Starting Point
"AI" is a big, technical-sounding word for something most of us have already used without thinking about it. If you have ever asked Siri a question, used autocorrect, or had your email suggest a reply, you have used AI.
What has changed in the last couple of years is a new kind of AI that you can have a full conversation with, in whole sentences, about almost anything. That is what this page is about.
You do not need to understand how a car engine works to drive a car. You do not need to understand how an AI works to use one. But a little bit of knowledge makes you a safer, smarter user.
What an AI Chatbot Actually Is
An AI chatbot, the ones called ChatGPT, Claude, Gemini and Copilot, is a computer program that is good at one thing: guessing what words should come next.
That sounds simple, almost silly. It is not. To get good at guessing the next word, the program has to handle context, grammar, facts, tone, and the shape of an answer. To get that good, these programs have been trained on an enormous amount of text; essentially most of the public internet, millions of books, news articles, conversations and reference works.
The result is something that feels like talking to a knowledgeable person. It is not a person. It does not have feelings. It does not remember you between conversations unless you turn that on. But it has read an incredible amount of writing by people, and it has learned to produce writing that sounds like that.
You will sometimes see AI chatbots called "LLMs", short for "Large Language Models". Same thing. "Large" means trained on a lot of text. "Language Model" means a program that predicts language. You do not need to use the term, but if you see it, now you know.
How It Works, Plainly
Imagine you have read every cookbook ever written. Now someone asks you, "what goes well with lamb?" You do not need to look anything up. Your brain just knows, because you have seen the word "lamb" next to words like "rosemary" and "mint" and "potatoes" thousands of times.
An AI chatbot works a bit like that. It has seen so many sentences that when you start one, it has a good sense of how people usually finish it. When you ask a question, it finishes the answer a word at a time, each word chosen because it is the most likely next word given everything that came before.
Two things follow from this, and they are important:
It is not looking anything up. Most of the time, an AI is writing from what it learned during training, not searching a live database. That is why it can be wrong with total confidence. Some newer versions can search the web, but that is a separate feature.
It works in patterns, not facts. It can tell you the general shape of an answer beautifully, even when the specific details are wrong. That is both its superpower and its weak spot.
See It Guess: The Next-Word Prediction SimulatorOpens in a new windowA Word on What You Type In
Because an AI learns by reading, it is worth knowing one thing about your side of the conversation. With many free AI tools, the things you type and the files you upload can be stored and used to help improve the tool. Some paid and workplace versions promise not to do this, but the free public versions often do.
The practical habit: treat a public AI chatbot like a postcard, not a locked drawer. It is fine for everyday questions and writing help. It is not the place for passwords, other people's private details, or information that is not meant to be shared widely.
A Tour of the Main Tools
There are four you will bump into most often. They all do basically the same core thing, but each has a different flavour.
ChatGPT
General writing, everyday questions, brainstorming. The one most people have heard of; the free version is capable.
Claude
Reading documents, drafting letters, careful explanation, and admitting when it does not know something.
Google Gemini
Quick answers while you are doing something else, use on Android phones, voice chats. Can see what is on your screen.
Microsoft Copilot
Work inside Word, Excel and Outlook, if your workplace already pays for it.
You do not need to pick a favourite. Many people keep two or three open and use whichever suits the moment. The free tiers of all four are strong enough for everyday use, but increasingly, the paid subscriptions provide access to maximum capabilities and often additional security settings and customisations that keep you in control of your data and the quality of your outputs.
What AI Is Good At
AI is genuinely, usefully good at a specific set of tasks. These are the places it will earn its keep for you.
Explaining things in plain English. Dense letters, technical words, legal forms, medical jargon. It can turn those into normal language faster than any human.
First drafts of writing. A complaint letter, a cover letter, a message you have been putting off. You get a decent starting point in seconds, and you adjust from there.
Summarising long things. A meeting transcript, a report, a book chapter, a long email thread. Paste it in and ask for the short version.
