1 essay on this topic.
Persona prompting generates the exact type of hallucination that automated LLM judges reward as 'depth.' Two experiments, blind evaluation, and a fact-check that flipped the finding.