How can you be sure what you're seeing is AI slop, as opposed to, for example, corporate-speak or just documents modeled on their past work wherever, etc?