Paste the same paragraph into six different readability formulas and you can walk away with six different opinions about how hard it is to read. One says eighth grade. Another says college sophomore. A third says "fairly difficult" without giving you a grade at all. None of them are lying. They are just measuring different things and calling the result the same name.
If you write for a living, or you are the person a writer's readability score gets forwarded to, this inconsistency matters more than it looks like it should. A style guide that says "keep everything under an eighth-grade reading level" is only enforceable if everyone agrees on which formula produced that number. Most teams never ask.
The formulas are not measuring "readability" the same way
Every popular readability formula is a proxy. None of them read for comprehension, tone, or whether the argument makes sense. They count things that are cheap to count, mostly syllables, word length, and sentence length, and then run those counts through a regression equation that was fit to a specific set of graded reading passages decades ago.
That last part matters. Flesch-Kincaid Grade Level and the Flesch Reading Ease score both come from Rudolf Flesch's original work, later recalibrated by the U.S. Navy for training manuals in the 1970s. Gunning Fog was built for business writing and leans harder on "complex words" (three or more syllables). SMOG was designed specifically for healthcare materials and consumer consent forms, where the cost of misjudging difficulty is unusually high. Coleman-Liau skips syllables entirely and counts characters per word instead, which makes it faster to compute but blind to some of what the others catch.
Photo by Esmanur Güler on Pexels
They were built for different documents, calibrated against different graded samples, and optimized for different failure modes. Expecting them to agree is a little like expecting a bathroom scale and a body fat caliper to report the same number just because they are both about "how heavy you are."
What each formula actually counts
Flesch Reading Ease produces a 0-100 score, higher meaning easier. It weighs average sentence length and average syllables per word. A score in the 60-70 range is considered "plain English" for a general adult audience.
Flesch-Kincaid Grade Level takes the same two inputs and outputs a U.S. school grade level instead of a 0-100 score. It is probably the formula you have seen most often, since it is built into word processors.
Gunning Fog Index counts sentence length plus the percentage of "complex" words (three-plus syllables, with some common exceptions like proper nouns and compound words). It tends to run higher than Flesch-Kincaid for the same text because it penalizes long words more aggressively.
SMOG (Simple Measure of Gobbledygook) only looks at polysyllabic word counts across a sample of sentences, then applies a square root transformation. It was designed to be conservative, so it often reports a higher (harder) grade level than Flesch-Kincaid on identical text.
Coleman-Liau Index uses characters per word and sentences per 100 words instead of syllables, which makes it resistant to mispronunciation-based syllable counting errors but more sensitive to things like technical abbreviations and hyphenated terms.
Automated Readability Index (ARI) is similar to Coleman-Liau in using characters rather than syllables, but weights sentence length more heavily, which can push technical or list-heavy writing toward a higher grade level than its actual complexity would suggest.
Why this shows up as a real problem, not just trivia
A newsroom style guide that targets "grade 8 or below" is really targeting whichever formula the CMS plugin happens to use. Swap the plugin, or the underlying library it calls, and the same house style can suddenly appear to be failing, even though not a word of the actual writing changed.
The same thing happens in healthcare and legal content, where regulations sometimes specify a target reading level for patient materials or disclosures. SMOG was built for exactly this use case and tends to be the more conservative (harder-scoring) choice, so a document that clears Flesch-Kincaid grade 6 might still fail a SMOG-based compliance check. If a regulation or internal policy references "readability" without naming the formula, that ambiguity is a real liability, not a nitpick.
"The number people usually want is 'will a reasonable adult understand this on the first read,' but every formula answers a narrower, more mechanical question than that. Pick the one that matches your actual audience and stick with it, instead of chasing whichever score looks best this week." - Dennis Traina, founder of 137Foundry
How to actually use these scores
Pick one formula for one purpose and be explicit about it in your style guide. If you are writing for a general web audience, Flesch Reading Ease or Flesch-Kincaid Grade Level are the most widely understood and the easiest to communicate to a non-technical stakeholder. If you are producing patient-facing healthcare content, default to SMOG since regulators and health literacy researchers reference it most often.
