AI Search Glossary

Shopping where an AI agent acts for the buyer: it searches, compares and can complete the purchase without the buyer visiting the store. Protocols like OpenAI and Stripe's Agentic Commerce Protocol give agents and merchants a shared way to handle orders.

An open standard from OpenAI and Stripe, announced in September 2025, that lets AI agents place orders with merchants. It powers checkout directly inside ChatGPT.

Marketing work carried out by AI agents that plan and execute tasks, such as drafting content, fixing technical issues or sending outreach, with people setting goals and approving what ships.

Search where an AI agent runs its own research loop: it plans queries, reads results, spots what is missing and searches again. Deep research modes work this way and read far more pages than a single search.

A record of the ads that appear across AI search engines, showing which advertisers bid on which prompts and topics, and what their creatives say.

An AI brand mention is any time an AI assistant names your brand in its answer, with or without a link to your site. Mentions are the base unit of AI visibility: share of voice, visibility score and sentiment are all calculated from them.

AI brand sentiment is the tone an AI assistant uses when it describes your brand: positive, neutral or negative. It is measured by classifying each mention and the claims around it, such as 'reliable', 'expensive' or 'poor support', then aggregating across prompts and engines.

AI citations are the source links an AI engine shows with its answer to credit the pages it drew on. A citation points to a specific URL, which may be your site or a third party like a review site, a news article or a Reddit thread. Citations tell you which sources shape what AI says about your category.

AI crawlers are bots that AI companies use to fetch web pages, either to collect training data, build a search index, or retrieve a page live when a user asks a question. Examples include GPTBot, OAI-SearchBot, ChatGPT-User, ClaudeBot, PerplexityBot and Google-Extended. If they cannot reach your pages, AI engines cannot cite them.

An AI hallucination is a confident statement from an AI model that is false, such as a wrong price, a feature you never offered or a review that does not exist. Hallucinations about brands usually come from missing, outdated or contradictory information online.

How AI models understand and describe your brand: the attributes, keywords and comparisons they attach to you, and the sources that shape those descriptions.

AI referral traffic is visits to your website from links inside AI answers, such as a cited source in ChatGPT, Perplexity, Gemini or Copilot. Analytics tools detect it from the referring domain, for example chatgpt.com or perplexity.ai, or from UTM tags the engine adds.

The umbrella term for improving how a brand appears across AI-powered search, from chat assistants to AI Overviews. It covers GEO, AEO and the technical, content and PR work behind them.

AI share of voice is your brand's portion of all brand mentions across a set of AI answers. If ChatGPT names 400 brands across your tracked prompts and 60 of those mentions are yours, your share of voice is 15%. It shows how much of the AI conversation in your category you own compared with competitors.

AI shopping is product discovery inside AI assistants, where ChatGPT, Perplexity, Gemini and others recommend specific products, often as cards with images, prices and merchant links. Which products appear depends on product data, reviews and the sources the assistant trusts.

An AI visibility score is the share of tracked AI answers that mention your brand. Run a fixed set of buyer prompts through engines like ChatGPT and Perplexity, count the answers that name you, and divide by all answers. A brand named in 18 of 100 answers has a visibility score of 18%.

Measuring how your brand shows up in AI answers over time, by running the same prompts across engines, countries and languages and comparing mentions, positions, sentiment and citations from run to run.

How ready a website is for AI engines: whether AI crawlers can reach and render its pages, and whether content, metadata and structured data are clear enough to be retrieved and cited.

A search tool that replies with a written answer instead of a list of links. ChatGPT, Perplexity, Gemini, Claude and Google AI Overviews are all answer engines.

Answer engine optimization (AEO) is the practice of structuring content so answer engines can lift it directly into a response: clear questions, direct answers in the first sentences, and facts that stand on their own. Today AEO and GEO are used almost interchangeably for optimizing visibility in AI answers.

