What Flesch-Kincaid Actually Measures and What to Aim For

Two numbers, one common mix-up

Flesch-Kincaid is actually two separate scores, and I spent an embarrassing amount of time confusing them. The Flesch Reading Ease score runs on a scale of roughly 0 to 100, where higher means easier. The Flesch-Kincaid Grade Level translates the same ingredients into a US school grade, where higher means harder. A piece can score 65 on the first scale and 8 on the second, and both numbers are describing the same text as fairly plain English.

I check both on almost everything I publish now, but only after learning what they actually measure, which turned out to be far less than I assumed. The formulas are almost insultingly simple, and that simplicity is both why they work and why they can be gamed into uselessness.

The formulas in plain English

Reading Ease starts at 206.835, then subtracts 1.015 times your average sentence length in words, then subtracts 84.6 times your average syllables per word. That is the entire formula. Grade Level is the same two ingredients reweighted: 0.39 times average sentence length, plus 11.8 times average syllables per word, minus 15.59.

Notice what is in there: sentence length and word length. Nothing else. Not vocabulary difficulty, not logic, not structure, not whether the sentences are in a sensible order. You could shuffle the sentences of this post into random nonsense and the score would not move by a point. Every strength and every failure of these metrics comes directly from that observation.

The practical translation: long sentences drag both scores toward hard, and multi-syllable words drag them harder, since the syllable term carries the heavier weight. Which is why swapping utilize for use and approximately for about moves the needle more than almost any structural edit.

What score should I aim for?

The conventional interpretation of the scale, the one published alongside the formula for decades, is that 60 to 70 counts as plain English, comfortable for a general audience, corresponding to roughly an eighth or ninth grade reading level. Scores below 30 are graduate-territory dense. Scores above 90 read like early primary school books.

My personal targets, which are opinions rather than laws: blog posts and marketing pages, 60 or better. Emails to customers, 70 or better, because nobody rereads a confusing email, they just do not reply. Technical documentation can legitimately sit in the 40s when precision demands it. What I never do anymore is sand a piece down to 85 for its own sake: past a point you are amputating nuance to please an equation from the 1940s.

The grade level number is also worth reframing. Aiming for grade 8 is not writing for children. It is writing that a smart, busy adult can process on a phone, on a train, at half attention, which describes essentially everyone reading anything I publish.

What the scores cannot see

The formulas are word-length detectors, so they are blind to every other way writing fails. Jargon built from short words sails through: churn, burn rate, run rate, top of funnel. A sentence like we need to right-size the org scores as beautifully readable while communicating dread and nothing else. Meanwhile a genuinely useful long word like immediately gets taxed at four syllables.

The federal plain language guidelines are refreshingly blunt on this point: they emphasize writing for your specific audience, organizing logically, and testing with real readers, and they treat formula scores as a rough signal at best. A document can hit every numeric target and still fail its reader, because clarity lives in structure and word choice, not word length. I treat a bad score as a smoke alarm: it reliably tells me something is burning, but a good score does not certify a five-star meal.

How I actually use a readability checker

My editing pass is mechanical and takes about ten minutes per post. I paste the draft into the readability checker and look at three things: the two scores, the longest sentences, and the average sentence length. Anything over 25 words gets read aloud. If I run out of breath or lose the thread, it becomes two sentences. Then one pass swapping the usual suspects: utilize, leverage, facilitate, and their Latinate friends all get replaced with the short words I would say out loud.

Two companion habits. I watch total length with the word counter, because the easiest readability improvement is often cutting the paragraph that was only there for my own satisfaction, and a text summarizer pass shows me quickly which points actually carry the piece. And since clear writing and search performance overlap heavily, the SEO content analyzer runs on the same drafts: short sentences and plain words help the same pages that good headings and meta tags do.

Does readability matter beyond human readers?

Two second-order effects convinced me to keep caring about this. First, comprehension at speed: most of my readers are skimming on phones, and plain sentences survive skimming while ornate ones die. Second, an oddly modern one: plain writing is cheaper to process with AI models. Shorter, more common words tokenize into fewer units, a point I unpacked in AI tokens explained, so the same clarity edit that helps a human on a train literally reduces an API bill. That is the only time the 1940s and the 2020s have agreed on anything in my working life.

What I would not do is chase readability as a search ranking tactic in itself. I have seen no convincing evidence that Google reads Flesch scores directly. The honest causal chain is simpler: readable pages get read further, shared more, and bounced from less, and those human behaviors are the thing worth optimizing.