AI agent
A software system that uses a model, instructions, context and tools to pursue a defined task through one or more steps.
Reference library
Plain-language definitions for concepts used across Zactra’s AI, product engineering and search-optimization work.
A software system that uses a model, instructions, context and tools to pursue a defined task through one or more steps.
A conversational application that interprets user messages and returns information, guidance or actions through natural-language interaction.
The practice of improving content and technical signals so answer systems can discover, understand and use a source when responding to a question. It builds on SEO rather than replacing it.
Software techniques that perform tasks commonly associated with human intelligence, such as language understanding, prediction, perception, planning and decision support.
The preferred URL for a page when the same or substantially similar content can be reached through multiple addresses.
Google metrics for loading performance, interaction responsiveness and visual stability, commonly represented by LCP, INP and CLS.
Numeric representations used to compare semantic similarity between text, images or other data. They are commonly used in vector search and RAG systems.
Improving the consistency and clarity of people, organizations, products, services and relationships across a website and trusted external sources.
Additional model training on examples intended to adapt behavior or task performance. It is different from retrieving current facts at request time.
A common industry term for improving visibility in generative search and AI-generated answers through useful content, clear entities, crawlability, authority and measurement.
A design in which people review, approve, correct or supervise model outputs or automated actions at appropriate points.
A protocol that lets a website notify participating search engines when URLs are added, changed or removed.
A machine-learning model trained to process and generate language and often other modalities. Applications add prompts, retrieval, tools, permissions and evaluation around the model.
A repeatable process for testing an AI system against representative examples and metrics such as task success, groundedness, safety, latency and cost.
An architecture that retrieves relevant information from approved sources and supplies it to a generative model to produce a more grounded response.
A shared vocabulary for structured data used to describe organizations, people, services, articles, jobs, breadcrumbs and many other entities on the web.
The underlying task or question a person is trying to complete with a search, such as learning, comparing, finding a company or requesting a service.
Machine-readable markup that describes entities and page content using a vocabulary such as Schema.org. It should match information visible to users.
A search-marketing concept describing the breadth, depth, quality and interconnection of useful content around a subject. It is not a formal score published by a search engine.
A database or search index optimized to store and retrieve embedding vectors, often alongside metadata and access controls.