Speech, in a human community, is not only a product. It is an act for which someone is answerable. A promise, a prayer, a proof, a diagnosis, a verdict, a love-letter — these can be typed by a machine, and they are not yet those acts until a person means them. The hearer is owed a speaker.
Even the lesser acts of study are like this. To paraphrase is to show that you have understood. To define is to take responsibility for a term. To object is to stand behind a denial. If a generator performs the wording, someone still has to have understood, defined, or objected — or else no one has. Judging the output is itself an act of those arts. A student who cannot tell a well-formed sentence from a broken one, a following from a leap, a persuasion from a fog, cannot use the machine. He can only accept it.
The cost of handing grammar and rhetoric to a generator is not that the page will look worse. It often looks better. The cost is that the student will not have performed the work those arts train: forming a thought so that it can be said, testing whether it follows, and offering it to a hearer. What is skipped is not typing. It is the becoming-able that typing used to force.
A fluent page with no speaker is a new kind of silence. It fills the ear and leaves the community without anyone who can be asked, “Did you mean that?”