Brainstorming. Ten ideas for a fundraiser. Five angles on a problem. Names for a program. It does not get bored and it does not judge your ideas.
Teaching and quizzing. A new job skill, study material, a tricky form. It is patient and it will change reading level when you ask.
Getting unstuck. Broken Wi-Fi, a confusing error message, a locked account. It is a fair help desk.
Where AI Gets Things Wrong
The honest part. AI is confident in a way that can fool you. Knowing where it fails is the single most useful thing on this page.
It makes things up. This is called "hallucinating". The AI writes something that sounds right, is written with total confidence, and is simply not true. Fake book titles, fake court cases, fake phone numbers. If the answer matters, verify it.
It is weak on current events. Most AI chatbots were trained months or even a year ago. They do not know yesterday's news unless they have a live search feature turned on.
It is unreliable on anything counted or measured. Poor at arithmetic without a calculator tool, poor at counting letters in a word, poor at exact dates.
It struggles with local or niche facts. It will be confidently wrong about a particular community organisation, a local place name, or who runs what in town. The more specific and local the question, the less you should trust the answer. For Central Australian people, places, language and history, an AI is often confidently wrong; check with people who know.
It guesses at images it has not actually seen. If you ask about a picture without uploading it, the AI will make up a description. Always attach the actual image.
The rule: AI is a great starting point and a terrible final word.
Free Versus Paid
Nearly every AI tool on this page has a free version and a paid version. The paid versions are usually around 20 to 30 Australian dollars a month.
What the free versions can do. Text chat, writing help, reading letters and documents, brainstorming, translation, basic image generation, some voice chat. For most people, most of the time, this is enough.
What paid adds. Usually the newest and smartest version of the AI, longer conversations, higher limits on file uploads and image generation, priority access when the free tier is busy, and sometimes extra features like video generation or advanced voice.
An honest suggestion. Do not pay for anything on day one. Use the free versions for a while. You will quickly know whether you are hitting the limits often enough to be worth the money, or whether free does everything you need.
Try It Yourself
The best way to understand what an AI chatbot is, is to open one and ask it something.
Pick one. Open it. Ask it three things.
- Something easy you already know the answer to. "What is a good meal to make with mince and a tin of tomatoes?" This lets you see how helpful the AI is, and whether the answer matches common sense.
- Something practical. "I am thinking of getting a dog but I live in a small house. What breeds should I consider?" This lets you see how the AI asks follow-up questions.
- Something you are curious about but might never ask out loud. Your choice. This lets you see that no question is too silly for an AI.
Then ask one more: "What kind of things are you not good at?" Read the answer carefully. That is a glimpse of the honesty the rest of this material builds on.
If the AI asks you a follow-up question, answer it. That is the AI doing its job, gathering context so it can do better.
Remember, as a general rule, and this is the biggest flaw with the technology in 2026: AIs are typically designed to be helpful assistants, and being helpful does not necessarily make them truthful.
Glossary
- AI (artificial intelligence)
- Computer programs that do tasks we used to think needed a person, such as understanding language or recognising images.
- AI chatbot
- A program you talk to in plain sentences, like ChatGPT, Claude, Gemini or Copilot.
- LLM (Large Language Model)
- The technical name for the kind of AI behind a chatbot; a program trained on a lot of text to predict language.
- Training data
- The huge collection of text and images an AI learned from. What is in the training data shapes what the AI can do, and what it gets wrong.
- Hallucination
- When an AI states something false as if it were true. The writing sounds confident; the facts are invented.
- Prompt
- The question or instruction you type into an AI.
- Free tier / paid tier
- The no-cost version of a tool versus the paid subscription, which usually adds speed, limits and newer features.
- Indigenous Cultural and Intellectual Property (ICIP)
- The rights and interests of Aboriginal and Torres Strait Islander peoples in their cultural knowledge, stories, language, art and data. See the resource library for sources.