AI visibility glossary

The AI visibility glossary.

The language of getting found inside AI answers. Every term that matters for ChatGPT, Perplexity and Google's AI Overviews, in plain English.

be the source AI cites ✍
Agentic search When an AI system plans and runs a multi-step research process on its own, issuing searches, reading pages and deciding what to do next before answering. AI crawler A bot that fetches web pages so AI systems can train on them or read them at answer time. GPTBot, ClaudeBot and PerplexityBot are the ones most publishers see in their logs. AI Mode Google's fully conversational search experience, where the entire results page is a generated answer with follow-up questions, rather than a list of links. AI Overviews Google's AI-generated summary above the organic results, written from several sources at once and shown with links to the pages it drew from. AI referral traffic Sessions arriving on your site from an AI assistant, identifiable in analytics by referrers such as chatgpt.com, perplexity.ai, claude.ai and gemini.google.com. Answer Engine Optimization (AEO) Structuring content so an answer engine can lift a clean, direct answer from it and credit you as the source.
Brand sentiment Whether AI answers describe your brand positively, neutrally or negatively when they mention it.
Chain-of-thought A model working through intermediate reasoning steps before giving an answer, rather than producing it in one shot. Citation share The percentage of all sources cited across a set of AI answers that belong to your domain. The AI-era equivalent of ranking share. Content chunking Structuring a page into self-contained sections so retrieval systems can pull the right passage without dragging in context that does not belong to it. Context window The maximum amount of text a model can hold at once, counted in tokens, covering the prompt, any retrieved sources, the conversation so far and the answer being written. Conversational search Searching in full natural sentences with follow-up questions, where the engine keeps context between turns instead of treating each query as new. Cosine similarity The measure retrieval systems use to score how related two pieces of text are, by comparing the angle between their vectors rather than their distance apart.
Deep research An assistant mode that spends minutes rather than seconds on a question, reading dozens of sources and producing a long cited report.
Entity SEO Optimizing around entities, the distinct people, products, brands and concepts an engine recognizes as things, and the relationships between them, rather than around keyword strings.
Fine-tuning Further training an existing model on a specific dataset to change how it behaves, as opposed to retrieving information at query time. Freshness signal The cues that tell a retrieval system how current your content is, including visible dates, structured data timestamps and substantive updates.
Grounding Tying a model's answer to real retrieved sources rather than to its own memory, so claims can be traced back and cited.
Hallucination When a model states something false or invented as though it were fact, with no signal that it is uncertain. Hybrid search Retrieval that runs keyword matching and vector similarity together, then merges the results, because each catches what the other misses.
Knowledge cutoff The date after which a model has no training data, so anything later is unknown to it unless it retrieves the information at query time. Knowledge graph A structured database of entities and the relationships between them, used by search and AI engines to answer factual questions and to resolve which thing a name refers to.
Large language model (LLM) A model trained on very large amounts of text to predict the next token, which is what lets it answer questions, summarize and write. GPT, Claude and Gemini are LLMs. llms.txt A proposed markdown file at your site root that lists your most important pages in a clean, structured form for AI systems to read.
Model training data The text corpus a model learned from, typically a mix of crawled web pages, licensed content, books, code and curated datasets. Multimodal search Search that accepts and understands images, voice and video alongside text, often combining them in a single query.
Passage retrieval Pulling the single most relevant section from a page to answer a query, rather than judging the page as a whole. Prompt The input a user gives an AI model: the question, instruction and any context supplied with it. Prompt engineering Writing instructions that reliably get better output from an AI model, through specificity, context, examples and explicit format. Prompt injection An attack where text hidden in a web page or document manipulates an AI system into ignoring its instructions or repeating attacker-controlled content. Prompt volume How often a given question gets asked of AI assistants. The AI equivalent of search volume, except no reliable public data exists for it.
Query fan-out When an engine silently expands one question into several related sub-queries, retrieves sources for each, and synthesizes them into one answer.
Reranking A second scoring pass that reorders retrieved candidates by relevance and quality before the best few are passed to the model. Retrieval-augmented generation (RAG) A method where the system fetches relevant documents at query time and the model writes its answer from those documents rather than from memory alone.
Semantic search Search that matches the meaning of a query against the meaning of content, using vector similarity rather than keyword overlap. Share of model How often, and how favorably, AI assistants name your brand across a defined set of buyer questions, measured against competitors. Source authority How much weight an AI system gives your domain when deciding which sources to retrieve, trust and cite. Synthetic content Text generated by AI with little human input, verification or original insight. Also called AI slop when published at volume.
Token The unit of text a model processes. Roughly a word fragment, and the thing context windows, pricing and limits are all counted in.
Unlinked brand mention Your brand named in text without a hyperlink, which still builds recognition and reinforces the association between your brand and its category.
Vector database A database built to store embeddings and find the nearest ones to a query vector quickly, which is what makes semantic retrieval fast enough to use. Vector embedding A list of numbers representing a piece of text's meaning, positioned so that similar meanings sit close together in the same space. Visibility score A blended metric combining how often AI assistants mention you, cite you and describe you favorably across a fixed set of queries.
Zero-click search A search that ends on the results page because the answer is displayed there and the user never clicks a result.

Know the terms. Want the content?

We turn these concepts into blogs, comparison pages and lead magnets that actually rank, in search and in AI answers.

Book a 30-min call →