Flesch-Kincaid explained: reading ease, grade level, and what a good score is

Flesch-Kincaid is the name for two readability formulas built on the same two counts: words per sentence and syllables per word. Flesch Reading Ease scores text from 0 to 100, where higher means easier. The Flesch-Kincaid Grade Level turns the same counts into a US school grade.

Both measure length, not meaning. A page can score well and still fail to answer its question. A clear page full of necessary technical terms can score badly. This guide gives both formulas with their sources, works one example by hand, sets out what a good score is according to the people who publish guidance, and shows where the numbers mislead. To score your own text, paste it into the free readability checker.

The Navy figure is from the 1975 report’s abstract^[3]. Microsoft’s range is from its Word help page^[5]. Flesch’s 83% is his own 1949 estimate, reprinted by DuBay, and is not a current measure^[2]. The 12.9 is the widest gap in a study of eight online calculators and ten health texts, even when the calculators used the same formula^[9].

The two Flesch-Kincaid formulas

Both formulas use three counts: words, sentences and syllables. From those come two averages. Average sentence length (ASL) is words divided by sentences. Average syllables per word (ASW) is syllables divided by words.

Microsoft prints both formulas in this form in its Word help, and explains the grade with an example: “a score of 8.0 means that an eighth grader can understand the document”^[5]. Here they are as plain text and as spreadsheet formulas, if you want to check a score yourself.

Flesch Reading Ease
  = 206.835 - 1.015 x (words / sentences) - 84.6 x (syllables / words)

Flesch-Kincaid Grade Level
  = 0.39 x (words / sentences) + 11.8 x (syllables / words) - 15.59

In a spreadsheet, with words in B1, sentences in B2 and syllables in B3:
  Reading Ease:  =206.835-1.015*(B1/B2)-84.6*(B3/B1)
  Grade Level:   =0.39*(B1/B2)+11.8*(B3/B1)-15.59

Where they come from

Rudolf Flesch published Reading Ease in 1948, in the Journal of Applied Psychology^[1]. His paper writes the word term with syllables per 100 words, as 0.846 × wl. Syllables per 100 words is 100 times syllables per word, so 0.846 × wl is the same as the 84.6 × ASW most tools print. Flesch set the scale so that a score of 100 “indicates reading matter that is understandable for persons who have completed fourth grade”^[2].

The grade-level version is not Flesch’s own. J. Peter Kincaid and three colleagues derived it in 1975 for the US Navy, in Research Branch Report 8-75. They recalculated three older formulas, Flesch’s among them, from tests of 531 Navy enlisted personnel and their understanding of 18 passages from Navy training manuals^[3]. The study tested each reader’s reading grade alongside their understanding of the passages, and used both to give each passage a grade level. That is why the result is a school grade. DuBay’s history of the field records that the US Department of Defense authorized the formula in 1978 for checking the readability of technical manuals^[4].

A worked example, step by step

Here is a three-sentence returns notice we wrote for this example. It is short enough to count by hand, so you can check every number.

You can return anything you buy within a month. Pack it in the original box and print the free label from your account. We refund the card you paid with when the parcel reaches the warehouse.

1. Average sentence length: 36 words ÷ 3 sentences = 12.
2. Average syllables per word: 49 ÷ 36 = 1.361.
3. Reading Ease: 206.835 - (1.015 × 12) - (84.6 × 49 ÷ 36) = 206.835 - 12.18 - 115.15 = 79.5.
4. Grade Level: (0.39 × 12) + (11.8 × 49 ÷ 36) - 15.59 = 4.68 + 16.06 - 15.59 = 5.15, about grade 5.

So the passage scores 79.5, in the band Flesch called fairly easy, and a grade level of about 5. Notice that the two results do not quite agree. Flesch’s own table puts 70 to 80 at seventh grade, while the Kincaid formula says fifth. Neither is wrong. They are fits to different readers, made decades apart: school reading tests for Flesch, Navy enlisted personnel for Kincaid^[2][3].

Paste the passage into a checker and you should see numbers close to these, but not always identical. Software has to estimate syllables, and that is the first place scores drift.

