
Feedback & Revision
What Reading Level Should a Novel Be? A Practical Guide
Memoirist and former hospice chaplain writing about grief, memory, and finding meaning late.
There is no required reading level for an adult novel, but popular fiction usually measures low on the formulas. Bestsellers tend to come out somewhere around a 4th to 9th grade level on grade-level formulas such as Flesch-Kincaid, even literary classics, so a manuscript that scores at grade 12 or above is unusually dense for fiction rather than unusually good.
Treat the number as a rough gauge of sentence length and word length, nothing more. Check it scene by scene, compare it with what each scene is trying to do, and change sentences only where the density is getting in the reader's way.
StoryWith AI is our own product, so weigh what we say about it accordingly.
Key takeaways
- Flesch Reading Ease and Flesch-Kincaid Grade Level measure two things only: average words per sentence and average syllables per word. They know nothing about plot, character or meaning.
- Famous novels score low. A Readable analysis puts To Kill a Mockingbird at Flesch-Kincaid grade 4.6 and 1984 at 8.9.
- A high grade level is not a quality verdict, and a low one is not a flaw. The formulas cannot see whether a long sentence is earning its length.
- The formulas are regressions fitted to English. Applied to other languages, the numbers do not mean what they say.
- Check per chapter or per scene. A whole-book average hides the three dense chapters that are actually slowing readers down.
- To lower the score, shorten the longest sentences, swap long words for plain ones, break dense paragraphs and let dialogue carry more of the scene.
What do readability formulas actually measure?
Both Flesch formulas take the same three counts from a passage: total words, total sentences and total syllables. Nothing else goes in.
Flesch Reading Ease (FRE):
FRE = 206.835 - 1.015 x (words / sentences) - 84.6 x (syllables / words)
Flesch-Kincaid Grade Level (FKGL):
FKGL = 0.39 x (words / sentences) + 11.8 x (syllables / words) - 15.59
Sally Jenkins's Goodreads blog post Flesch Reading Ease and Flesch-Kincaid Grade Level test (published 3 April 2017) describes the two scales plainly: "The Flesch Reading Ease index works on a 100 point scale, the higher the index, the easier a document is to understand." And: "The Flesch-Kincaid Grade Level gives a manuscript a US school grade level."
The two run in opposite directions: higher Reading Ease means easier text, higher grade level means harder text.
The Secret Importance of Readability on DIY MFA, bylined Jeanette the Writer (published 26 August 2019), names the limit directly: readability scales "only look at the style of the text, not the content." She adds: "No scale can measure how compelling a character is or how beautifully a theme is conveyed."
What reading level are popular novels?
Lower than most writers expect. The sources below each used their own sample and method, so read them as indications, not a standard.
| Source | What it measured | Finding |
|---|---|---|
| Readable, "Popular fiction and readability" (Laura Kelly, 23 January 2020, updated 21 July 2020) | Flesch-Kincaid grade of popular Goodreads picks | To Kill a Mockingbird 4.6, The Great Gatsby 5.5, 1984 8.9 |
| Same Readable article | A general claim it attributes to George Klare | "most popular novels are written at a 7th-grade reading level" |
| Shane Snow, "What Reading Level Should You Write At?" (no date shown; read on 1 November 2026) | Bestselling books, scored as an average of five readability calculators | "Note how none of these guys wrote above a 9th grade level." |
| DIY MFA (26 August 2019) | General guidance for wide audiences | "Universal readability (about 8th-grade level) is recommended to reach the widest audience possible." |
Snow's post also notes that "Cormac McCarthy, Jane Austen, and Hunter S. Thompson join J.K. Rowling in the readability realm of pre-teens." Readable says of the school-age grade levels: "This doesn't mean they're teen fiction, necessarily."
On the Reading Ease scale, Jenkins's only guidance is general rather than fiction-specific: "A score between 60 and 70 is acceptable for most documents."
None of these sources breaks the numbers down reliably by genre, so we will not invent genre ranges here. The honest pattern is simpler: novels that a very wide audience reads tend to score at or below about grade 9.
Is a high reading level bad for a novel?
Not by itself. A grade level is a summary of sentence and word length, and long sentences are sometimes exactly right: a slow, winding sentence in a moment of dread, or a narrator whose voice is ornate on purpose.
What a high score does tell you is where to look. If one chapter measures far denser than its neighbours, ask whether that density is a choice. If a fight scene reads at grade 13, the sentences are probably too long for the speed the scene needs, which ties into pacing in writing.
