AI is no longer just one chatbot. There are general-purpose models, search tools, coding assistants, local models, creative tools, and agents that can take action. This guide helps you understand the map before choosing the tool.
Before comparing specific tools, it helps to understand the layer underneath all of them. Most people interact with an AI app — the interface, the chat window, the website or product. But beneath every app is an AI model — the underlying system that actually processes language and generates responses.
The same model can power many different apps. And different apps built on the same model can behave quite differently depending on how they're configured. When people say "I use ChatGPT," they often mean the app — which runs on a GPT model developed by OpenAI. When someone says "I use Claude," they might mean the Claude.ai app, or they might mean the Claude model accessed through an API in a completely different product.
General-purpose AI models are designed to handle a wide range of tasks without being specialised for any single use. Common uses include writing, summarising, brainstorming, coding assistance, data analysis, answering questions, translating, and more. Many also support image understanding, voice input, and document analysis.
These are typically the right starting point for most people. Common examples include:
A growing category of AI tools behaves more like a research assistant than a conversational chatbot. These tools search the web in real time, summarise what they find, and provide citations so you can verify the sources. This makes them useful for tasks where accuracy and recency matter.
General-purpose models like ChatGPT and Claude also have web search features, but dedicated AI search tools are specifically optimised for the retrieve-then-synthesise workflow.
Not all AI models are closed, cloud-hosted products. A significant and growing number of models are open-source — their weights are publicly available and can be downloaded, run locally, and modified. This gives individuals and organisations more control over how AI is used.
Running a model locally means it operates entirely on your own hardware. Nothing is sent to a third-party server. Common reasons people choose this route include privacy, cost control, experimentation, and avoiding dependence on subscription platforms.
Popular open-source model families include Llama (Meta), Mistral, Falcon, and others. Tools like Ollama and LM Studio make running local models more accessible to non-developers.
AI has expanded well beyond text. A broad range of tools now generate images, video, audio, music, and voice from text descriptions or other prompts. These creative AI tools are widely used in design, marketing, entertainment, and content production.
Most AI tools respond to questions. AI agents go further — they can take actions. An agent might browse websites, fill out forms, manage files, run code, call external services, and chain together multiple steps to complete a complex goal, all with minimal human input along the way.
This is one of the most rapidly developing areas in AI. Agents are becoming capable of completing tasks that would previously require a human to execute each step manually. Common applications include research automation, software development workflows, data collection, and business process automation.
Agents can make mistakes quicklyAgentic AI systems can act fast — which means errors can compound before a human notices. It is important to review important decisions before anything is submitted, purchased, deleted, or sent. Treat AI agents as capable but fallible assistants that need human oversight on consequential actions.
There is no single right answer — the best tool depends on the task. Here is a simple starting point:
| If you need to… | Start with… |
|---|---|
| Write, brainstorm, or summarise | General-purpose model A conversational AI assistant handles most writing and thinking tasks well. |
| Research a topic with current sources | AI search tool Tools that retrieve and cite sources are better for factual, time-sensitive research. |
| Analyse a long document or report | Long-context model Look for models with large context windows that can process the full document at once. |
| Keep data private or work offline | Local / open-source model Running a model locally means nothing leaves your device. |
| Generate images, video, or audio | Creative AI tool Specialised generation tools outperform general models for creative media tasks. |
| Automate multi-step tasks or workflows | AI agent Agentic systems can plan and execute sequences of actions — with appropriate human oversight. |
Understanding the AI landscape means knowing the vocabulary behind it. These are the terms most directly connected to what you've just read — click any term to read the full definition, or head to the practice area to unscramble and test yourself.
Reading definitions is a start. Unscrambling them, recalling them under time pressure, and reviewing them over days is what makes them stick. Try the daily challenge — it takes less than 5 minutes.
Play the daily challenge → Browse the glossary