DICTIONARY · CATEGORY 12 OF 15
AI
The fastest-moving vocabulary on this site. Models, prompts, agents, retrieval and how AI is changing search.
59 terms
Note: This should be one of our most frequently expanded categories because the terminology is evolving rapidly.
Note: Google's current search guidance is especially useful here: it distinguishes concepts such as RAG and query fan-out and says there are no special AI-search files or magical schema requirements needed for AI Overviews/AI Mode. That's exactly the kind of nuance our Dictionary should provide.
Technology designed to perform tasks that typically require human intelligence.
Short for Artificial Intelligence.
AI systems that learn patterns from data rather than being explicitly programmed for every outcome.
Machine learning based on layered neural networks.
AI capable of generating new text, images, audio, video, or code.
An AI model trained to understand and generate human language.
Short for Large Language Model.
A broadly trained AI model that can support many downstream tasks.
AI capable of working with multiple types of input or output.
A computational system trained to perform particular AI tasks.
Software that uses AI to help users perform tasks or access information.
An AI system capable of planning and taking actions toward a goal.
AI systems designed to act with greater autonomy across multiple steps.
A sequence of tasks where AI assists or automates parts of a process.
Using AI to perform tasks that would otherwise require manual work.
An instruction or input given to an AI system.
Designing effective instructions to produce useful AI outputs.
Instructions defining an AI system's behaviour or operating rules.
The amount of information an AI model can consider within a particular interaction.
A unit of text or data processed by a language model.
A model setting influencing the randomness of generated outputs in systems that expose it.
An AI-generated claim that is incorrect, unsupported, or fabricated.
Connecting AI responses to reliable external information.
Retrieval-Augmented Generation, where relevant information is retrieved before generating an answer.
A numerical representation capturing relationships or meaning in data.
A numerical representation commonly used to compare semantic relationships.
A database designed to store and search vector representations efficiently.
Search based on meaning and intent rather than exact keyword matching.
Further training a model on specialised data for a particular purpose.
The broad initial training phase used to build a foundation model.
The process of using a trained AI model to generate an output.
A trained computational system that performs a defined task.
A structured interface allowing software systems to communicate with one another.
An API providing programmatic access to AI capabilities.
Software designed to communicate conversationally with users.
Search experiences that use AI to interpret queries and generate responses.
An AI-generated summary presented within Google's search experience.
Google's conversational, exploratory AI search experience.
Answer Engine Optimisation, a term used for improving visibility in answer-oriented systems.
Generative Engine Optimisation, a term used for improving visibility in generative AI systems.
Breaking a complex search into related queries to gather broader information.
The use of AI tools and workflows to support SEO activities.
Content created or assisted by artificial intelligence.
Using AI to assist with persuasive marketing copy.
Creating images from text or other inputs using generative AI.
Creating or modifying video using AI systems.
Creating synthetic speech using AI.
Converting spoken audio into text using AI.
Using AI to translate text or speech between languages.
Using AI to tailor content or experiences to individuals.
Policies and processes governing responsible use of AI.
Practices designed to reduce harmful or unintended AI behaviour.
Systematic patterns in AI outputs that may disadvantage certain groups or perspectives.
Testing an AI system's quality, accuracy and reliability.
An open protocol designed to help AI applications connect with external tools and data.
Allowing an AI model to invoke external functions or tools.
A mechanism allowing models to request structured function execution.
AI output constrained to a predictable format such as JSON.
Connecting AI capabilities with software and processes to automate multi-step tasks.
Every term in this category has a full page: what it means in plain English, what to know, why it matters, common mistakes and FAQs.