• Basics
AI Chatbot Glossary: 25 Terms Explained
Updated July 9, 2026 · 7 min read
Quick answer
This glossary defines the core AI chatbot terms in plain language: large language model, token, prompt, context window, RAG, agent, hallucination, fine-tuning, multimodal, and more.
Core model terms
These terms describe the technology under a chatbot.
- Large language model (LLM): a neural network trained to predict text, the engine behind a modern chatbot.
- Token: a chunk of text, such as a word or part of a word, that a model reads and writes.
- Context window: the maximum number of tokens a model holds at once in a conversation.
- Parameters: the internal values a model adjusts during training; more can mean more capability.
- Multimodal: a model that handles more than text, such as images, audio, or video.
Behavior and quality terms
- Prompt: the text you send to a chatbot, plus any hidden instructions it adds.
- Hallucination: a confident answer that states wrong information.
- Grounding: connecting a model to source data so answers stay accurate.
- Reasoning model: a model that works through steps before it answers, for harder problems.
- Temperature: a setting that controls how varied a model's output is.
Training and tuning terms
- Pretraining: the first training stage, where a model learns to predict text from a large body of data.
- Fine-tuning: further training on curated examples to shape behavior for a task.
- Alignment: safety and feedback training that steers a model toward useful, safe answers.
- Open weights: model files a provider releases so others can run or tune the model.
- Distillation: training a smaller model to copy a larger one for lower cost.
Application terms
- RAG (retrieval-augmented generation): a method where a chatbot pulls from documents and answers from that content with citations.
- Agent: a chatbot that takes actions across steps and tools, not just replies.
- Knowledge base: the approved content a business chatbot answers from.
- Handoff: the moment a chatbot passes a conversation to a human agent.
- Answer engine: a search-first chatbot that returns cited answers, such as Perplexity.
- System prompt: hidden instructions that set a chatbot's role and rules.
- API: an interface that lets software send prompts to a model and receive responses.
- Fine print on privacy: the data policy that states whether a tool trains on your inputs.
- SSO (single sign-on): a login control common in enterprise chatbot plans.
- Deflection: a support metric for tickets a chatbot resolves without an agent.