> They can pretend to do it, but no more effectively than an untrained human would.

Which is exactly as much as Generative AI should be trusted.