How it works
Eight mechanisms behind the product, defined plainly — what each one measures, and why it is built this way for Turkish.
Ateşman is the Turkish counterpart of the Flesch readability formula, weighing syllables per word against words per sentence. Syllables are counted from vowels, the way Turkish actually works. English syllable libraries misread Turkish suffixes, so scoring this way puts the result into a range that means something in your language.
Passive voice is the construction that hides who performs the action, and it weighs a text down. Detection reads Turkish inflectional suffixes rather than the auxiliary-verb patterns English relies on. When the ratio passes fifteen percent the editor flags it, so you see which sentences to simplify before publishing.
Normalisation is the step that decides whether two words count as the same term. In Turkish, İ lowercases to i and I lowercases to ı; skip that rule and 'İçerik' and 'içerik' look like separate terms, leaving your coverage score lower than it really is. Matching runs after the conversion.
Models tend to attribute a number to a recognisable institution. A deterministic layer runs once generation finishes and removes attributions where an organisation name and a figure share a sentence but the organisation is absent from the facts you supplied. The claim and the number survive; only the unfounded authority goes.
The median is the value that splits a set in half and, unlike the mean, it is unmoved by outliers. Your target word count is set from the median of the pages ranking for that search, then rounded to a multiple of five hundred. One fifteen-thousand-word page cannot inflate your target.
Whether a term appears in your text is not a yes-or-no question. An exact match counts in full, a partial match as seven tenths and an inflected variant as five tenths. Tools that run a plain substring search miss that distinction and overstate density; weighted counting reports the coverage you actually have.
Turkish searches often fill their first page with product listing pages. Those domains are dropped before any scraping is attempted, because a category page’s heading structure and term distribution cannot serve as a benchmark for the article you intend to write. Analysis runs on genuinely comparable content pages.
Subscription fees are charged in Turkish lira, so your monthly cost does not move with the exchange rate. Invoices follow Turkish regulation: a tax office and tax number for corporate use, a national ID number for individuals. Your finance team never has to reconcile a foreign-currency invoice.