Readability Scores Explained: Flesch, Flesch-Kincaid and Gunning Fog
The formulas behind Flesch Reading Ease, Flesch-Kincaid Grade and Gunning Fog, what scores mean for web, email and academic text, and how to improve them.
A readability score is a number that predicts how hard a text is to read. It is not a measure of how good the writing is, how correct it is, or whether anyone will enjoy it. It is arithmetic on three things: how long your sentences are, how long your words are, and (for one formula) how many long words you use. That is all. Once you know the formulas, the scores stop being mysterious and start being useful.
The three formulas
Flesch Reading Ease
The oldest and most widely used, developed by Rudolf Flesch in the 1940s. Higher is easier.
206.835 − 1.015 × (words ÷ sentences) − 84.6 × (syllables ÷ words)The output is nominally 0 to 100, though dense text can go negative and very simple text can exceed 100. Word length is weighted far more heavily than sentence length: one extra syllable per word on average costs about 85 points; one extra word per sentence costs about one point.
| Score | Reading level | Typical text |
|---|---|---|
| 90–100 | Very easy | Children’s books, simple instructions |
| 70–90 | Easy to fairly easy | Popular fiction, conversational blog posts |
| 60–70 | Standard | Most newspapers, general web content |
| 50–60 | Fairly difficult | Quality broadsheets, trade magazines |
| 30–50 | Difficult | Academic writing, technical documentation |
| 0–30 | Very difficult | Legal text, dense scientific papers |
Flesch-Kincaid Grade Level
The same two inputs, rescaled to a US school grade. Lower is easier. It was developed for the US Navy in the 1970s to grade training manuals.
0.39 × (words ÷ sentences) + 11.8 × (syllables ÷ words) − 15.59A result of 8.0 means a typical US eighth-grader (age 13–14) should be able to read it. Because it shares inputs with Reading Ease, the two always move together; the grade level is just easier to explain to a client.
Gunning Fog Index
Robert Gunning’s 1952 formula also outputs a grade level, but it counts complex words (three or more syllables) rather than average syllables.
0.4 × [ (words ÷ sentences) + 100 × (complex words ÷ words) ]Strictly, the complex-word count excludes proper nouns, compound words, and words that reach three syllables only through a common suffix such as -es, -ed or -ing. Most online tools skip some of these exclusions, so Fog scores vary between calculators more than Flesch scores do.
A Fog of 12 corresponds to a US high-school senior; Gunning’s recommendation for wide readership was under 12, and for really broad audiences, under 8.
The same passage, three ways
Passage A: “We reviewed the plan. It costs too much. We will cut two items and try again next month.”
Three sentences, 18 words, 21 syllables, no complex words.
Passage B: “Following a comprehensive evaluation of the proposed implementation strategy, the committee determined that the anticipated expenditure significantly exceeded the allocated budgetary parameters.”
One sentence, 22 words, 63 syllables, roughly 13 complex words depending on the suffix rule.
| Passage A | Passage B | |
|---|---|---|
| Avg sentence length | 6 | 22 |
| Avg syllables per word | 1.17 | 2.86 |
| Flesch Reading Ease | ~102 (very easy) | below 0 (very difficult) |
| Flesch-Kincaid Grade | ~0.5 | ~27 |
| Gunning Fog | ~2.4 | over 30 |
Both passages say the same thing. Passage B is not more precise; it is just longer per word and per sentence. That is the whole lesson of readability formulas in one table.
Syllable counting is done by algorithm in every tool, and English syllables are irregular, so two tools will give slightly different numbers for the same text. Differences of a few points are noise. Differences of twenty points are real.
