Plausibility, persuasion, and truth: why language models may appear designed to deceive