syllables meant a person with a pencil. ARI replaced syllables with characters, which a machine can count without judgment. The original work reportedly fitted counters to an electric typewriter, so a text could be measured as it was typed.

Eight years later, the US Navy study that produced the Flesch–Kincaid Grade Level also recalculated ARI for Navy readers. Both formulas output a US school grade.

The formula

ARI = 4.71 × (characters ÷ words) + 0.5 × (words ÷ sentences) − 21.43

• Characters means letters and digits. Spaces and punctuation don’t count.
• Words ÷ sentences is average sentence length.

Each part has a clear price:

• One extra character per word, on average, adds 4.71 grades.
• One extra word per sentence, on average, adds 0.5 of a grade.

So a 10-word increase in average sentence length costs about 5 grades, roughly the same as one extra character in every word. ARI gives sentence length real weight, more than Coleman–Liau does.

A worked example

Before: Users must authenticate utilizing their organizational credentials before accessing the reporting dashboard, which aggregates performance metrics across all regional departments.

20 words · 1 sentence · grade 21.7

The grade shown with each example is Flesch–Kincaid. The arithmetic below is ARI, so expect different numbers.

Counts: 20 words, 157 characters, 1 sentence.

• Characters per word: 157 ÷ 20 = 7.85, and 4.71 × 7.85 ≈ 36.97
• Words per sentence: 20 ÷ 1 = 20, and 0.5 × 20 = 10
• ARI: 36.97 + 10 − 21.43 ≈ 25.5

After: Sign in with your work account to open the dashboard. It shows sales figures for every region.

17 words · 2 sentences · grade 4.4

Counts: 17 words, 76 characters, 2 sentences.

• Characters per word: 76 ÷ 17 ≈ 4.47, and 4.71 × 4.47 ≈ 21.06
• Words per sentence: 17 ÷ 2 = 8.5, and 0.5 × 8.5 = 4.25
• ARI: 21.06 + 4.25 − 21.43 ≈ 3.9

Most of the drop came from word length. “Authenticate utilizing their organizational credentials” became “sign in with your work account.” The rewrite also names the reader (“your”) and replaces the vague “performance metrics” with something concrete. In real use, check that the concrete version is true: if the dashboard shows more than sales, say so.

Short samples like these exaggerate. Use a few paragraphs to get a grade you can compare.

A quick check by hand

Because the inputs are so simple, you can estimate ARI without software. Suppose a 100-word passage has 470 characters and 5 sentences:

• 4.71 × (470 ÷ 100) = 4.71 × 4.7 = 22.137
• 0.5 × (100 ÷ 5) = 0.5 × 20 = 10
• 22.137 + 10 − 21.43 ≈ 10.7

Strengths

Consistency. Any two tools that agree on what counts as a word, a sentence, and a character will produce the same ARI. Syllable-based scores can’t promise that, because syllable counters disagree. See counting syllables.

Speed. Counting characters is trivial for software, so ARI suits long documents and live feedback while you type.

A sensible weight on sentence length. ARI reacts clearly when you split a long sentence, which matches most editors’ experience of what helps readers.

Weaknesses

Characters are a rougher proxy than syllables. “Strengths” has nine characters and one syllable. “Idea” has four characters and three syllables. Over a long text this mostly evens out. Over a paragraph it may not.

Numbers distort it. Digits count as characters, so a year like “2026” is a four-character word and a long order number is a very long word, though neither is hard to read. Tables of figures can push ARI up for no good reason.

Sentence boundaries matter a lot. With 0.5 grade per word of average length, a missed sentence break inflates the score fast. Bulleted lists without end punctuation, headings, and abbreviations like “e.g.” all affect where a tool thinks sentences end. Tools handle these differently, which is one reason the same text can score differently in two checkers.

All the usual blind spots. Like every formula, ARI ignores meaning, word order, familiarity, and layout. “KPI” looks easy; “grandmother” looks hard. Readability scores explained covers these limits in more detail.

Lowering an ARI grade

Work on the two inputs in order of payoff:

1. Use shorter words where they mean the same thing. “Use” for “utilize,” “start” for “initiate,” “show” for “demonstrate.” The plain words list has more.
2. Split long sentences at natural joints. The sentence length checker flags anything over 25 words. Our guide to shorter sentences shows where to cut.
3. Leave needed terms alone. If your reader needs the word “authentication,” use it. A lower grade is not worth a vaguer instruction.