AI Glossary
Key AI terms like generative AI, LLM, RAG and prompts, explained simply for beginners. Meanings and related terms in one place.
An umbrella term for AI that creates new content such as text, images, audio and code. AI chatbots like ChatGPT and Gemini are a type of generative AI.
A large-scale AI model trained on vast amounts of text that understands and generates language. It is the brain behind AI chatbots like ChatGPT.
An AI service that answers questions and helps with tasks through conversation. ChatGPT, Gemini, Claude and Perplexity are examples.
The instruction or input you give an AI. How you write the prompt has a big impact on the quality of the output.
The smallest unit an AI uses to process text. Pricing and the maximum input length (context) are both measured in tokens.
The maximum number of input + output tokens an AI can consider at once. The larger it is, the longer the documents and conversations it can handle.
When an AI generates content that is factually wrong but sounds plausible. Always verify important information against primary sources.
A method where an AI searches external documents and answers based on their content. It is effective for handling up-to-date information and internal documents, and for reducing hallucinations.
A numeric vector that captures the meaning of text or an image. It powers “vector search,” which finds and classifies items by semantic similarity.
AI that handles not only text but multiple types of information such as images, audio and video. It can describe an image it sees or hold a spoken conversation.
Retraining an existing AI model on additional data to optimize it for a specific use. It is used to adapt models to specialized domains or a particular writing style.
An AI that, given a goal, autonomously plans and uses tools to carry out a series of tasks. It goes a step beyond one-off question answering.
An open standard for securely connecting AI models to external tools and data sources. It underpins the spread of AI agents.
Search that answers a question directly and with citations by searching the web. Perplexity and AI modes are typical examples.
Optimizing so that AI search and chat AIs cite your information correctly. Think of it as SEO for the AI era.
A connection point that lets software work together. By calling AI through an API, you can embed generative AI into your own apps and tools.
An LLM whose model weights are published, so anyone can use, modify and self-host it. Llama, DeepSeek and Qwen are examples.
The instruction that sets an AI's role, assumptions and tone at the very start. It takes priority over the user's input in shaping overall behavior.
An attack that makes an AI misbehave using malicious instructions hidden in external data or web pages. It is a key security concern in the age of AI agents.
The concept of a general-purpose AI that can autonomously handle a wide range of tasks at a human level. Today's generative AI is generally seen as a step along the way.
⚠ Reference values based on public data and editorial research; not a complete count of actual users.