What AI chatbots are (and aren't)
A plain picture of how chatbots work, so you know when to trust them.
An AI chatbot, like ChatGPT, Claude or Gemini, is a program you talk to in plain language. You type a question or a request, and it writes an answer.
How it works, in one paragraph
These tools learned from a huge amount of text. From that, they got very good at predicting which words should come next. That's why they can explain, summarise, translate and draft almost anything in seconds. It is also why they can sound confident while being wrong: they produce likely text, not checked facts.
What they're great at
- Explaining a topic at your level, as many times as you need.
- Drafting: emails, posts, summaries, plans. You edit; it saves you the blank page.
- Rewording: shorter, friendlier, more formal, in another language.
- Brainstorming: ten ideas in ten seconds, so you can pick the good ones.
What they're not
- Not a search engine. Some can search the web, but by default they answer from what they learned, which may be out of date.
- Not always right. They can invent facts, quotes and sources. You'll learn to spot this in lesson 4.
- Not a person. They don't know you, remember you between chats (unless a memory feature is on) or have opinions of their own.
The right mindset
Treat a chatbot like a fast, well-read assistant on their first day: brilliant at drafts and explanations, but you check anything important before you rely on it.
Key takeaways
- Chatbots predict likely text; that's why they're fast and sometimes confidently wrong.
- Use them to explain, draft, reword and brainstorm.
- Check anything important before relying on it.
Try it now
Open a chatbot and ask: "Explain what you are and what you're not good at, in five short bullet points, for someone who has never used AI." Read the answer and compare it with this lesson.
Open ChatGPTFree · paid from $20/moYou've got it when: You got five bullets, and at least one of them mentions that it can make mistakes.