Flesch Reading Ease score bands

Flesch described his scale in seven bands. The style names, school grades and reader estimates below are from his 1949 book The Art of Readable Writing. The typical sentence lengths, word lengths and reading matter are from his 1948 paper. Both tables are reprinted in William DuBay’s collections of the classic studies^[2][4].

Two details the band table hides. First, the scale has no floor or ceiling. Flesch wrote that scores fall between 0 and 100 “for almost all samples taken from ordinary prose”^[2], not for all. Text made only of one-word sentences of one-syllable words would score 121.22, and very difficult text can score below zero^[10]. Second, Flesch warned that the link between his score and school grade “holds true only up to about seventh grade; beyond that, the formula under-rates grade level to an increasing degree”^[2].

What is a good Flesch-Kincaid score?

There is no official target for web pages. What exists is guidance from a handful of sources. For general readers, most of it lands in the same place.

Microsoft’s help page gives the 60 to 70 range and the 7.0 to 8.0 grade target^[5]. WCAG’s reading level criterion asks for supplemental content, or a simpler version, when text “requires reading ability more advanced than the lower secondary education level”. It defines that as more than nine years of school^[6]. Digital.gov’s plain language guide, where plainlanguage.gov now redirects, calls it a myth that you have to “dumb down” content, and adds: “Only write for 8th graders if your audience is, in fact, an 8th-grade class”^[7].

Yoast is worth a note. In version 19.3 its SEO plugin replaced the Flesch score with word complexity and sentence length checks. The score now sits in an Insights tab, and Yoast says “we won’t use this assessment in our readability analysis anymore”^[11].

Our reading of that guidance, for a page meant for the general public: aim for 60 or higher on Reading Ease, or grade 8 or below, and treat it as a prompt to look, not a pass mark. For an audience of specialists, a lower score is fine when the terms are ones they use every day. Write for the reader you actually have.

Where the scores mislead

Every readability formula reduces a page to two averages. That is what makes them fast, and it is also where they go wrong. Five problems are worth knowing before you act on a score.

Syllables are estimated, and tools disagree

A person can count syllables by ear. Software cannot, so it estimates them from spelling rules or a word list. Is “every” two syllables or three? Is “our” one or two? People say both. Flesch’s own instructions told readers to “use any good dictionary” when in doubt, and to count numbers and symbols as they are read aloud: four syllables for 1918, two for $^[2]. A tool that counts “1918” as one short word gets a different answer.

Small differences add up. In the 36-word example above, one disputed syllable moves Reading Ease by 2.35 points (84.6 ÷ 36) and the grade by 0.33. Across tools the gap is far larger. In a 2022 study, eight online calculators scored the same ten health texts, and “the same text produced scores that varied by up to 12.9 grade reading levels even when using the same formula”^[9]. Preparing the text first, for example by removing incomplete sentences, reduced the spread in most cases. But for 5 of the 10 texts, that meant leaving out more than 20% of the text. Microsoft Word’s Flesch-Kincaid scores were among the few that agreed well with scores counted by hand^[9].

Wagner tested five open source readability libraries and found the same thing from the code side: “subtle differences in how they compute word and sentence complexity, which can significantly alter the resulting readability scores”^[10]. The practical rule: compare scores from one tool only. A 62 in one checker and a 58 in another may be the same text.

Short words and short sentences can still be hard

The formulas assume a short word is an easy word. Often it is, but not always. The Language Portal of Canada gives the examples “ennui”, “writ”, “fez” and “nadir”: short, and hard for many readers, yet “a formula will score them as easy words”^[8]. Jargon and acronyms work the same way. “Set the TTL. Purge the CDN. Check the ETag.” is three short sentences that mean nothing to most readers. By our count it scores above 70 whichever way a tool counts the acronyms. The same article quotes the US Agency for Healthcare Research and Quality: the formulas “can’t tell you whether the words you are using are familiar to your readers”^[8].

Lists, headings and missing full stops

Formulas find sentences by looking for full stops. A bulleted list with no full stops can be read as one very long sentence, which drags the score down even though lists help readers. Ginny Redish, quoted by the Language Portal of Canada: “If you use bulleted lists to chunk your material and lay your text out with white space, readability formulas will say you have long sentences”^[8]. Headings can push the other way. A tool that counts each heading as its own short sentence lowers the average sentence length and lifts the score. Wagner traced part of the gap between libraries to exactly this: differences in “how partial sentences, such as headings, are counted”^[10]. Score the body text, and know what your tool does with the rest.

