LLMs and chatbots can't truly "understand" language, they just pattern-match and generate convincing-sounding text, which is a fundamentally different thing.