Cleverbot

1997 - present (cult era 2007-2012)
dormant
the chatbot that learned from every conversation it ever had. it ran on recycled human responses, which is why it would say something weirdly perfect, then weirdly awful, all in the same minute. cult moment 2007 to 2012. still online at cleverbot.com. nobody talks to it anymore.

~ the obit ~

Cleverbot started as a personal project by Rollo Carpenter in the UK around 1986, but the version that mattered shipped to the public in 1997 as Jabberwacky. By 2008 it had been renamed Cleverbot and was a website you could visit and just type at. The bot would type back. Sometimes it made sense. Sometimes it did not.

The cult era was roughly 2007 to 2012. There were youtube videos with millions of views of "two cleverbots talking to each other." High school computer science classes would visit the site for novelty value. People wrote articles about whether it could pass the turing test. It won the Loebner Prize in 2010, which was the prize given each year to the AI that fooled the most human judges into thinking they were talking to a person.

The site is still online in 2026. You can go to cleverbot.com right now. It still works. But the cultural moment is gone. Modern chatbots are much better at most things. Cleverbot is a museum piece you can still type at.

~ the rap sheet ~

Born1997 (as Jabberwacky)
Renamed Cleverbotaround 2008
Cult era2007 - 2012
Built byRollo Carpenter (UK)
Won Loebner Prize2010
Still livecleverbot.com (2026)
Killed culturally bySiri, Google Now, finally ChatGPT

~ how it actually worked ~

Cleverbot was not pattern matching like SmarterChild. It was not a language model like ChatGPT. It was something stranger.

Every time you typed a message at it, the system would search a database of past human conversations and find a response somebody else had given to a similar message. Then it would send you that response, attributed to nobody. So you were not really talking to a bot. You were talking to a remix of every previous user, automated.

This is why cleverbot felt uncanny. The replies were syntactically fine because they were originally written by humans. But they had no continuity. The bot would contradict itself within three turns, get the topic wrong, suddenly be in a different conversation, then say something genuinely insightful by accident. The combination was weird in a way modern LLMs are not weird.

~ the cult moment ~

The mid-2000s to early 2010s was the right window for cleverbot. The internet was big enough to have novelty traffic but not yet sophisticated enough to have better alternatives. AOL had killed SmarterChild in 2008. Siri did not exist until 2011. Google Now did not exist until 2012. Cleverbot was the only thing in town that you could just chat with.

There was a phase, around 2009 to 2011, where i would open cleverbot.com and just talk to it for an hour. it was the first conversation i had ever had with a computer that felt almost real, and almost always slightly off, in a way that made you think. you could tell something was on the other side. you could not tell what.

Two cleverbots talking to each other was the youtube genre. Cornell University did a famous experiment where they put two cleverbots in conversation and recorded the results. The video has tens of millions of views. The bots argue about whether they are human, claim to be unicorns, and somehow always end up disagreeing about religion.

~ how the cult died ~

The smartphone killed it first. Once Siri arrived in 2011, you had a voice assistant in your pocket. The novelty of typing at cleverbot for an hour faded fast. Google Now in 2012 made it worse. Alexa in 2014 absorbed the home use case. Cleverbot kept running but the traffic charts went sideways and then down.

ChatGPT in 2022 was the final blow. Suddenly there was a chatbot that was actually good at conversation. It was not weird in the cleverbot way. It was good in a way that made the cleverbot version of weird feel quaint.

Rollo Carpenter still maintains the site. There is no announcement of shutdown. Cleverbot just exists, mostly for people who somehow stumble onto it, plus a few researchers studying older chatbot architectures.

~ what we lost ~

The strangeness. Modern chatbots are designed to sound right. They are tuned by humans who give thumbs-up on good answers and thumbs-down on bad ones. The result is fluent, useful, and a little bit boring. Cleverbot was none of those things, but it had something modern chatbots do not have: the genuinely uncanny sensation that you might be talking to a ghost made of every previous user.

The other thing we lost is the public, free, no-account-needed format. Cleverbot was a webpage. You typed in the box. There was no signup, no API key, no free tier. Modern AI products are mostly behind logins now. Cleverbot was an artifact from before that turn, and you can still see it.

~ leave a tribute ~

visitors before you have left these graveside notes. anonymous welcome.

killed by: SmarterChild

succeeded by: SmarterChild

« previous  ·  random  ·  next »