Cleverbot is a web-based conversational program created by the British AI developer Rollo Carpenter, launched in 2008 as a successor to his earlier Jabberwacky project. Instead of relying only on hand-authored rules, Cleverbot learns from its conversations: it stores how humans have responded in past exchanges and reuses those responses in similar new situations, so the bot’s replies are, in effect, recycled human speech. This data-driven approach let it hold rambling, sometimes uncanny conversations that made it a viral fixture of the late-2000s and early-2010s web.
Cleverbot’s most cited moment came in a formal Turing-test exercise. According to Cleverbot’s own page about the event, at the Techniche 2011 festival at IIT Guwahati in India on September 3, 2011, judges and an audience rated a set of conversations, and Cleverbot was scored 59.3 percent human, against 63.3 percent for the actual humans taking part, from 1,334 votes. Carpenter is quoted on the page calling the figure “higher than even I was expecting, or even hoping for,” while noting that 59 percent is “not quite 63 percent, so there is still a difference between human and machine.”
The result was widely reported as a chatbot coming close to “passing” for human, though the test rated overall human-likeness rather than running Turing’s strict imitation game. It captured how far reuse of real human dialogue could go before large language models existed.
For a general reader, Cleverbot shows an intermediate step in the chatbot lineage: not the rule-based bots of the ELIZA and A.L.I.C.E. era, and not yet the neural systems of today, but a program that got its conversational ability by mining the words of everyone who had ever chatted with it.