When faced with a blank page some percentage of people will not copy a known solution, they'll draw on their experiences and eg., write a book about them.
For example, I could write a book about software development without research only inventorying my opinions, arguing for them, illustrating them, and so on.
If I were to start with ChatGPT I would, necessarily, only say what has been said.
There are so many dangers with the Eliza-effect, and this is one of them. I think the narrative on how these systems work needs more strongly wrestled from their salesmen -- and the real downsides exposed.
A client of mine seemed quite insistent that ChatGPT was "creative". I hope in a few years that opinion will be rare, as everyone ends up writing the same article over-and-over.
And likewise, in development, using earlier versions of libraries with the most "stackoverflow training data"; using dying languages with the most github repos; and so on.
In other words: what we have written the most about is not what we now need to write the most about. And the former is the training data for modern AI.
It is not in the world with us, with our concerns: it is a statistical average over "done and dead" cases.