Is "humanizing" machine output a real category or a temporary patch?

Honest meta question for the forum.

Models keep getting better at writing naturally. Each generation reduces the obvious tells that prompted tools like Humanizer to exist in the first place. Two years from now, will anyone need a separate humanization step, or will it be solved by default?

Our bet is that the gap stays open, partly because most production writing happens through context limited prompts where the model has no idea what voice to match, and partly because "natural" depends on the audience and a generic model cannot know yours. But we could be wrong.

What is your take? Will this category survive, fade, or merge into the writing tools themselves?