Jump to a section
You don't need any of these words to use AI well. But knowing a handful, like prompt, hallucination, knowledge cutoff and deepfake, makes the news, the settings and the warnings much easier to follow.
The basics
The words that come up in almost every conversation about AI.
- AI (artificial intelligence): computer systems that do things we'd normally think need a human, like writing, recognising pictures or understanding speech.
- AI chatbot or AI assistant: an app you talk to in ordinary language, such as ChatGPT, Gemini, Claude or Copilot.
- Generative AI: AI that makes something new, such as text, pictures, music, voices or video, rather than just sorting or analysing.
- Model: the AI 'engine' inside an app. One app can offer several models, often with different strengths.
- Prompt: whatever you type or say to an AI. A better prompt usually gets a better answer.
- Free plan and paid plan: most AI apps have a free version and one or more paid subscriptions with more features or higher limits.
How to ask an AI chatbot a good question (the four-part trick)
Is AI bad for the environment? An honest look at the energy and water
How it works
The words behind the scenes, kept as simple as possible.
- Large language model (LLM): the kind of AI behind chatbots. OpenAI describes it as learning patterns from large amounts of text, then predicting the most likely next piece of language.
- Training data: the huge collection of text, and sometimes images or audio, a model learns from.
- Token: the small chunks of text an AI reads and writes, often common runs of letters rather than whole words.
- Context window: how much text a model can take in at once. Past that limit, the earliest parts of a long chat effectively drop out of view.
- Pre-training and post-training: first the model learns general patterns from text; then it's coached to follow instructions and be helpful.
- RLHF (reinforcement learning from human feedback): part of that coaching, where people rank the model's answers so it learns which kinds of reply are most helpful.
How AI chatbots actually work, and why it changes how you use them
OpenAI Academy: AI fundamentals IBM: what is ChatGPT? Giles Thomas: what AI chatbots are doing under the hood TDWI: context windows in AI
When it goes wrong
Why a chatbot can sometimes be confidently wrong.
- Hallucination: when an AI states something false as if it were fact. It isn't lying; it's predicting what a good answer would look like, and getting it wrong.
- Knowledge cutoff: the date a model's training data stops. Anything after that, it won't know unless it searches the web.
- Web search (or grounding): when a chatbot looks things up online before answering. It doesn't always choose to, so ask it to for anything recent.
- Bias: when an AI's answers lean one way because of patterns in the data it learned from.
AI chatbots make things up. Here's how to check what they tell you
Features and settings
The words you'll see in an AI app's menus.
- Memory: a feature that lets an assistant remember facts about you between chats. You can usually see and delete what it's kept.
- Custom instructions: standing rules you write once, like 'use UK spelling' or 'keep answers short'.
- Temporary chat (or private or incognito chat): a one-off chat that's kept out of your history.
- Model training setting: the switch that decides whether your chats can be used to improve the company's AI.
- Multimodal: an AI that can handle more than text, such as photos, voice or video.
- Voice mode: talking to an AI out loud, and hearing it reply.
Custom instructions and memory: make AI remember what matters to you
How to turn off AI (or at least quieten it) in Google, WhatsApp, Windows and iPhone
Agents and connections
The newer words, for when AI starts doing things as well as saying things.
- Connectors or connected apps: links that let an assistant use your other apps, like your email or calendar.
- Agent: an AI that can take actions for you, such as booking, sending or buying, rather than just answering.
- Prompt injection: a trick where hidden instructions in an email, web page or document make an AI do something it shouldn't. The UK's National Cyber Security Centre describes it as an input 'designed to make the model behave in an unintended way'.
Connecting AI to your email, calendar and files: do it safely
What is agentic AI? AI agents explained for everyday life
NCSC: AI and cyber security Privacy International: risks in turning chatbots into agents
Fakes and scams
The words you'll see in news stories and safety warnings.
- Deepfake: a fake video, photo or audio clip made with AI to look or sound like a real person.
- Voice cloning: copying someone's voice with AI so it can be made to say anything.
- AI watermark: an invisible mark added to AI-made content. Google's is called SynthID, and Gemini can check for it.
- Content Credentials: an industry standard, from a group called C2PA, for recording where an image or video came from and how it was edited.
How to tell if a photo or video is AI-made, and what to do when you can't
Sources
These sources support the guide. Tool settings, prices and policies change, so dated source boxes show when those details were checked.
- Google Gemini Help: verify AI-generated images, videos and audio
- Advanced Television: Ofcom advises how to identify deepfakes (11 July 2025)
- Columbia Journalism Review, Tow Center: AI chatbots can search the web, so why don't they? (27 January 2026)
- OpenAI Academy: AI fundamentals (10 April 2026)
- IBM: what is ChatGPT? (19 April 2025)
- Giles Thomas: what AI chatbots are actually doing under the hood (29 August 2025)
- TDWI: context windows in AI (May 2026)
- UK National Cyber Security Centre: AI and cyber security, what you need to know (13 February 2024)
- Privacy International: risks in turning AI chatbots into AI agents (18 September 2025)
AI helps with research and drafting. Dan checks the sources and is responsible for what is published.
How we check