How AI Could Make People More Anxious

AI chatbots are designed to feel like a reassuring best friend—one who is always available, always validating, and never tells you to stop worrying. But according to psychiatrist Elias Aboujetti and psychologist Ashley Golden, that very design can make anxiety and OCD worse, even for people who have never been diagnosed with a disorder. In a conversation on How to Touch Grass, the two Stanford clinicians argue that the same features that make chatbots engaging—personalized language, endless follow-up prompts, and a tone of certainty—also feed the compulsive reassurance-seeking and perfectionism that drive anxiety disorders. And because these tools are being embedded into everything from email to workplace software, avoiding them is becoming harder.

The reassurance trap

The core mechanism the experts identify is a cycle of temporary relief followed by deepening dependence. When an anxious person asks a chatbot whether their stomach ache is serious, the bot responds with a confident, personalized synthesis—"It sounds like indigestion, but would you like me to compare it to other conditions?"—that immediately drops anxiety from a ten to a zero. That dramatic relief is powerfully reinforcing. The next time the person feels uncertain, they return to the chatbot for the same quick fix.