Run a second formula as a sanity check rather than a target. If Flesch-Kincaid says grade 7 and Gunning Fog says grade 12 on the same paragraph, that gap usually means you have a handful of long technical words dragging the Fog score up, not that the writing is secretly much harder than it reads. Checking the Reading Level Analyzer against a paragraph you already know well is a fast way to build intuition for how much any single formula's number should be trusted.
Photo by Buse Çolak on Pexels
Sentence length is the one lever nearly every formula responds to, so if a document is scoring harder than you want across the board, splitting long sentences will move the number more reliably than swapping in shorter synonyms. Watch out for technical vocabulary and proper nouns too. A grade-level score inflated by product names, chemical compounds, or legal terms is not really telling you the prose is hard to follow, it is telling you the formula does not know those words are supposed to be there.
Where the formulas actively mislead you
None of these formulas understand meaning, so they can be gamed in ways that make text objectively worse while the score improves. Chopping a sentence in half with a period instead of a comma will lower a Flesch-Kincaid score even if the resulting two fragments are harder to follow than the original sentence was. Replacing a precise multisyllabic word with a vaguer short one can "improve" readability scores while making the writing less accurate.
This is why readability scores work best as a screening tool, not a quality bar. Use them to catch paragraphs that have clearly gotten away from you, long compound sentences stacked with jargon, not as the final word on whether a piece is well written. A brief that specifies both a target formula and a target audience up front will save more editorial arguments than a bare "keep it readable" instruction ever will.
A quick worked example
Take a single sentence: "The mitochondria, often described as the powerhouse of the cell, generates the majority of the chemical energy needed to power biochemical reactions through a process called cellular respiration." Run it through Flesch-Kincaid and you will likely land around grade 14 to 16, mostly because of "mitochondria," "biochemical," and "respiration" pushing the syllable count up. Run the same sentence through Coleman-Liau, which counts characters instead of syllables, and the score shifts because long technical nouns affect character-per-word averages differently than they affect syllable averages.
Neither number is wrong. They are answering "how hard is this by my counting method," and the two methods count different things. The practical lesson is not to memorize which formula runs high or low on jargon, it is to always report which formula produced a number whenever you share a readability score with someone else. "This scores grade 9" means nothing on its own. "This scores grade 9 on Flesch-Kincaid" is a claim someone else can actually verify or reproduce.
Building this into an editorial workflow
Teams that get real value out of readability scoring usually do three things consistently. First, they pick one formula as the house standard and write it into the style guide by name, not just "check readability." Second, they set the target range by section or content type instead of site-wide, since a technical API reference and a marketing landing page should not be held to the same grade level. Third, they re-check scores after edits, not just on the first draft, since a well-meaning copyedit that swaps in more precise vocabulary can raise the grade level even as it makes the writing better.
That third point trips up more teams than the other two combined. A junior editor who sees a readability score go up after a revision often assumes they made a mistake, when in fact they replaced a vague word with an accurate one and the formula, which cannot tell the difference between "vague" and "precise," just registered a longer word. Knowing which formula you are using, and what it can and cannot see, is what keeps that kind of false alarm from turning into a wasted rewrite.
The takeaway
Readability formulas disagree because they were built to answer different questions using different cheap-to-compute proxies for difficulty, not because one of them is right and the others are broken. Flesch-Kincaid and Gunning Fog will diverge most on jargon-heavy text. SMOG will run more conservative than almost anything else. Coleman-Liau and ARI trade syllable counting for character counting, which changes how they treat technical vocabulary.
Pick the formula that matches your actual use case, name it explicitly in whatever style guide or compliance policy references it, and treat the number as a screening signal rather than a verdict. For a deeper look at how these calculations work in practice, the EvvyTools blog covers more of the tools writers and editors use to sanity-check their own drafts, and the tools directory has the full readability and writing toolkit, including the six-formula breakdown, in one place. You can also start directly from EvvyTools if you would rather browse the full catalog first.
Sources worth a look if you want the formulas straight from the people who study them: the Center for Plain Language publishes practical guidance on plain-language writing standards, Wikipedia's overview of readability links out to the original papers behind each formula, and the National Institutes of Health publishes health literacy guidance that references SMOG directly for patient materials.