Shane Snow, who argues for simpler writing, still adds a caveat on the same page: "Of course, nobody's going to be excited enough to read or talk about something just because it's easy."
The reverse also holds. A very easy score across a whole book can mean every sentence has the same short, flat shape.
Do readability formulas work for books not written in English?
No. Both Flesch formulas are fitted constants: the 206.835, 1.015 and 84.6 came from English text and English reading tests. The syllable counting that feeds them also depends on English spelling patterns, such as silent final e.
Spanish, Lithuanian or Japanese prose run through the same arithmetic produces a number, but it does not describe how hard that text is for its readers.
Sentence length is different. Words per sentence and paragraph length can be counted in any language, and they remain useful signals on their own.
Should you check reading level for the whole book or per chapter?
Per chapter at least, and per scene where you can. A whole-book average of grade 7 can contain a prologue at grade 12 and a run of dialogue chapters at grade 3. The average describes neither.
Per-scene numbers let you ask a better question: does this scene's density match its job? A reflective chapter can carry longer sentences than a chase.
Two cautions:
- Very short passages score erratically. A 20-word scene is two or three sentences, and one long one swings the result. Judge short scenes by reading them aloud.
- Compare with yourself, not with a famous novel. Your own lightest and densest scenes show you your range. A gap of several grades between similar scenes is more useful information than a gap between you and The Great Gatsby.
How do you lower the reading level of your book?
The formula has two levers, so every effective change pulls one of them.
| Change | Formula lever | When it helps | When to leave it alone |
|---|---|---|---|
| Split long sentences | Words per sentence | Action, tension, openings, any sentence a reader has to reread | A deliberate long sentence that builds rhythm or mood |
| Swap long words for plain ones | Syllables per word | Latinate narration ("commenced", "approximately", "deliberated") | Technical words your world needs, or a character's own vocabulary |
| Break dense paragraphs | Neither directly, but readers feel it | Blocks where the action, speaker or focus shifts mid-paragraph | A paragraph that is one continuous thought |
| Let dialogue carry more | Usually shorter sentences | Scenes summarised in narration that could be played out | Interior chapters where the point is what a character does not say aloud |
| Cut filler and hedges | Both, slightly | "She began to slowly realise that" type constructions | Voice-driven narration where the hedging is characterisation |
Splitting sentences does the most work, because the formula weighs average sentence length heavily. Word choice comes second, because each extra syllable per word costs a lot on the Reading Ease scale (84.6 points per unit).
Small, repeated habits add up too. Filter words such as "she noticed that" or "he felt" add words to sentences without adding anything a reader can see.
What does lowering the reading level look like, with real numbers?
A hypothetical example, invented for this guide. Syllables were counted by hand using standard dictionary syllable breaks; software that estimates syllables may land a little differently.
Before (3 sentences):
Eleanor, considering the extraordinary circumstances surrounding her grandmother's unexpected disappearance, deliberated carefully before deciding that investigating the abandoned observatory herself was her only reasonable alternative. The building, which had remained unoccupied for approximately thirty years, stood on an isolated hillside overlooking the village. She gathered her equipment, notified nobody, and departed immediately.
Counts: 52 words, 3 sentences, 132 syllables. The sentences run 25, 18 and 9 words.
- Words per sentence: 52 / 3 = 17.33
- Syllables per word: 132 / 52 = 2.538
- FRE = 206.835 - (1.015 x 17.33) - (84.6 x 2.538) = 206.835 - 17.59 - 214.71 = -25.5
- FKGL = (0.39 x 17.33) + (11.8 x 2.538) - 15.59 = grade 21.1
Yes, the Reading Ease score is negative. The scale is not truly capped at 0; very long words push it below. Notice that the sentences are not especially long. The syllables are doing almost all the damage.
After (7 sentences):
Gran was gone. Nobody knew where. Eleanor thought about it for a long time, and then she made up her mind. She would go up to the old observatory herself. No one had lived there for thirty years. It sat alone on the hill above the town. She packed a bag, told no one, and left.
Counts: 56 words, 7 sentences, 69 syllables.
- Words per sentence: 56 / 7 = 8.00
- Syllables per word: 69 / 56 = 1.232
- FRE = 206.835 - (1.015 x 8.00) - (84.6 x 1.232) = 206.835 - 8.12 - 104.23 = 94.5
- FKGL = (0.39 x 8.00) + (11.8 x 1.232) - 15.59 = grade 2.1
The rewrite is four words longer and reads at grade 2. It also lost some things: "grandmother's unexpected disappearance" had a formality that might have been Eleanor's voice. A real revision would probably land between the two, perhaps keeping one longer sentence for the decision itself. The score tells you what changed; it cannot tell you which version is right for the book.
