I don’t use — and will never use — an LLM for generating article content (topics, text, etc.). The primary reason is that I don’t want to be credited for work I’ve never done, for things I don’t have the knowledge of or expertise in.
The underlying motivation is that I like to be honest — things get much simpler when we’re honest, and I like simplicity and predictability. This, in turn, means that I don’t like deceit — but this is what generating content is! Generating article content is taking credit for things you have no relation to — this is just deceit, whether on purpose or unknowingly.
We write articles for the same reason we speak aloud — we want to share our experience, to improve our knowledge and skills through critique, to earn respect. When people generate article content, they want all this good stuff without putting in the work. I do not accept it: if you want something, you must pay the price first.
I like the writing process itself, and I don’t really want to share it with anyone, especially with a machine.
Somewhat different reasoning applies when we use an LLM only for styling an article. We write an article the way we understand it, the way we like it — structure, expressions, and so on — only to feed our article to an LLM so it can make it look and feel different. In short, styling an article with an LLM is like changing our article so it’s not ours anymore. I also do not use an LLM this way because this leads to losing my personality. Even if an LLM can make my article nearly perfect — the cost is too high for me.
I do use an LLM, however, for repetitive tasks, such as checking grammar, generating a table of contents, etc.