A chatbot optimized to generate plausible-looking text is a very good fit for these known weak spots in human judgment, it's effectively punching below the waist all the time. And when something (or anyone - including humans!) systematically spews high-quality bullshit at you, that is effectively gaslighting, which is harmful to your perception of reality and mental health.

i personally started having different dreams after using gpt and stable diffusion for few weeks

i suspect this will be studied in the future