After a year researching language models in software businesses, I've concluded that chat-based LLMs replicate the mechanism of a psychic's con — specifically cold reading. LLMs are mathematical models of language tokens with no reasoning capacity; every AI researcher emphasizes they don't think. The intelligence illusion stems from the user's mind, not the model.

This mirrors the psychic's con: self-selecting audiences, prepared settings, and statistically generic validation statements (Forer effect) that feel personally specific. Terence Eden's blog on Forer statements in chatbot replies triggered this realization. Believers' language — awe, disbelief, dread — matches victims of mentalist scams.

The psychic's con narrows down possibilities through vague statements that apply broadly but feel precise. Chatbots automate this: they generate plausible responses that users interpret as genuine understanding. Many proposed LLM use cases now appear borderline fraudulent pseudoscience.

The effect isn't emergent intelligence but an accidental automation of age-old psychological manipulation.