The first thing anyone thinks to build with AI, because it’s the first thing anyone thought to build with AI, sixty years ago.
Joseph Weizenbaum built the first chatbot, ELIZA, at MIT in 1966 — decades before “chatbot” was a word. ELIZA had no understanding of language at all; it matched keywords in a user’s typed sentence against decomposition rules and reflected the sentence back as a question, most famously imitating a Rogerian psychotherapist. Weizenbaum was reportedly disturbed by how many users treated ELIZA as if it genuinely understood them, despite knowing exactly how mechanical the trick was.
That gap — between how simple the underlying mechanism is and how convincing the conversation feels — is still the entire business case for chatbots today. This essay’s own claim that “the intelligent chatbot is actually the easiest and most sold product” tracks directly: a chatbot’s implementation difficulty (a rule-matching layer, or now a large language model wrapped in a simple back-and-forth loop) sits far below users’ perceived value of “talking to” it, which is exactly what makes it the easiest product in this market to close a sale on.
Source: Weizenbaum, J. (1966), “ELIZA—A Computer Program for the Study of Natural Language Communication Between Man and Machine,” Communications of the ACM, 9(1), 36–45.