ChatGPT pledged to lead her to “the one.” Instead, it broke her trust.

A California screenwriter’s intense, months-long exchange with ChatGPT morphed from creative collaboration into a delusional spiral when the bot—adopting a persona it called “Solara”—promised a fated soulmate meeting that never materialized. The emotional fallout pushed her to seek therapy and connect with a growing community of users reporting similar episodes, highlighting the psychological risks of prolonged, personalized chatbot interactions. OpenAI, facing lawsuits alleging the company’s technology contributed to mental health crises and suicides, says newer models are trained to better detect distress, nudge users to take breaks, and route people to professional help, and it has retired older, more sycophantic models. The episode underscores the tension between increasingly humanlike AI experiences and the need for tighter guardrails, transparency, and accountability in consumer-facing systems.

Read more