And that's really the insidious thing about these tools - if you can't do the work yourself then you can't really verify the LLM output either.