I Scored 85 on the Flesch Reading Ease Scale — Then Got Flagged as AI. Here's What I Found Out

Alex Chen had been a content marketing manager for six years. She wrote everything by hand — no ChatGPT, no shortcuts. But in early 2025, a new enterprise client started running her articles through an AI content verifier before publishing. Three of her first four submissions came back flagged. "Over 80% AI probability," the report said. She was stunned.

What followed was a two-week investigation that led her somewhere she never expected: the Flesch Reading Ease scale, a formula invented in 1948 that was quietly shaping how modern AI detectors judged her work.

What Is the Reading Ease Scale?

The reading ease scale — most commonly the Flesch Reading Ease score — assigns text a number from 0 to 100 based on two variables: average sentence length and average syllables per word. A score of 90–100 reads like a children's book. A score of 60–70 is standard for most web content. Scores below 30 tend to appear in academic or legal writing. Higher numbers mean easier, faster reading.

Rudolf Flesch developed the formula to help educators assess textbook difficulty. Today it's baked into Microsoft Word, the Hemingway App, and dozens of writing tools. Most professional writers chase a high score as a sign of quality. That instinct, it turns out, can backfire badly.

The Problem With Being "Too Readable"

When Alex ran her own articles through a readability checker, she found her average Flesch score was 81. Excellent by most standards. But the paragraph-by-paragraph breakdown told a different story: every single paragraph scored between 78 and 84. Almost zero variation across the whole document.

That is exactly what AI detectors look for. Understanding how AI detectors work means recognizing they measure statistical consistency, not just word choice. When readability scores stay in a tight band across an entire piece, the text reads as machine-generated. Humans vary. One paragraph is punchy. The next runs long and wanders. A well-placed fragment breaks the cadence. AI systems — even without instruction — naturally optimize toward smooth, consistent output. Alex had spent years training herself to write that way. That was the problem.

What the Reading Ease Scale Actually Measures (And What It Misses)

The formula only cares about sentence length and syllable count. Nothing else. It cannot measure originality, voice, or rhetorical rhythm. A sentence like "Go." scores a perfect 100. A long, vivid sentence built around a concrete image might score a 38. Neither number tells you whether a person or a machine wrote it.

But AI detectors are not using the raw score — they are using variance. Low variance in readability across paragraphs is a statistical fingerprint of AI. Unpredictable variance across a document is a fingerprint of someone who was not thinking about readability at all while they wrote.

This is why AI detection false positives hit professional writers disproportionately. The more deliberately polished your writing, the more it resembles a well-prompted model. Alex was a victim of her own competence.

How Alex Fixed It — And What Changed

She made three changes over about two weeks.

• She stopped optimizing every paragraph for clarity. Long sentences stayed long when the idea warranted it. Short ones stayed short. She kept contractions she used to edit out and let the occasional fragment land where it felt right.
• She started checking before submitting, not after. Using WriteMask's free AI detector, she could see which sections were flagging and roughly why — rather than receiving a rejection with no explanation and no path forward.
• For any draft that had AI-assisted portions, she ran it through WriteMask. The tool does not just swap synonyms — it restructures sentence patterns and introduces the kind of readability variance that human writers produce naturally. Her pass rate went from one out of four to consistently above 90%, which aligns with WriteMask's documented 93% pass rate across major detection platforms.

Two weeks. From "what is happening to my content" to a workflow that held up every time.

What This Means for Anyone Who Writes Professionally

The reading ease scale is worth understanding — not so you can game it, but because it reveals what AI detectors are actually measuring. They are not reading your ideas. They are reading your patterns.

Vary sentence length deliberately. Let a complex thought stay complex. Let a simple one be blunt. Use a readability checker to audit your paragraph-level variance, not just your document average. And if you are getting flagged despite writing your own work, the techniques covered in how to humanize text for AI detectors apply across nearly every platform — not just Turnitin.

Alex still writes everything herself. She just stopped trying to be perfectly readable. As it turns out, that is exactly what makes writing sound human.