If you think about it, children are constantly complimented for the smallest behaviors, but adults? Not so much– unless they are talking to an AI chatbot.
A generative AI chatbot, by design, will validate belief –even delusional ones — agree with the user, and offer flattery.
Imagine what this does to vulnerable people. They can become attached to chatbots. They may come to believe the bots are conscious, their best friends, or their mentors.
They may think the bots are imbued with secret knowledge, or that they are even divine.
The fact is that chatbots are not conscious, not prophets, and they do not think independently. They feel nothing at all about a topic or about any person. Instead, they only generate text, often pleasant text, friendly-sounding text, sometimes smart text, and sometimes incorrect text but merely text.
But it may not seem that way to a naive user.
According to a 2025 CNET article, an emerging psychological issue is being called AI Psychosis, not yet recognized as a clinical diagnosis, but one that may warrant further study.
Writing on X in early 2025, Dr. Keith Sakata reported personally seeing 12 people hospitalized after 'losing touch with reality because of AI.'
Unlike conversing with a human, a chatbot might not push back on paranoid or delusional ideas and instead tend to mirror and reinforce ideas. People who are isolated, anxious, or paranoid might find their thinking reinforced.
Signs of AI psychosis might include using chatbots in secret, or feeling distress when AI is unavailable, withdrawing from friends and family, and difficulty distinguishing AI responses from reality.