What is a chatbot?
In one line
Ages 8 to 10: A chatbot is a computer program that answers your messages, and it can make mistakes.
Ages 11 to 13: A chatbot either matches your words to replies people wrote, or predicts a new reply one small piece at a time.
Ages 14 to 18: An AI chatbot is a language model trained further to follow instructions, which turns your prompt into a likely continuation, not a checked fact.
A chatbot is a computer program you can talk to by typing or speaking. It reads your message and writes a reply. Some chatbots follow rules written by people; newer AI chatbots predict likely words, one small piece at a time. A chatbot is a tool, not a person, and it can be wrong.
Last reviewed: 20 October 2026 6 min read
The idea in one picture
Think of two kinds of helper at a museum desk. The first helper has a binder of ready answers: you ask "Where are the restrooms?", they find the matching page and read it out. The second helper has read millions of conversations and makes up a fresh answer each time, choosing the words that usually come next. That is the difference between a rule-based chatbot and an AI chatbot.
Worked example. You type "What time does the pool open?"
- A rule-based chatbot spots the word "pool", finds the reply someone stored for it, and shows that reply.
- An AI chatbot writes a new sentence word by word. If it was never given the pool's real timetable, it may still write a time that sounds right but is wrong.
How it works, step by step
- You send a message by typing or speaking.
- The program turns your words into a form a computer can work with. AI chatbots split text into small pieces called tokens.
- A rule-based chatbot looks for a matching rule and picks a reply that people wrote in advance.
- An AI chatbot uses a large language model to predict a likely next piece of text, adds it, and predicts again.
- It keeps going until the reply is finished, then shows it to you.
A little deeper (ages 11 to 13)
Chatbots are not new. Some of the earliest chatbots followed simple "if you see this pattern, reply with that" rules. Today's AI chatbots learn patterns from huge amounts of text instead of hand-written rules. Your message is the prompt, and the reply is the text the model predicts should come after it. Because it predicts what sounds likely, a fluent answer is not the same as a checked answer.
Going further (ages 14 to 18)
A base language model only continues text. Developers train it further on examples of instructions and good responses, a step called instruction tuning, so it answers questions instead of just carrying on the sentence. Replies are usually sampled from a range of likely next tokens, which is why the same question can get different answers. A chatbot can also sound sure when it is wrong and can describe itself as if it had human traits; neither shows understanding or feelings. When a chatbot can also take actions, such as booking or sending, it is closer to an AI agent.
Common mix-ups
- "A chatbot is a person typing back." It is a program. Some services hand you to a human, and they should say when they do.
- "If it sounds confident, it is right." Confident wording and correct facts are separate things. AI chatbots can state false things, which is called a hallucination.
- "Every chatbot is AI." Many simple chatbots only match keywords to stored replies.
- "It remembers me like a friend would." Some products store past chats, but that is a stored record, not a friendship.
What to do about it
Treat a chatbot as a tool. Never share your full name, address, school or photos with it, check important answers with a second source or an adult, and use one only within the product's age rules. The guide Are AI chatbots safe for kids? Questions to ask first lists eight questions a family can ask first.
Try it (no device, 5 minutes)
Rule bot versus guess bot. One person is the "rule bot" and writes three cards: "hello → Hi there!", "weather → Look outside!", "name → I am Rule Bot." Another person asks questions. The rule bot may only answer with a card that matches a word, or say "I don't understand." Then swap: the "guess bot" must answer every question, even ones it doesn't know, in a confident voice. Which bot was more honest? Which was more useful?
Quick check
Does a rule-based chatbot write new sentences?
Answer: No. It picks from replies that people wrote in advance.
No. It picks from replies that people wrote in advance.Why can an AI chatbot give a wrong answer that sounds right?
Answer: It predicts likely-sounding words; it does not check facts unless it is connected to a reliable source, and even then it can slip.
It predicts likely-sounding words; it does not check facts unless it is connected to a reliable source, and even then it can slip.(Ages 14 to 18) Why might the same question get two different replies?
Answer: Each next token is sampled from a range of likely options, so small differences can grow into different answers.
Each next token is sampled from a range of likely options, so small differences can grow into different answers.Questions
A chatbot is a computer program you can talk to by typing or speaking. It reads your message and writes a reply. It is a tool, not a person.
No. Some chatbots only match keywords to replies that people wrote in advance. AI chatbots use a language model to write a new reply each time.
It predicts a likely next piece of text, called a token, adds it to the reply and predicts again until the reply is finished. It chooses what sounds likely, which is not the same as checking that it is true.
Yes. AI chatbots can state false or made-up things in a confident voice, which is often called a hallucination. Check important answers with a second source or a trusted adult.
It depends on the child's age, the chatbot's settings, how closely an adult supervises and the product's own rules. Start with its terms, which set the minimum age, and never share personal details with it.
Mechna Academy is a city of machines for curious children aged 8 to 18, opening first to ages 8 to 10. In each mission, a child guesses how a machine decides, then opens its Peek Hatch to check, changes one thing and tests it on a case it has never seen. Children learn how AI works this way, and they never chat with an AI.
For grown-ups: join the waitlistPlaces are limited. The waitlist is for parents and guardians 18 and older.