AI glossary · 66 terms
The vocabulary behind
the pricing page.
Each term has a one-paragraph definition you can quote, a note on why it matters when choosing a tool, and links to the categories and tools where it shows up.
A
Agent memory
Mechanisms that let an assistant or agent remember facts, preferences, and history across sessions and tasks.
Read definition →Agentic workflow
A process where one or more AI agents carry out multi-step tasks, often with human checkpoints, inside a defined pipeline.
Read definition →AI agent
A system that uses a model to plan and take actions with tools over multiple steps toward a goal, not just answer once.
Read definition →AI content detection
Tools that estimate whether text or media was produced by AI, which are probabilistic and prone to false positives.
Read definition →AI crawler
A bot operated by an AI company that fetches web pages for training data, search indexing, or live answer retrieval.
Read definition →AI search
Search experiences that use models to understand queries semantically and generate direct answers with sources.
Read definition →AI upscaling
Increasing an image or video's resolution using models that infer plausible detail rather than simple interpolation.
Read definition →Answer engine
A search product that returns a synthesised answer with citations instead of a list of links, such as Perplexity or AI Overviews.
Read definition →Answer engine optimization (AEO)
Structuring content to answer specific questions directly so search engines and assistants can surface it as the answer.
Read definition →API access
Programmatic access to a model or tool so developers can integrate its capabilities into their own products and workflows.
Read definition →C
Chain-of-thought
Prompting or training a model to work through intermediate steps explicitly, which often improves reasoning accuracy.
Read definition →Citations in AI answers
Links or references an AI tool attaches to its statements so users can verify claims against the original source.
Read definition →Commercial rights for AI outputs
Whether and under what conditions you may use generated content in products, marketing, or sales, as set by the vendor's terms.
Read definition →Context window
The maximum amount of text, measured in tokens, that a model can consider at once, including your prompt, documents, and its reply.
Read definition →Copilot
An AI assistant embedded in a specific application to help with tasks in context, such as coding, writing, or spreadsheets.
Read definition →D
Data residency and privacy
Where and how an AI vendor stores and processes your data, whether it trains on it, and which regional or contractual guarantees apply.
Read definition →Diffusion model
A generative model that creates images, audio, or video by learning to reverse a gradual noising process.
Read definition →E
Embedding
A numerical vector that represents the meaning of text, images, or other data so similar items sit close together.
Read definition →Enterprise search
Search across a company's internal applications and documents, now often paired with AI answers that respect permissions.
Read definition →Evaluation (evals)
Systematic testing of an AI system's outputs against expected results or quality criteria to measure and track performance.
Read definition →G
Generative AI
AI systems that create new content such as text, images, audio, video, or code from prompts rather than only classifying existing data.
Read definition →Generative engine optimization (GEO)
Shaping content so AI answer engines and assistants cite it accurately, extending SEO to generative search.
Read definition →Grounding
Anchoring a model's answers in specific, verifiable sources such as your documents, a search index, or live data.
Read definition →Guardrails
Controls that constrain what an AI system will do or say, such as topic limits, content filters, action permissions, and approvals.
Read definition →I
Inference
Running a trained model to produce outputs, as opposed to training it; the compute that powers every response.
Read definition →Inpainting and outpainting
Editing part of an image by regenerating a selected region, or extending an image beyond its original borders.
Read definition →IP indemnification
A vendor's promise to defend and cover customers against intellectual property claims arising from generated outputs.
Read definition →L
Large language model (LLM)
A neural network trained on vast text corpora that predicts and generates language, powering chat assistants and writing tools.
Read definition →Latency
The delay between sending a request to an AI system and receiving its response, critical for voice and real-time uses.
Read definition →llms.txt
A proposed plain-text file at a site's root that summarises the site and links key pages for AI assistants and crawlers.
Read definition →LoRA (low-rank adaptation)
An efficient fine-tuning method that trains small adapter weights instead of the whole model, popular for custom image styles.
Read definition →M
Model card
A document describing a model's intended uses, training data summary, evaluation results, limitations, and safety considerations.
Read definition →Model Context Protocol (MCP)
An open standard for connecting AI assistants and agents to tools and data sources through a common interface.
Read definition →Multimodal AI
Models that understand or generate more than one kind of input or output, such as text, images, audio, and video.
Read definition →P
Per-seat pricing
Charging a fixed monthly fee per user, common for assistants and copilots embedded in daily tools.
Read definition →Prompt
The instruction, question, or context you give an AI model to produce a response.
Read definition →Prompt engineering
The practice of designing and refining prompts, examples, and instructions to get reliable outputs from AI models.
Read definition →R
Rate limit
A cap on how many requests or tokens you can send to an AI service in a period, enforced to protect capacity.
Read definition →Reasoning model
A model trained to spend extra computation thinking through steps before answering, improving accuracy on complex problems.
Read definition →Retrieval-augmented generation (RAG)
A technique where a system fetches relevant documents and gives them to the model so answers are grounded in your data.
Read definition →S
SDK
A software development kit that packages an API into libraries, types, and helpers for a programming language.
Read definition →Self-hosting
Running AI software and models on infrastructure you control rather than using a vendor's hosted service.
Read definition →Speech-to-text (transcription)
Converting spoken audio into written text, the foundation of meeting assistants and captioning tools.
Read definition →Streaming responses
Delivering a model's output progressively as it is generated instead of waiting for the full response.
Read definition →Structured data (schema.org)
Machine-readable markup, usually JSON-LD, that describes a page's entities such as products, reviews, FAQs, and organisations.
Read definition →Synthetic data
Artificially generated data used to train or test models when real data is scarce, sensitive, or expensive to label.
Read definition →System prompt
Hidden instructions set by a product or developer that define an assistant's role, rules, tone, and constraints.
Read definition →T
Temperature
A setting that controls randomness in generation; low values give consistent answers, high values give more varied ones.
Read definition →Text-to-image
Generating images from written prompts using models such as diffusion or transformer-based image generators.
Read definition →Text-to-speech (TTS)
Synthesising natural spoken audio from written text, often with selectable voices, emotions, and languages.
Read definition →Text-to-video
Generating short video clips from text or image prompts, with controls for motion, camera, and keyframes.
Read definition →Token
A unit of text, roughly three-quarters of an English word, that models read and generate and that many providers bill by.
Read definition →Tool use (function calling)
A model's ability to call external functions, APIs, or tools by producing structured requests that software executes.
Read definition →Training data
The text, images, audio, or code used to teach a model, which shapes its capabilities, biases, and legal exposure.
Read definition →Training opt-out
A setting or contract term that prevents a vendor from using your prompts, files, or outputs to train its models.
Read definition →V
W
Watermarking and content credentials
Techniques that embed signals or attach metadata to AI-generated media so its origin can be verified.
Read definition →Workflow automation
Connecting apps and steps so tasks run automatically on triggers, increasingly with AI steps that read, decide, and write.
Read definition →