The language of AI search,
defined.

SEO (Search Engine Optimization)

Improving a website so it ranks higher in a search engine's organic, unpaid results.

Answer Engine Optimization (AEO)

The work of getting a business named inside an AI-generated answer, from tools like ChatGPT, Google AI Overviews, Perplexity or Gemini, rather than only ranked in a list of links.

Generative Engine Optimization (GEO)

A term from the paper "GEO: Generative Engine Optimization" (Aggarwal et al., Princeton, Georgia Tech, Allen Institute for AI and IIT Delhi, arXiv 2311.09735, November 2023), describing methods to improve how often content is cited in AI-generated search answers. In practice, AEO and GEO describe the same goal. GEO is the academic term, AEO is the term more common with agencies and business owners.

Source: Aggarwal et al., arXiv 2311.09735

AI Overview

Google's AI-generated summary shown at the top of some search results. Per Google's own documentation, a page has no special requirements to appear in one beyond being indexed and eligible to show a normal search snippet.

Source: Google Search Central, AI features

AI Mode

A conversational, AI-first search experience inside Google Search, separate from the standard results list, that supports follow-up questions in the same session.

Source: Google, AI Mode updates

Answer engine

Any tool that answers a question directly in its own generated response, instead of returning a list of links to click. ChatGPT, Perplexity, Google's AI Overviews and AI Mode, and Gemini are all answer engines.

LLM (large language model)

The type of AI model, such as GPT, Gemini or Claude, that powers chatbots and AI search answers, trained on large amounts of text to generate responses.

Prompt

The text a person types to an AI system to ask it something or tell it to do something. Every AI answer starts from a prompt, so the wording of the prompt shapes what comes back.

Context window

The amount of text a model can work with at one time, counting everything already said plus what it is about to write. Anthropic describes it as the model's working memory, separate from what it learned in training.

Source: Anthropic docs, context windows

Knowledge cutoff

The date up to which a model learned things during training. Anything newer is unknown to the model unless the system retrieves fresher information at answer time. Anthropic publishes a knowledge cutoff date for each of its models.

Source: Anthropic docs, model comparison

Crawling

The process by which a bot, such as Googlebot or GPTBot, visits and downloads pages so they can be indexed or used to generate answers.

Indexing

Storing and organizing crawled pages so they can be retrieved in response to a search or a prompt. A page must be indexed to be eligible for an AI Overview.

Source: Google Search Central, AI features

AI crawler / bot

A bot run by an AI company to collect pages for training, search, or live answers. The named ones are GPTBot (OpenAI, collects content that may train its models), ClaudeBot (Anthropic, collects web content for training), PerplexityBot (Perplexity, surfaces sites in its search results, not for training models), and Google-Extended (a robots.txt token controlling whether Google may use crawled content for Gemini training and grounding).

Sources: OpenAI · Anthropic · Perplexity · Google

Retrieval

The step where an AI system searches an index for the documents most relevant to a question, before writing an answer.

RAG (retrieval-augmented generation)

An AI approach that retrieves relevant documents before generating a response, so the answer is grounded in that material rather than only the model's training data.

Passage ranking

A Google system that weighs individual sections of a page, not just the page as a whole, when judging relevance. Google uses it to identify single passages and better understand how relevant a page is to a search.

Source: Google, ranking systems guide

Semantic search

Matching what words mean rather than the exact words typed. Google's RankBrain returns relevant content even without all the exact search words, by understanding how words relate to concepts.

Source: Google, ranking systems guide

Query fan-out

When an AI search feature runs several related searches behind one question to cover subtopics before writing its answer. Google says its AI features use query fan-out to find more supporting pages and show a wider set of links.

Source: Google Search Central, AI features

Featured snippet

A box at the top of Google results that shows the answer text first, with the source page linked below it. Google's systems choose which page gets shown there, and a site owner cannot mark a page as one.

