AI search glossary
AI search has a vocabulary of its own, from how answers are formed to how brands measure visibility. These 68 terms are defined in plain language and linked to relevant Allswap resources.
Core concepts
Answer engine optimizationAnswer engine optimization is the practice of getting your brand named and cited when AI assistants answer questions. Where SEO works to rank a page in a list of links, AEO works to be the recommendation inside the answer itself.Generative engine optimizationGenerative engine optimization is the practice of improving how generative AI systems describe and recommend a brand. It is used interchangeably with answer engine optimization, with GEO emphasising the generated text and AEO the answer it forms.LLM SEOLLM SEO is optimising for large language models rather than for search result pages. It covers everything that shapes what a model says about a brand: retrievable facts, trusted third-party sources, and clearly written pages.Answer engineAn answer engine is a system that responds to a question with a direct written answer instead of a list of links. ChatGPT, Claude, Gemini, Perplexity, Google AI Overviews, Copilot, Grok and Meta AI are all answer engines.AI searchAI search is searching by asking a question in natural language and receiving a written answer, usually with citations. It covers assistants like ChatGPT and Perplexity as well as AI summaries inside traditional search engines.AI visibilityAI visibility is how often and how prominently a brand appears in answers from AI assistants. It is usually expressed as the share of tracked questions where an engine names the brand, weighted by position in the answer.Zero-click searchA zero-click search is one where the person gets their answer without visiting any website. AI answers and search summaries produce many of them, which makes being named inside the answer more valuable than ranking below it.
Measurement
AI visibility scoreAn AI visibility score is a single number for how often answer engines name a brand across a tracked set of questions, weighted by where in the answer the brand appears. It is usually reported per engine and as one rolled-up figure.Share of voiceShare of voice in AI search is the percentage of all brand mentions across a set of answers that belong to one brand. If engines name brands forty times across your tracked questions and eight of those are you, your share of voice is twenty percent.Brand mentionA brand mention is any time an answer engine names a brand in its response, whether or not it links to that brand's website. Mentions are counted separately from citations because a brand can be recommended without being linked.CitationA citation is a link an answer engine attaches to its response to show where information came from. Citations reveal which sources shaped the answer, which makes them the clearest signal of what to work on.Cited sourceA cited source is a page or domain that an answer engine links to when answering a question. In most categories a small group of sources accounts for the majority of citations, and that group decides which brands get recommended.Citation gapA citation gap is a source that answer engines trust in your category, that already cites your competitors, and that has not yet cited you. Closing one is among the most direct ways to improve AI visibility.Answer positionAnswer position is where a brand appears in the order an engine lists recommendations. Being named first carries substantially more weight than being named fourth, because readers rarely act past the first two or three options.Sentiment scoreA sentiment score measures the tone an answer engine uses about a brand — positive, neutral or negative. It is best scored on the sentence containing the mention rather than the whole answer, which can be warm about a category and specific about a brand.Brand perceptionBrand perception in AI search is how engines rate a brand on the attributes buyers care about — price, quality, support, speed. Mapping it trait by trait shows where a brand is already strong in the model's description of it.Topic coverageTopic coverage is how visible a brand is inside each subject its category is made of — pricing, integrations, security, support. It turns a scattered list of prompt results into a short list of subjects worth working on.Competitive setA competitive set is the group of brands an answer engine treats as alternatives to yours. It is read from the answers themselves rather than chosen, so it often includes companies a sales team would not have listed.AI referral trafficAI referral traffic is website sessions that arrive from an AI assistant, identified by the referrer the assistant attaches to its links. It usually converts well, because the assistant has already shortlisted before the click.
