Content readability: why Flesch scores matter for AI
What is readability and why does it matter for AI?
Readability measures how easy a text is to understand. For human readers this has been an important quality criterion for decades, but for AI models it may be even more important. Language models process text based on patterns in word usage, sentence structure and word frequency. The clearer and more consistent your writing style, the better an AI model can interpret and summarize your message.
Research shows that AI models cite texts with average readability most accurately. Extremely simple texts lack nuance, while extremely complex texts lead to interpretation errors. The sweet spot lies with content that is professional and informative, without unnecessary jargon or endless sentences.
Readability is one of the core pillars of Answer Engine Optimization (AEO). It is closely connected to your heading structure and your E-E-A-T signals: together these factors determine how well AI models can process your content and present it as a reliable source.
The Flesch Reading Ease formula
The most widely used measure of readability is the Flesch Reading Ease (FRE) score. This score was developed by Rudolf Flesch in 1948 and calculates readability based on two factors: average sentence length and average number of syllables per word.
Flesch Reading Ease = 206.835
- (1.015 x average words per sentence)
- (84.6 x average syllables per word)
Example calculation:
- Text with 20 words per sentence, 1.5 syllables per word
- FRE = 206.835 - (1.015 x 20) - (84.6 x 1.5)
- FRE = 206.835 - 20.3 - 126.9
- FRE = 59.6 (fairly easy to read)Score interpretation
The FRE score ranges from 0 to 100, with higher scores indicating simpler text. For web content and AI readability, the following ranges are relevant.
- 90 to 100: very easy, suitable for elementary school level. Too simple for professional content.
- 80 to 89: easy, conversational language. Good for consumer-facing content.
- 70 to 79: fairly easy. Ideal for most web content.
- 60 to 69: standard. Good for professional blogs and articles. This is the sweet spot for AEO.
- 50 to 59: fairly difficult. Acceptable for technical trade publications.
- 30 to 49: difficult. Academic level, too complex for most web content.
- 0 to 29: very difficult. Legal texts and scientific papers. Avoid this for AEO.
Dive deeper: The heading hierarchy for humans and machines | Alt texts for images | How AI Overviews are changing search results
Adapting Flesch for different languages
The original Flesch formula was developed for English texts. For Dutch, the Flesch-Douma variant exists, which accounts for the typically longer words in Dutch. The formula is adjusted but the principle is the same: shorter sentences and simpler words lead to a higher score.
In practice, Dutch texts score slightly lower on average than English texts with a comparable complexity level. Take this into account when interpreting your scores: an FRE of 55 for a Dutch text is comparable to an FRE of 65 for an English text.
// Flesch-Douma formula for Dutch:
Flesch-Douma = 206.835
- (0.93 x average words per sentence)
- (77.0 x average syllables per word)
// Difference from original Flesch:
// - Factor for sentence length: 0.93 (was 1.015)
// - Factor for syllables: 77.0 (was 84.6)
// - This compensates for longer Dutch wordsAim for a Flesch score between 50 and 65 for Dutch web content. For English content, the ideal range is 60 to 70. This provides the best balance between professionalism and machine readability.
How readability influences AI citations
The link between readability and AI citations is more concrete than many people think. When an AI model generates an answer, it selects sources based on multiple factors. One of these is how well it can summarize the source text into a coherent, accurate answer.
A practical example
Compare these two passages about the same topic:
// Version A: low readability (FRE ~25)
"The implementation of structured data markup utilizing the
JSON-LD serialization format facilitates the machinal
interpretation of web page entities by artificially intelligent
systems, resulting in a significantly improved probability of
source attribution in AI-model-generated answer constructions."
// Version B: good readability (FRE ~62)
"Add JSON-LD markup to your web pages. This helps AI models
understand what your page contains. The result: you get cited
more often in AI-generated answers."
// Version B is shorter, clearer and produces better
// AI citations. Same message, better execution.Practical writing tips for better readability
Improving your Flesch score does not mean oversimplifying your content. It is about removing unnecessary complexity while maintaining depth of content.
Sentence length
The biggest factor in readability scores is sentence length. Aim for an average of 15 to 20 words per sentence. This does not mean every sentence should be the same length. Vary with short and longer sentences for a natural rhythm. But avoid sentences of more than 35 words, as these are difficult for both humans and AI to process.