Source: Google Search Central, featured snippets

Grounding

Anchoring an AI's answer in retrieved sources or verified data, rather than relying only on what the model learned during training. Anthropic's own guidance on reducing hallucinations recommends grounding responses in direct quotes from source material.

Source: Anthropic docs, reduce hallucinations

Hallucination

When an AI model produces a response that sounds confident and plausible but is factually wrong, because it filled a gap with fluent text instead of a verified fact.

Source: Anthropic support, on hallucinations

Embedding

A numeric representation of text that captures its meaning, letting an AI system measure how similar two pieces of text are even if they share no words.

Entity

A specific, identifiable thing, such as a business, person, product or place, that a search or AI system can recognize and connect to facts across sources.

Listing

A directory or profile entry for a business, such as a Google Business Profile, Yelp or Bing Places listing, stating its name, address, phone number and category.

Google Business Profile

Google's free profile for a local business, controlling how it appears on Google Search and Maps. Google tells owners to claim it to manage how their business shows up.

Source: Google, establish business details

AI mention

An AI assistant naming a specific business in its answer to a user's question.

Source authority

How much weight a search or AI system gives a source when deciding whether to use it. Google describes many separate ranking systems rather than one authority score, so treat any single authority number as the tool vendor's estimate.

Source: Google, ranking systems guide

Brand mention

Your business name appearing in text anywhere on the web: a review, a news article, a forum thread, or an AI answer. Linked or not, each mention gives engines another independent sighting of your name.

E-E-A-T

A framework in Google's Search Quality Rater Guidelines, standing for Experience, Expertise, Authoritativeness and Trustworthiness, used by human raters to judge content reliability. Google has stated E-E-A-T is not itself a direct ranking factor, but reflects qualities its ranking systems are built to reward.

Source: Google Search Central, December 2022

Schema markup / structured data

Code added to a page, usually as JSON-LD, that explicitly labels what content is, so search and AI systems don't have to guess from layout.

Knowledge graph

A structured database of entities and the verified relationships between them, which some systems draw on to answer factual questions directly.

Zero-click search

A search where the person gets their answer on the results page itself and never clicks through to a website. 68.01% of Google searches did this in the first four months of 2026.

Source: SparkToro and Similarweb, 2026

Share of voice

The share of a defined set of AI answers where a brand appears, compared with competitors. No single standard methodology exists for measuring this; different tools and agencies calculate it differently, so treat any specific share-of-voice number as tied to whatever method produced it, not a universal standard. Webiyo measures this for clients every week in Answer Check.

Prompt tracking

Running the same set of real customer questions across AI engines on a fixed schedule, then recording which businesses get named. One round of answers is a snapshot; repeated rounds show a trend. This is what Webiyo's Answer Check does every week.

Source: Webiyo Answer Check

robots.txt

A text file at a site's root telling crawlers, including AI bots, which parts of the site they may access. Most crawlers honor it voluntarily; it is not a technical block.

llms.txt

A proposed text file, published by Jeremy Howard of Answer.AI in September 2024, giving AI models a short map of a site's key pages, since models have far smaller context windows than a full site. A voluntary community convention, not an official standard adopted by any AI company.

Source: llmstxt.org proposal

Canonical URL

The one address a site owner picks to represent a page when several addresses show the same content. Google calls this selecting a representative URL for duplicate content, and treats that page as the primary one to evaluate.

Source: Google, canonicalization

Crawl budget

The time and resources Google can spend crawling one site, set by how much the server can handle and how much demand there is for fresh crawls. Google says most small sites never need to think about it.

Source: Google, crawl budget guide

Sitemap

An XML file listing a site's pages, submitted to help search engines discover and crawl it.

Rich Results Test / structured data testing

Google's free tool for checking which rich results a page's structured data can generate, with a preview of how they look. Google recommends starting with the Rich Results Test; the old Structured Data Testing Tool now lives on as the Schema Markup Validator.

Source: Google, test structured data