Average position is where your brand tends to appear when an AI answer lists several brands. A brand named first in one answer and third in another has an average position of 2. Lower is better, because buyers read the first recommendations most closely.

A branded prompt names your company, like 'Is Flatpay good for small shops?'. An unbranded prompt describes a need without naming anyone, like 'best card terminal for a small shop'. Unbranded prompts show whether AI recommends you to buyers who have not heard of you yet.

OpenAI's AI assistant and the largest AI answer engine by usage. It answers from what its model learned in training and, when it searches, from live web results that it cites as sources.

ChatGPT ads are paid placements that OpenAI shows inside ChatGPT conversations, clearly labelled as sponsored and kept separate from the organic answer. They are shown to users on the Free and Go plans and expanded to 31 European markets in August 2026, giving brands a paid route into AI answers alongside organic visibility.

The web search capability inside ChatGPT. When a question needs current information, ChatGPT runs searches, reads the results and shows the pages it used as cited sources.

Anthropic's AI assistant. It can search the web and cite sources when a question needs current information.

In AI search, competitor gap analysis finds the prompts where AI engines recommend a competitor but not you, and the sources they cite when they do. Each gap points to a concrete fix: a missing page, a review site you are absent from, or a publication that only covers your rival.

Splitting pages into smaller passages that AI systems can retrieve and quote on their own. Writing in clear, self-contained sections makes each passage easier to lift into an answer.

Signals that a page is current, such as a visible updated date, recent figures and references to the current year. AI engines that retrieve live results often prefer recent sources for questions where things change.

Searching by asking full questions in natural language and refining them with follow-ups, while the engine keeps the context of the conversation.

A Chinese AI lab and assistant known for open-weight models. Its assistant is used widely enough that brands with global audiences track it alongside ChatGPT and Gemini.

Earning mentions on the publications, review sites and communities that AI engines cite for your category. The goal is less about backlinks and more about being named in the sources models trust.

Experience, expertise, authoritativeness and trustworthiness: the qualities Google's quality rater guidelines use to judge content. Pages with clear authorship, first-hand evidence and a trustworthy reputation also tend to be cited more by AI engines.

Earned media is coverage of your brand that you do not pay for or publish yourself: press articles, reviews, listicles, forum threads and creator content. AI engines lean heavily on these third-party sources when they recommend brands, which makes earned media one of the biggest levers in AI visibility.

Numerical representations of meaning. Text with similar meaning gets similar numbers, which lets AI systems match a question to a relevant passage even when they share few words.

Making your brand a clearly defined entity that search engines and language models can recognise, with consistent facts about what you are, what you sell and who you compete with across your site and the wider web.

Google's family of AI models and its standalone assistant. Gemini models also power Google AI Overviews and AI Mode inside Google Search.

Generative engine optimization (GEO) is the work of getting your brand mentioned, recommended and cited in answers from AI engines like ChatGPT, Perplexity, Gemini and Google AI Overviews. It combines content built for retrieval, technical access for AI crawlers, and presence on the third-party sources those engines trust.

A conversational mode in Google Search that answers with an AI-written response, supports follow-up questions and splits complex questions into many sub-searches behind the scenes.

Google AI Overviews are AI-written summaries shown at the top of some Google search results, with links to the pages they draw on. They launched widely in May 2024 and replaced the Search Generative Experience test. Being cited in an Overview now matters as much as ranking below it.

xAI's AI assistant, built into X. It draws on web results and posts on X, so social conversation can shape how it describes brands.

Tying an AI model's answer to specific source material, such as live search results or a document set, instead of relying only on training data. Grounded answers can show citations.

How much new information a page adds compared with what already exists on the topic. Original data, first-hand testing and specific details give AI engines a reason to cite you instead of the ten pages that say the same thing.

The date after which a model has no training data. Anything newer, like a rebrand or new product, is only known to the model if it searches the web while answering.

A database of entities and the relationships between them. Knowledge graphs help search engines and AI systems work out which brand, person or product a question refers to.

