ChatGPT, Gemini, Llama3 .. oh wait, you asked "How to .. ?" not "What tool .. ?"
The process I'm using it's something like this (I automate with Python but can be done relatively easiy with Excel or even Notepad, manually):
1. Generate ideas/keywords. Save them somewhere
2. For each idea, create an outline (a set of unique headlines) for the content. Save again
3. For each headline, generate 2-3 paragraphs. Instruct the LLM to not use information reserved for other headlines. Save the output
4. Stitch the pieces together. If the final result is not exceeding context window of LLM , send it for proofreading. If longer, proofread manually.
5. Delete final text, start over. Or maybe, if you really like the result, better save it, I don't know, maybe even use it lol