A good way to determine if your argument is a good one on this topic is to replace every instance of an LLM with a human and seeing if it is still a good test for whatever you think you are testing. Because a great many humans are terrible at logic and argument and yet still think.