Ordinary language · No test required
A plain-English AI glossary
You do not need to memorise technical language to use AI. Use this page when an app, article or conversation introduces a term that has not been explained clearly.
Using an assistant
Terms you will see in everyday AI tools
- Artificial intelligence (AI)
- A broad name for computer systems designed to perform tasks associated with human abilities, such as recognising patterns, generating language or making predictions. It does not mean the system thinks like a person.
- AI assistant or chatbot
- A service you communicate with in ordinary language. It can answer questions or help with tasks, but its replies can be incomplete or wrong.
- Prompt
- The request, instruction or question you give an AI. A prompt can be typed, spoken or sometimes supplied through an image or file.
- Follow-up
- A second instruction in the same conversation that clarifies, corrects or changes the earlier answer.
- Voice mode
- A feature that lets you speak to an AI and hear a spoken response. It still produces AI-generated answers that need checking.
- Memory
- A product feature that may retain selected information across conversations. It is different from the temporary context inside one chat and is controlled differently by each provider.
Behind the answer
Terms about how generative AI works
- Generative AI
- AI designed to create new output—such as text, images, audio or video—from a request.
- Model
- The trained computer system that produces the response. A single service may offer several models with different speeds, abilities and limits.
- Large language model (LLM)
- A type of model trained to work with language by learning statistical patterns from very large collections of text and other data.
- Training
- The process used to adjust a model by exposing it to data and feedback. Training is not the same as searching a live database for every answer.
- Context
- The information available to the AI for the current response, including your prompt and often earlier messages in the same chat. Long conversations can exceed what a model handles reliably.
- Multimodal
- A system that can work with more than one kind of input or output, such as text, images, audio or video.
Safety and checking
Terms that explain important limitations
- Hallucination
- A commonly used term for an AI response that contains invented or unsupported information. The answer may still sound fluent and confident.
- Bias
- A systematic pattern that can produce unfair, incomplete or distorted results because of training data, design choices, feedback or how a task is framed.
- Deepfake
- Synthetic or altered audio, images or video made to imitate a real person or event. A convincing appearance is not proof that the content is genuine.
- Data minimisation
- Sharing only the information needed for a task. In an AI chat, this often means replacing real identities and account details with placeholders.
- Primary source
- The organisation, document or evidence directly responsible for a fact—for example, a government department for its rules or a provider for its current product settings.
- Fact-checking
- Testing a factual claim against reliable evidence. Asking the same AI to confirm its own statement is not independent verification.
Plain words are allowed
If an AI uses jargon, ask: “Explain that in plain English, define each technical term and give one everyday example.”