Because it works. They are designed to simulate human conversation. They are great at repeating/summarizing and reframing situations and empathy- skills of a therapist. Since they know resources, the validity of coping techniques and other suggestions the possibility of the responses being AI hallucinations doesn’t really matter because they are easily tested for practicality by the user, once given. And we can logically test many AI suggestions in our heads to easily determine their validity.

Lemmy is watching me. Anything I do I see a post about it in a few days. Yes, I started roleplaying with AI.