An AI model trained on very large amounts of text to understand and generate language. LLMs are the engines inside ChatGPT, Claude, Gemini and most AI search tools.

Search optimization aimed at large language models rather than ranked links. It focuses on what models retrieve and repeat: your own pages, and the third-party sources that mention you.

llms.txt is a proposed standard for a Markdown file at the root of a website (example.com/llms.txt) that gives language models a short summary of the site and links to its most useful pages. Jeremy Howard proposed it in September 2024. It is cheap to add, but no major AI engine has confirmed that it uses the file.

Meta's AI assistant inside WhatsApp, Instagram, Facebook and Messenger, which puts AI answers in front of people where they already chat.

Microsoft's AI assistant across Windows, Edge, Bing and Microsoft 365. Its web answers are grounded in Bing results, so Bing visibility feeds Copilot visibility.

The Model Context Protocol (MCP) is an open standard, introduced by Anthropic in November 2024, for connecting AI assistants to outside tools and data. An MCP server describes its tools in a common format, so assistants like Claude, ChatGPT and Cursor can use them without a custom integration for each.

What a language model knows from training, stored in its weights, as opposed to what it retrieves live. Brands that were widely discussed before the cutoff are often recommended even when the model does not search.

An AI answer engine that searches the web for almost every question and shows numbered citations next to its answer, which makes its sources unusually easy to analyse.

Structured product data, such as titles, prices, availability and attributes, sent to or read by AI shopping surfaces. Complete, accurate feeds make it easier for assistants to recommend the right product.

Prompt tracking means running the same set of buyer questions through AI engines on a fixed schedule and recording which brands, positions and sources appear in each answer. Because AI answers vary run to run, repeated tracking turns noisy single answers into trends you can act on.

An estimate of how often people ask AI assistants a given question or topic. AI platforms do not publish query data, so prompt volume is modelled from search demand and other signals and is best used to prioritise, not as an exact count.

Query fan-out is when an AI engine breaks one question into several narrower searches, runs them in parallel, and merges the results into one answer. A question like 'best CRM for a 10-person agency' might fan out into searches about pricing, integrations and reviews.

Retrieval-augmented generation (RAG) is how most AI search works: the system first retrieves documents relevant to the question, then the language model writes its answer from them. Because the sources are fetched at answer time, RAG lets AI use current information and cite where it came from.

Using robots.txt rules to allow or block individual AI crawlers. Blocking a training crawler and blocking a search or user-triggered crawler have different effects, so each user agent needs its own decision.

Search that matches the meaning of a query rather than its exact words, so 'cheap card reader for a market stall' can find a page about 'affordable payment terminals'.

Building a page's full HTML on the server before sending it. Many AI crawlers do not run JavaScript, so content that only appears after client-side rendering can be invisible to them.

Sources are every page an AI engine retrieved while building an answer. Citations are the subset it shows to the reader as links. A page can influence an answer without being cited, so tracking only visible citations underestimates which domains shape AI answers about your brand.

Any paid unit shown in or beside an AI-generated answer, such as sponsored product cards or ads below a response. They sit next to the organic answer rather than changing what the model says.

Structured data is code, usually schema.org vocabulary in JSON-LD, that labels what a page contains: a product with a price, an organization with its logo, an FAQ, an article and its author. It helps search engines and AI systems read your facts without guessing.

How thoroughly and credibly a site covers a subject across many connected pages, rather than with one article. Deep coverage makes a site a more likely source when AI engines fan out related queries.

Places your brand is named without a link, such as in an article, forum thread or AI answer. Language models read the brand name itself, so unlinked mentions still shape AI visibility even though they pass no link value in classic SEO.

A file listing the URLs you want crawled, often with the date each changed. It helps search and AI crawlers discover new and updated pages faster.

A zero-click search is one where the user gets their answer without visiting any website, because the search page or AI assistant answers directly. As AI answers spread, more discovery ends inside the answer, so being named there matters even when no click follows.