Content Creation, Readability, and Weasel Words.

Categories:

Rudolf Flesch was an Austrian immigrant who moved to America in 1938 to escape Nazi persecution. He was a committed academic and a graduate of the University of Vienna. He obtained his PhD in library science from Columbia University in New York. In his 1943 dissertation, “Marks of a Readable Style,” he proposed a formula to estimate adult reading materials’ difficulty. He later simplified the formula based on the average sentence and word lengths. His first book, The Art of Plain Talk (1946), became a best-selling book, and Flesch introduced his adult readability formula to the public. Known as “Flesch’s Reading Ease formula,” it is still used today:

RE = 206.835 – (1.015 x Total Words / Total Sentences) – (84.6 x Total Syllables / Total Words)

It works by assigning a number from 0 to 100 to a text. The lower the score, the more complex the sentences. The later Flesch-Kincaid readability grade is similar but uses a different ascending scale.

Score	Difficulty (Education Level)	Approximate CEFR level	Average words	Average syllables
0-29	Academic (Postgraduate level)	C2+	29 or more	1.92 or more
30-50	Very difficult (Higher Education)	C2	25	1.67
50 – 60	Fairly difficult (10th or 12th grade)	C1	21	1.55
60 – 70	Standard (8^th or 9^th grade))	B2	17	1.47
70 – 80	Fairly easy (7^th grade)	B1	14	1.39
80 – 90	Easy to read (6^th grade)	A2	11	1.31
90 – 100	Very easy to read (5^th grade)	A1	8 or fewer	1.23 or less

In the table, I have roughly mapped the CEFR to readability, which is plausible but problematic. The reason for this is that the two scales are very different animals. Flesch Reading Ease is based purely on sentence length and word syllable count. It’s a mechanical formula that doesn’t care about meaning or context. The CEFR (Common European Framework of Reference for Languages) measures a learner’s overall language ability, not just reading. It includes grammar, vocabulary range, listening, speaking, interaction, and writing. The descriptors are qualitative, based on can-do statements, not formulas.

Another critical point is that Flesch’s readability formula emphasizes the “product” over the writing process, limiting its usefulness for learners. Formulas are also not absolute and cannot assess all factors influencing readability. Therefore, we should regard such formulas as only rough indicators of readability. They cannot make definitive assessments of individual work.

However, Flesch’s readability has a lot of very good applications. It is very valuable in content creation and SEO (Search Engine Optimization). Readability is an essential consideration when creating language training materials. If your text is too complex for the student’s language level, it will create a lot of unnecessary cognitive load which will demotivate the learner.

One way to check the level of a text is to look at the distribution of common words in the text. This will give us a rough yardstick on which to judge the material. One of the best tools is the Oxford Text Checker:

For example, we can see that the words in this text are standard A1 terms with a smattering of A2 vocabulary. We can easily grade the text for an A2 student using these results. This is my primary go-to tool for content writing for language learners.

Using a Flesch readability calculator, I also found that the text’s readability is 79.2, suitable for the 7^th grade (the average words per sentence is 9.1, and the average syllables per word is 1.4). The Flesch readability data cannot evaluate the complexity in terms of the types of clauses, etc. However, it has given us a rough estimate for this text and validated our CEFR mapping as A2. This shows that decreases in readability also come with increases in grammatical complexity.

Flesch supported the “plain English” movement, promoting simple, user-friendly writing. He was against overly complicated, academic-type, unnecessarily complex, and wordy writing. The plain English campaign is still active in modern times, promoting the same goal: access to clear and concise information. This goal has also been adapted in the corporate world, where companies evaluate the readability and clarity of their workers’ written communications. They teach staff to avoid complex language and unnecessary words. The goal is clear, concise, and compelling communication. To achieve this, the writer must avoid weasel words and other undesirable expressions like vague adjectives, unnecessary or vague adverbs, modal verbs, auxiliary verbs, empty, meaningless expressions, and vague numerical expressions.

Jumping forward to 2025, one of the most common problems with AI-generated writing from models like ChatGPT is that it tends to default to longer, more complex sentence structures. This occurs even with simple topics that do not merit this treatment. This happens because the models are trained with “smart-sounding” stuff. This tends to favor more formal, structured, and dense writing. So, the model learns that longer sentences, embedded clauses, and abstract vocabulary are “normal,” even when they are not optimal for readability. Style is prioritized over function, and many longer and complex sentences have a high density of multisyllabic words. Another factor is that most users forget to explicitly tell the Large Language Model (LLM) whether it is writing for a ten-year-old child or a business professional. The result is output that is not matched to the reader. Learners or readers with limited attention span, fluency, or time may misunderstand key information. It also creates the illusion of complexity or credibility at the cost of clear communication.

One good tool to evaluate AI-generated text is a Flesch readability score calculator. This will flag an unreasonably complex text. If you are writing for an A2 learner, and you get the following, you should be very worried:

The Flesch readability calculator immediately flags this ChatGPT text as highly complex and unsuitable for any language learner.

Evaluating readability from different perspectives, such as vocabulary distribution, grammatical complexity, and the Flesch readability scores, is essential.

Oxford Text Checker: https://www.oxfordlearnersdictionaries.com/text-checker