Models regress toward the category mean
A language model predicts the most probable next token given everything it has read. Ask it to write about your category and it will return the centre of that category — the average of every competitor page it has seen.
That is genuinely useful for structure and speed. It is actively harmful as a source of positioning, because the average is precisely the thing you are trying not to sound like.
The inputs are the differentiator
The output is only as distinctive as what you put in. A model does not know your founder's actual reason for starting the company, the objection your sales team hears every week, or the specific way you are better and worse than the incumbent.
Those things have to be decided by people and written down. Once they are, a model can carry them across a hundred pages faithfully. That is the real leverage — not generation, but consistent propagation of something that was already sharp.
Use it for volume, not for judgement
Good division of labour: humans decide what is true and what matters, models handle variation, format, and first drafts. Humans review anything that goes out with your name on it.
The teams getting hurt by AI are the ones that inverted this — letting the model decide what to say and reserving human time for editing. That produces a lot of content and no position.