Readability scores count syllables and punish every German compound word, including ordinary ones. This check measures what actually decides whether a page can be read and navigated: is there one h1, does the heading tree skip levels or contain empty headings, is the main content marked as such, how long are the sentences and paragraphs, which abbreviations appear in the introduction without ever being explained in brackets, do images carry alt text, and how much of the HTML is text at all. Sentences are measured strictly inside paragraphs, with code removed first, because a configuration file counted as prose produces a 140-word sentence that never existed.
One request of at most 1.5 MB goes from this server to the address entered; nothing on the page is executed. Hosts in private, loopback and link-local networks are refused. The query is protected by reCAPTCHA v3; the calling IP address and the target are stored for one hour to limit the rate.
Limits worth knowing: only server-rendered HTML is measured, content that JavaScript inserts later is invisible; the main content is taken from main, article or a role=main element, otherwise from the body without navigation, header, footer and aside; sentence splitting is a heuristic that keeps single-letter abbreviations and day numbers together but has no dictionary; the abbreviation check accepts a bracket explanation anywhere on the page and skips a list of everyday acronyms; and there is no syllable count on purpose.