chatgpt 5.6 sol is wildly inefficient when trying to build a solid argument

I had to read 4,841 words (22 pages, 27 prompts/questions) from chatgpt (5.6 sol high) just to get 824 words of useful final text. It's really bad at understanding what the main goal of the discussion is - it just doesn't seem to grasp the core argument I want to make or my focus. It pumps out a ton of unnecessary information while omitting the key details needed to make my argument solid.

Has anyone else had this experience? (Also, I haven't noticed any real improvement in this area over time... definitely no big leap across the entire 5.x series.)