Word choice
Use everyday words where possible. Replace "implement" with "apply", "facilitate" with "enable", "optimize" with "improve" when the context allows. Technical jargon is acceptable when your audience expects it, but explain complex terms on first use.
- Write in active voice: "We analyze your website" instead of "Your website is analyzed by us".
- Avoid double negatives: "It is not impossible" becomes "It is possible".
- Use concrete words: "within three days" instead of "in the short term".
- Split long sentences at conjunctions: replace a comma-and with a period and a new sentence.
- Read your text aloud: if you stumble, the sentence is too complex.
- Use subheadings and lists to break up long blocks of text.
The inverted pyramid
Use the journalistic "inverted pyramid" structure. Start each paragraph with the core message. Then provide the supporting evidence and details. This helps both human readers and AI models quickly find the essence. AI models often prefer texts that present their conclusions early.
A good writing style goes hand in hand with a good heading structure. Descriptive headings combined with readable paragraphs form the ideal foundation for AI citations. Also don't forget to write your image alt texts in the same way: clear, descriptive and not too long.
Measuring and improving readability
Various tools are available to measure the readability of your content. For English texts, tools like Hemingway Editor and Yoast SEO provide real-time feedback. For Dutch texts, the Leesindex tool from Utrecht University is a good option.
Our AEO scanner automatically measures the Flesch Reading Ease score of your homepage content and displays it as part of the AEO score. A score that is too low results in an improvement recommendation with concrete tips to increase readability.
A workflow for readability
- Write your first draft without worrying about readability. Focus on content and completeness.
- Measure the Flesch score of your text with a tool of your choice.
- Identify the longest sentences (more than 25 words) and split them up.
- Replace complex words with simpler alternatives where possible.
- Add subheadings after every 2 to 3 paragraphs for structure.
- Measure again. Repeat until you are in the desired range (50 to 65 for NL, 60 to 70 for EN).
- Have someone else read your text as a final check.
The balance between readability and expertise
A common concern is that readable content looks less professional. This is a misconception. The best experts are precisely those who can explain complex topics clearly. Albert Einstein put it well: "If you cannot explain it simply, you do not understand it well enough." The same logic applies to AI: if a language model cannot summarize your text well, the text is probably unnecessarily complex.
This principle connects directly to E-E-A-T optimization: true expertise is demonstrated not by writing in a difficult manner, but by making complex material understandable. AI models recognize this distinction and prefer sources that combine authority with clarity.
Key takeaways
- Readability directly determines how well AI models can process, summarize and cite your content in their answers.
- The Flesch Reading Ease score (and the Flesch-Douma variant for Dutch) is the most widely used metric. Aim for 60 to 70 (EN) or 50 to 65 (NL).
- Shorten sentences, use everyday words and write in active voice. This improves your score without losing content depth.
- Readability, heading structure and E-E-A-T signals work together: improve them as a whole for the best AEO results.
- Measure your readability structurally with tools and build it in as a fixed step in your content publication process.
Frequently asked questions
Can content be too readable for AI?
Yes, extremely simple texts (FRE above 80) often lack the nuance and depth that AI models look for in quality answers. A text at elementary school level typically does not contain enough information to serve as a reliable source. The sweet spot for professional content lies between 50 and 70.
Do I need to rewrite all my existing content?
Not necessarily. Start with your most important pages: the homepage, your most visited blog posts and your product or service pages. First measure the current scores and focus on pages that score below 40. Small adjustments (splitting sentences, replacing jargon) can already make a big difference.
How do Flesch scores relate to Google's Helpful Content criteria?
There is strong overlap. Google's Helpful Content Update rewards content written for people, not for search engines. Readable, well-structured texts typically score well on both criteria. A high Flesch score is no guarantee for a good ranking, but a low score is often a red flag.
Does readability matter equally for all AI models?
The exact weighting differs per model, but the principle is universal. ChatGPT, Claude, Perplexity and Gemini all benefit from readable source texts. The effect is strongest for summarization tasks: the more readable the source, the more accurate the summary.
Are there industry-specific readability guidelines?
Yes. For medical content (patient information), an FRE of 70 or higher is recommended. For legal content aimed at consumers, aim for at least 50. For technical documentation aimed at professionals, 40 to 55 is acceptable. The key is knowing your audience and adjusting the score accordingly.
Good writing is editing. Every sentence you make shorter makes your content more readable for both humans and machines.
How does your website score on AI readiness?
Get your AEO score within 30 seconds and discover what you can improve.