Where StoryWith AI fits
StoryWith AI reads the manuscript you wrote and returns findings. It does not rewrite your prose to hit a score.
The relevant feature is the Readability Flow page, listed as Readability under Diagnostics in the studio, for both fiction and non-fiction books. "Recompute all" measures every scene in your browser with no AI model involved, so it does not use AI text actions, and the page is not restricted to a paid plan. For each scene of at least 25 words it shows:
- Cognitive load, a bar from 0 to 100 that blends sentence length, word length and paragraph density, with a band: Light, Smooth, Dense or Heavy. Scenes are sorted heaviest first.
- Reading ease, the Flesch Reading Ease score shown as a word and a number: very easy (90 and up), easy (70 to 89), plain (60 to 69), fairly hard (50 to 59), hard (30 to 49) or very hard (below 30). On this page the score is kept between 1 and 100.
- Words per sentence, the scene's average.
A "Scenes to look at" list then names each scene that trips a check: more than 25% of sentences over 30 words, paragraphs averaging over 180 words (or one unbroken block), more than 1.7 syllables per word on average, or an average sentence over 28 words. If the manuscript changes after a run, the page says the results are out of date and offers a re-run. Collaborators without edit rights can view results but not recompute them.
Reading ease and cognitive load are only scored when the book's language is English. For other languages the page says so up front and measures structure alone: sentence length, paragraph density and the sentence and paragraph checks still appear.
On the Write page, the quality panel shows the open scene's numbers as you work, also computed locally: words, sentences, average and longest sentence, Flesch ease (unclamped, so a dense passage can show a negative score), Flesch-Kincaid grade level, dialogue percentage and adverb percentage. It flags individual sentences over 35 words, three sentences in a row that start with the same word, and sentence lengths that stay monotonous, plus English-only checks for passive voice, weak verbs and adverb-heavy passages. The whole-book drift scan in the same panel is a separate AI feature on the Pro plan.
Plans, listed on storywith.ai/pricing on 1 November 2026: Free has 50 one-time AI text actions; Author is $19 a month or $149 a year with 500 AI text actions a month; Pro is $39 a month or $299 a year with unlimited AI text actions. StoryWith AI runs on the web and Android, not iOS, and manuscripts are not used to train AI models.
It is the wrong tool in three cases. If you write for children or teens and need a school leveling measure, StoryWith AI does not compute Lexile or any similar system; MetaMetrics describes a Lexile text measure (no date shown; read on 1 November 2026) as "a measure of the text complexity for a book or piece of text," assigned "when it's analyzed by MetaMetrics." If your book is not in English, it will not give you a reading ease score at all. And if you want a judgment on whether your style works, the formulas are crude for everyone; a human reader or editor, or a pass with our self-editing checklist, will tell you more.
Frequently Asked Questions (FAQ)
What reading level should an adult novel be?
There is no rule, but widely read novels tend to measure low. A Readable analysis puts To Kill a Mockingbird at Flesch-Kincaid grade 4.6 and 1984 at 8.9, and Shane Snow's look at bestsellers found none above a 9th grade level. A score well above that suggests your sentences or words run long for fiction, which is worth checking but is not a flaw by itself.
What is a good Flesch Reading Ease score for fiction?
Higher scores mean easier text on a scale that mostly runs from 0 to 100. Sally Jenkins's general guidance is that "a score between 60 and 70 is acceptable for most documents," and popular fiction often scores easier than that. Use your own scenes as the comparison: look for chapters far denser than the rest of your book.
How do I find the grade level of my novel?
Run your text through a Flesch-Kincaid calculator, such as the one built into many word processors, or compute it yourself: 0.39 x words per sentence, plus 11.8 x syllables per word, minus 15.59. Check chapter by chapter rather than the whole file, because a whole-book average hides the dense sections.
How can I lower the reading level of my book?
Split the longest sentences first, since average sentence length drives the score. Then replace long, formal words with plain ones where the meaning allows, break dense paragraphs where the action or speaker shifts, and play summarised moments out in dialogue. Recheck the scene afterwards and read it aloud to make sure the voice survived.
Do I need Lexile for a children's or YA book?
If schools, libraries or an educational publisher are part of your plan, a dedicated leveling system such as Lexile is the measure they work with, and Flesch-Kincaid is not a substitute. Lexile text measures are assigned through analysis by MetaMetrics. For adult fiction, Flesch scores are enough as a rough density check.





