You are describing false equivalency. There is a fundamental flaw to LLMs. The data is not weighted. Thats why it cant be reliably directed.