AI visibility, term by term

AI citation

An AI citation is a source link that an AI engine attaches to part of its generated answer, crediting the web page the information was drawn from. Citations are distinct from mentions: a page can be cited without its brand being named in the answer, and a brand can be named without any of its pages being cited.

AI crawler

An AI crawler is an automated bot that reads web pages to feed an artificial intelligence system. AI crawlers fall into three functional types: training crawlers such as GPTBot collect content to train future models, search index crawlers such as OAI-SearchBot build the retrieval indexes behind AI search, and user-triggered fetchers such as ChatGPT-User load a specific page live during a conversation.

AI hallucination (brand context)

An AI hallucination is a confident, false statement generated by a language model. In a brand context it means an AI assistant misdescribing your product, inventing features or pricing, attributing fake reviews or quotes, or recommending a company that does not exist. OpenAI research from 2025 argues hallucinations persist because standard training rewards confident guessing over admitting uncertainty.

AI mention

An AI mention is an appearance of a brand name in the text of an AI-generated answer, independent of whether the engine links to the brand's website. Mentions are the raw unit of AI visibility measurement: the rate at which a brand is mentioned across repeated runs of a fixed question set is its visibility score.

AI visibility

AI visibility is the degree to which AI assistants such as ChatGPT, Gemini, Perplexity, Claude and Grok mention or recommend a brand when users ask relevant buying questions. It is measured by running a fixed set of unbranded questions repeatedly across engines and recording how often each brand appears in the generated answers.

AI visibility score

An AI visibility score is the percentage of unbranded buying questions for which an AI engine names a brand in its answer, sampled over repeated runs because answers vary between identical queries. Three choices decide whether the score is trustworthy: the question set, the sampling method, and the parsing method that decides what counts as a mention.

Answer engine

An answer engine is an AI system that responds to a question with a direct, synthesized answer instead of a list of links. ChatGPT, Perplexity, Google AI Mode, Gemini and Microsoft Copilot are answer engines: they retrieve sources, compose a single response, and typically name a small set of brands or citations rather than ranking ten results for the user to evaluate.

Answer engine optimization (AEO)

Answer engine optimization (AEO) is the practice of structuring content so that answer systems, from featured snippets to AI chatbots, select and present it as the direct answer to a user's question. Its signature techniques are answer-first paragraphs, question-shaped headings and one page per question. AEO grew out of featured snippet optimization and now substantially overlaps with generative engine optimization.

Branded vs unbranded questions

A branded question contains a brand's name, such as asking whether Acme is good for invoicing. An unbranded question describes the need without naming any vendor, such as asking for the best invoicing tool for freelancers. Only unbranded questions test whether an AI assistant recommends a brand unprompted, which makes them the basis of honest visibility measurement.

Citation share

Citation share is the percentage of source citations in AI-generated answers that point to your domain, measured across a defined set of prompts and engines. If answers about your category cite 200 sources in a month and 14 link to you, your citation share is 7 percent. It is the AI search counterpart of ranking share in SEO.

ClaudeBot

ClaudeBot is the web crawler Anthropic operates to collect publicly available content that may be used to train and improve its Claude AI models. It identifies itself with the user agent string Mozilla/5.0 AppleWebKit/537.36 (KHTML, like Gecko; compatible; ClaudeBot/1.0; +claudebot@anthropic.com) and honors robots.txt. It is distinct from Claude-User and Claude-SearchBot, Anthropic's fetch and search crawlers.

Generative engine optimization (GEO)

Generative engine optimization (GEO) is the practice of increasing how often generative AI engines such as ChatGPT, Perplexity, Gemini, Claude and Grok mention or cite a brand when composing answers. The term was coined by a Princeton-led research paper published at KDD 2024, which found that adding statistics, quotations and cited sources lifted visibility in generative responses by up to roughly 40 percent.

Google AI Mode

Google AI Mode is a conversational search experience inside Google Search, a dedicated tab where a Gemini model answers the question end to end, with citations and follow-up questions, instead of returning a ranked list of links. Google reported over 1 billion monthly AI Mode users at I/O 2026. It works through query fan-out, running many background searches and synthesizing one answer.

