LLMs, by design, are making plausible guesses.