Other languages

Both formulas were fitted to English readers and English text. Yoast’s methodology notes say that “in languages other than English, Flesch is unreliable”^[12]. For some European languages inside an English document, Microsoft Word “displays only information about counts and averages, not readability”^[5]. WCAG suggests scoring each language on a page separately and judging the page on the worst result^[6]. For text in another language, use a formula built for that language, or skip the score.

They cannot see meaning

No formula reads for sense. It cannot tell whether the answer comes first, whether the steps are in order, or whether the page answers the question at all. Research summarised by the Language Portal of Canada goes further. Richard Kern’s work, published in 1980, found that rewriting to lower the reading-grade score “does not increase comprehension”^[8]. The ISO plain language standard, published in 2023, sets formulas aside in so many words: plain language “focuses on how successfully readers can use the document rather than on mechanical measures such as readability formulas”^[8].

How to raise a Flesch-Kincaid score

Start with the arithmetic, because it tells you where the points are. In Reading Ease, each word you cut from the average sentence adds 1.015 points. Each 0.1 you cut from the average syllables per word adds 8.46 points. In the grade formula, the same two changes lower the grade by 0.39 and 1.18. So cutting 0.1 syllables per word is worth about as much as cutting eight words from every sentence.

Here is a corporate version of the returns notice from the worked example. It says the same thing.

Customers may return merchandise within one month of purchase. Items should be repackaged in the original packaging, and a prepaid label can be generated through the customer account portal. Reimbursement to the original payment method is processed after the parcel arrives at the distribution facility.

The score rose by about 51 points. About 48 of them came from shorter words: the syllable term fell from 163.56 to 115.15. Shorter sentences gave about 3: the sentence term fell from 15.225 to 12.18. “Reimbursement” became “refund”. “Merchandise” became “anything you buy”. The passive “is processed” became “we refund”. What helped most:

• Swap long words for short, familiar ones. “Use” for “utilise”, “buy” for “purchase”, “help” for “assistance”. Familiar matters more than short.
• Split long sentences at “and”, “which” and “but” when each half can stand alone.
• Say who does what. Active sentences are often shorter. See active and passive voice for when passive is still the right choice.
• Cut filler. “In order to” is “to”. “At this point in time” is “now”.
• Keep the terms your readers need, and explain each one once. Swapping a precise term for a vague one lifts the score and costs clarity.
• Mark up lists as lists, so a tool does not read them as one long sentence.

Do readability scores matter for AI search?

Start with what is documented. Google says “there are no additional requirements to appear in AI Overviews or AI Mode, nor other special optimizations necessary”^[14], and its guidance on helpful content names no readability score^[15]. OpenAI’s, Anthropic’s and Perplexity’s documentation for site owners, read on 2 October 2026, does not mention readability either^[16][17][18]. Nobody outside those companies knows whether a Flesch score plays any part in what gets cited.

Some SEO vendors say it does. Yoast’s guide to writing for AI advises: “aim for a Flesch score above 60, especially for top-of-funnel or FAQ-style content you want to be quoted or summarized”^[13]. It cites no evidence from an AI engine for that number. Treat it as a vendor’s opinion.

What we can say without inventing data is narrower, and it is reasoning, not a measured effect. An AI answer quotes or paraphrases passages. A passage that answers the question on its own, in plain sentences, near the top of the page, is easier to lift than an answer spread over five paragraphs. That is a point about structure more than about scores. A direct answer under the H1 is one of the checks in our AI search checklist, and how AI assistants choose what to cite covers what is actually known about selection. For writing answers in that shape, see the guide to answer engine optimization.

Checking a whole site

A readability checker scores one passage at a time. CoreCited’s Site Audit runs across your site and flags pages that lack a short, direct answer straight under the H1. It does not grade your prose; that is a job for an editor and a checker like the one above. The free account audits 10 pages, and every limit is on the pricing page.

Questions people ask

Sources