What scores to aim for
There is no universal target. There is a target for each kind of reader.
| Content | Flesch Reading Ease | FK Grade | Why |
|---|---|---|---|
| Web content, blog posts, product pages | 60–75 | 6–9 | People scan on screens; most readers are not in the mood to work |
| Marketing email, newsletters | 65–80 | 5–8 | Read on phones, in a hurry, competing with everything else |
| Internal business email | 55–70 | 7–10 | Colleagues share context, but still want it short |
| Instructions, help content, government | 70+ | 6 or lower | Readers may be stressed, distracted or reading in a second language |
| Journalism | 50–65 | 8–11 | Educated general readers |
| Academic writing | 30–50 | 12–16 | Specialist vocabulary is unavoidable; sentence length is not |
| Legal and technical | 20–40 | 14+ | Precision beats ease, but the ranges above are still a warning sign |
Academic and legal writing score badly by design, because their vocabulary is polysyllabic. Within those fields, though, the writers who score 45 are still easier to read than the ones who score 20, and it is almost always sentence length that separates them.
How to improve a score without dumbing down
“Improve readability” is often heard as “write for children”. It is not. Look at the formulas: they punish long sentences and long words. Neither is the same as sophistication.
Split sentences at the conjunction. A 40-word sentence held together by “and”, “which” and “however” is usually two or three sentences wearing a coat. Splitting changes nothing about the ideas.
Swap the Latinate word for the plain one when the meaning is identical. “Utilize” is “use”. “Approximately” is “about”. “Sufficient” is “enough”. “Commence” is “start”. Each swap saves one to three syllables and loses nothing. Keep the technical terms that carry meaning (“photosynthesis”, “amortisation”); cut the ones that only carry tone.
Cut the ceremony. “It should be noted that”, “in order to”, “for the purpose of”, “with regard to”. These add words and syllables and no content. The list of AI-typical phrases overlaps heavily with this, because language models learned formality from the same corporate prose.
Vary sentence length, do not just shorten it. Uniformly short sentences read as choppy and, as it happens, score as machine-like on AI detectors. Aim for a low average with a wide spread: some 25-word sentences, some 5-word ones.
Keep the vocabulary that the reader needs. A page about mortgage amortisation must say “amortisation”. Explain it once, then use it. The score will be lower than a page about cats; that is fine.
Before (FK Grade ~15): In the event that the customer is unable to access their account subsequent to a password reset, it is recommended that they clear their browser cache and attempt authentication again prior to contacting support.
After (FK Grade ~6): Can’t log in after a password reset? Clear your browser cache and try again. If that doesn’t work, contact support.
Same instructions, same precision, less than half the words. Nobody was talked down to.
What the scores do not measure
- Structure. A wall of short sentences with no headings scores well and reads badly.
- Coherence. Sentences in random order score exactly the same as sentences in the right order.
- Jargon. “Use the API key” is three short words and impenetrable to most people.
- Tone. A curt, easy-to-read email can still be rude.
Use the score as a smoke alarm. When it says a paragraph is hard, look at that paragraph. When it says everything is easy, that is necessary but not sufficient.
A working routine
- Draft without thinking about the score.
- Paste into the readability checker. Look at the grade level, the longest sentences, and the density of complex words.
- Fix the three worst sentences. Usually splitting them is enough.
- Replace ceremonial vocabulary with plain vocabulary, leaving technical terms alone.
- Re-check. If the score is in range for the audience, stop. Do not chase a number past its target.
For long documents, run the summarizer first to check the argument is clear at all; sentence surgery does not fix a missing point. The word counter shows average sentence length directly, the fastest single thing to watch.
FAQ
What is a good Flesch Reading Ease score?
For general web content, 60–70 is the standard target and 70+ is comfortable for broad audiences. Specialist writing (academic, legal, technical) normally sits between 30 and 50, and that is appropriate as long as sentence length is under control. Match the score to the reader, not to a universal ideal.
Why do different tools give different readability scores?
Mostly syllable counting. English has no reliable rule for syllables, so each tool uses its own heuristic, and Gunning Fog tools differ on which words count as “complex”. Small differences between tools are expected; use one tool consistently and watch the direction of change rather than the absolute number.
Does improving readability help SEO?
Not directly; search engines do not publish a readability ranking factor. Indirectly, easier pages tend to keep readers longer and get shared more, and clear writing is part of what Google’s helpful content guidance describes. Treat readability as a reader-experience metric that happens to align with what search engines reward. See Humanizing AI Text for SEO for the broader picture.