There is a difference between an AI checking a text on the basis of a common LLM and there being a defined set of rules for a customer against which a text is checked. Especially if there is a second customer with different rules, an AI (with one global LLB) is no longer reliable.

(Added from a tester email)