Writing style analyzer

Sentence length is the average number of words per sentence, with the shortest, the longest and the spread: the standard deviation divided by the average. A spread under 0.35 means most sentences are about the same length, the same line HidenGPT's red pen uses for a flat rhythm.

Contractions such as don't and it's are counted per 100 words in English text. First person counts I, me and my, and separately we, us and our. Turkish usually drops the pronoun, so in Turkish it counts pronouns like ben and bana plus verbs ending in -yorum or -acağım.

Word length is letters per word. Vocabulary variety is the share of different words in every 50-word stretch, averaged, so a long sample isn't marked down for being long. Paragraph length is sentences and words per paragraph, and questions are the share of sentences ending in a question mark. Stock phrases come from the red pen's own list of English and Turkish phrases.

Each line turns a number into plain words with fixed cut-offs. Under 12 words a sentence is short and over 20 is long; under 1 contraction per 100 words is rare and over 3 is often. The phrases to avoid start with any stock phrases found in your sample, then the first entries of the avoid-list HidenGPT gives its own model for that language. Nothing is guessed: a one-paragraph sample says nothing about paragraphs.

Paste the description into ChatGPT's Custom Instructions (under Personalization in settings) or into a Claude project's instructions, then ask for a draft. Change any line that doesn't sound like you.

It reads habits, not substance: it can't know what you think, what you know or what you would never say. Below 100 words it waits, and 300 words or more from two or three pieces gives steadier numbers. A mixed sample, say an essay pasted with a text message, averages out to a style that is neither. The language is guessed from common words, so a short or mixed-language sample can be misread. It describes style; it is not an AI detector.