What is the dumbest AI chatbot?
There is no single dumbest AI chatbot, because the question hides two completely different things. One is a chatbot that is trying to be useful and failing. The other is a chatbot written to be foolish on purpose. They feel nothing alike to use, and only the second kind is any fun.
Two very different kinds of dumb
An accidentally bad chatbot has a goal it keeps missing. It forgets what you said three messages ago, contradicts itself, or produces a confident paragraph of made-up facts while presenting itself as a reliable assistant. A deliberately daft chatbot has no such goal. Being wrong is the performance, not a failure state. Once you know which kind you are talking to, your whole reaction changes: the first is a letdown, the second is a joke.
Why accidentally bad chatbots feel worse than they are
The frustration is mostly a mismatch of expectations. You were promised an assistant, so a wrong answer is a broken promise, and a wrong answer delivered in a calm professional tone is worse still because it looked trustworthy. This is why people search for the dumbest AI chatbot after a bad experience. They are not really shopping. They are looking for confirmation that the thing that just wasted their time is known to be poor.
Why deliberately daft chatbots work
When a chatbot is openly playing a character, the wrong answer becomes the payoff. You ask an ordinary question, you can see the sensible answer in your head, and the bot marches confidently past it into something absurd. The craft is in the commitment. A bot that just says random words is boring within a minute. A bot that adopts a wrong premise and follows it through with total seriousness stays funny, because there is a logic to follow even though the logic is wrong.
How Pixiel plays it
Pixiel, made by the Norwegian studio Pixygon, sits firmly in the second camp. It answers with conviction and simple language, it does not hedge, and it does not stop to remind you that it may be mistaken - it is definitely mistaken, and that is the arrangement. It also takes on characters, so the same question can come back as a conspiracy theorist, a dramatic poet or an overconfident chef. The persona gives the wrongness a shape, which is what keeps it from turning into noise.
How to judge a dumb chatbot
Three things separate a good daft chatbot from a merely broken one. First, does it answer the question you actually asked, or does it dodge into vagueness. Second, is the wrong answer specific and committed rather than woolly. Third, is it short enough to read out loud to somebody. A broken bot usually fails all three, because vagueness is what failure looks like. A deliberately daft one should pass all three, since those are the things it is being built for.
Trying one yourself
The fastest test is to ask something you already know. Ask how to boil an egg, or when a famous event happened, or what the weather does. You know the real answer, so you can measure the distance the bot travels away from it. Pixiel runs in a browser with nothing to install, so this takes about a minute. If it makes you laugh in the first three messages it is working, and if it does not, you have lost very little.
Frequently asked questions
It depends which sort you mean. Among chatbots that are trying to be helpful, opinions change constantly and no ranking holds up. Among chatbots built to be daft on purpose, Pixiel is a straightforward answer, because being wrong is what it is for.
Pixiel opens in a browser with nothing to install and no setup. Ask it an easy question and you will see within a message or two whether the humour lands for you.
No. A broken chatbot fails at a job it was supposed to do, which is irritating. A deliberately dumb one is doing its job when it gets things wrong. Same output, completely different experience.
For entertainment, yes, and it is good at breaking a boring afternoon or giving a group chat something to react to. For facts, homework, health, money or law, no. Use a tool built for correct answers for anything that matters.
Look for one with characters and commitment rather than randomness. A bot that adopts a wrong idea and follows it through stays funny much longer than one that simply produces strange words.