Google AI Overviews

Google AI Overviews are AI-generated summaries that appear above the traditional results on Google Search, composed by Gemini models from retrieved web pages and displayed with a panel of source links. Rolled out broadly from May 2024, they appeared on roughly 48 percent of tracked queries by March 2026 and substantially reduce clicks to the organic results below them.

Google-Extended

Google-Extended is a robots.txt control token, introduced by Google in September 2023, that lets site owners decide whether their content may be used to train and ground Google's Gemini AI models. It is not a separate crawler and never appears in server logs; Googlebot does the crawling, and the token governs data use. Blocking it does not affect Google Search rankings or inclusion in AI Overviews.

GPTBot

GPTBot is OpenAI's web crawler that collects publicly available content which may be used to train OpenAI's generative AI models. It identifies itself with the user agent string Mozilla/5.0 AppleWebKit/537.36 (KHTML, like Gecko); compatible; GPTBot/1.4; +https://openai.com/gptbot, honors robots.txt rules addressed to the GPTBot token, and is separate from OAI-SearchBot, which feeds ChatGPT search.

Grounding (AI search)

Grounding is the process of anchoring an AI model's answer in retrieved, verifiable sources at generation time, so its claims trace back to real documents instead of relying on training memory alone. In AI search, grounded answers quote and cite live web pages. Google offers Grounding with Google Search for Gemini, and ChatGPT grounds answers through its built-in web search.

Knowledge cutoff

A knowledge cutoff is the date after which a language model's training data ends. Anything that happened later, including product launches, pricing changes, rebrands and new companies, is absent from the model's built-in memory. Web search and retrieval features work around the cutoff, but when an AI answers from memory alone, it describes the world as it stood at that date.

llms.txt

llms.txt is a proposed web standard: a markdown file at a site's root giving AI systems a curated summary of the site and links to its key pages in a format language models parse easily. Proposed by Jeremy Howard of Answer.AI in September 2024, it has real adoption among site owners but weak evidence of use by AI engines, and Google says it does not use the file.

OAI-SearchBot

OAI-SearchBot is the web crawler OpenAI operates to build the search index behind ChatGPT's search features. Pages it can read become candidates for retrieval and citation in ChatGPT search answers. Its robots.txt token is OAI-SearchBot, and its user agent string ends with: compatible; OAI-SearchBot/1.4; +https://openai.com/searchbot. It is separate from GPTBot, OpenAI's training crawler.

PerplexityBot

PerplexityBot is the web crawler Perplexity operates to build the search index behind its AI answer engine. Perplexity documents that it surfaces and links websites in results and is not used to train AI models. Its user agent string is Mozilla/5.0 AppleWebKit/537.36 (KHTML, like Gecko; compatible; PerplexityBot/1.0; +https://perplexity.ai/perplexitybot), and it honors robots.txt rules addressed to the PerplexityBot token.

Retrieval-augmented generation (RAG)

Retrieval-augmented generation (RAG) is an AI architecture that fetches relevant documents from an external source at answer time and passes them to a language model, so the response is built on retrieved evidence rather than training data alone. Introduced by Lewis et al. in 2020, RAG-style pipelines power modern AI search, including ChatGPT search, Perplexity, Google AI Overviews and AI Mode.

robots.txt for AI crawlers

robots.txt for AI crawlers means using the standard robots exclusion file at a site's root to allow or block AI bots individually, by addressing each bot's documented user agent token, such as GPTBot, OAI-SearchBot, ClaudeBot, PerplexityBot or Google-Extended, with Allow and Disallow rules. Because each token governs a different system, training, search indexing or live fetching, a precise file makes separate decisions per bot rather than one blanket choice.

Share of voice (AI search)

Share of voice in AI search is the percentage of AI assistant answers that mention a brand across a fixed set of buyer questions, benchmarked against the competitors named in the same answers. Computing it honestly requires unbranded questions, repeated runs of each question, a separate score per engine, and stored answer text that can be audited.