How engines work
Large language modelA large language model is a system trained on very large amounts of text to predict and generate language. It is the component that writes an answer engine's response, drawing on what it learned in training plus anything retrieved at the moment of asking.Retrieval-augmented generationRetrieval-augmented generation is when a model searches for relevant documents and writes its answer from what it finds, rather than from training alone. It is why a page published this week can appear in an answer today.GroundingGrounding is tying a model's answer to specific retrieved sources rather than to its own recollection. A grounded answer can be traced back to documents, which is what makes citations possible and answers more reliable.Query fan-outQuery fan-out is when an answer engine turns one question into several searches of its own before answering. Asking about the best option in a category may quietly become separate searches for price, reviews, warranty and alternatives.HallucinationA hallucination is a confident statement from a model that is not supported by its sources — a wrong price, a feature attributed to the wrong company, an outdated fact stated as current.Training dataTraining data is the body of text a model learned from. It sets the model's default understanding of a category and which brands it associates with what, before any live retrieval happens.Knowledge cutoffA knowledge cutoff is the date after which a model's training data ends. Anything that changed after that date reaches an answer only if the engine retrieves it live at the time of asking.EmbeddingAn embedding is a numeric representation of a piece of text that places similar meanings close together. Embeddings let a system match a question to relevant passages even when they share no words.Vector searchVector search finds documents by comparing embeddings rather than matching keywords, retrieving passages whose meaning is close to the question. It is the retrieval method behind most answer engines.Semantic searchSemantic search retrieves results by meaning rather than by exact wording, so a page can be returned for a question that shares none of its words. It underpins how answer engines find material to answer from.RerankingReranking is a second pass that reorders retrieved documents by how well they actually answer the question, before the model writes from them. It decides which of the candidates becomes a citation.Context windowA context window is how much text a model can consider at once, covering the conversation so far plus any retrieved documents. It limits how many sources can inform a single answer.Multi-turn conversationA multi-turn conversation is an exchange where the person asks follow-up questions and the assistant carries the earlier context forward. Recommendations often change between the first turn and the one where a decision is made.
Crawlers and access
AI crawlerAn AI crawler is an automated agent that collects web pages for an AI company, either to inform model training or to build the search index an assistant retrieves from. Each is identified by its own user-agent string in robots.txt.GPTBotGPTBot is OpenAI's crawler for collecting web pages that may inform future model training. It is controlled in robots.txt by its user-agent name, separately from OpenAI's search and live-fetch agents.OAI-SearchBotOAI-SearchBot is OpenAI's crawler for building the search index that ChatGPT draws on when it searches the web. Allowing it is what makes your pages eligible to be retrieved and cited in ChatGPT answers.PerplexityBotPerplexityBot is Perplexity's crawler, which indexes pages so they can be retrieved and cited in its answers. Because Perplexity cites heavily, allowing it is one of the most direct routes to being named in AI answers.ClaudeBotClaudeBot is Anthropic's crawler, collecting public web pages that may inform Claude's models. It is one of three Anthropic agents, alongside Claude-SearchBot for search quality and Claude-User for pages a person asked Claude to read.Google-ExtendedGoogle-Extended is the robots.txt control over whether your content may be used for Gemini training and for grounding in Gemini Apps and Vertex AI. Google states it does not affect inclusion in Google Search and is not used as a ranking signal.robots.txtrobots.txt is a file at the root of a site that tells crawlers which paths they may fetch. For AI search it is the primary control over whether assistants can reach your pages at all.llms.txtllms.txt is a plain-text file at the root of a site that gives AI systems a curated summary of what the site offers and where its key pages are. It is a proposed convention, not a requirement, and it is inexpensive to publish.Crawl budgetCrawl budget is how much crawling a site receives in a given period. Keeping sitemaps accurate and avoiding duplicate URLs helps crawlers spend that budget on pages you want retrieved.SitemapA sitemap is an XML file listing the URLs a site wants crawled, with optional hints about change frequency and language alternates. It is how a crawler discovers pages that navigation alone might not surface.
Google surfaces
AI OverviewsAI Overviews is Google's AI-generated summary that appears above the ordinary search results. For many queries it is the first answer a person reads, and the brands it names are seen before any organic link.AI ModeAI Mode is Google's conversational search experience, where a person asks a question and follows up in a dialogue rather than scanning a results page. It sits alongside AI Overviews as an AI-first way into Google.Featured snippetA featured snippet is a short extract Google promotes above its results to answer a query directly. It was the earliest widely-used answer format and rewards the same writing that AI answers do: a direct answer, stated early.SERPA SERP is the page of results a search engine returns for a query. Modern SERPs mix organic links with AI summaries, snippets and other panels, so a ranking position is only part of what a searcher sees.E-E-A-TE-E-A-T is Google's framework for assessing content quality: experience, expertise, authoritativeness and trust. Named authors, clear credentials and consistent facts are the practical signals behind it.Knowledge graphA knowledge graph is a structured store of entities and the relationships between them — companies, people, products and how they connect. Search and AI systems use it to know what a name refers to.
Content and entities
EntityAn entity is a distinct thing a system can recognise and reason about — a company, a product, a person, a place. Search and AI systems work with entities rather than strings, so being recognised as one matters.Brand aliasA brand alias is an alternative name for a brand — a product name, a shortened form, a legal entity name or a common misspelling. Tracking aliases is what stops a mention being missed when the company name is never used.Structured dataStructured data is machine-readable markup that states what a page is and what it contains. It helps search engines show rich results and helps AI systems understand a page before deciding whether to use it.Schema markupSchema markup is structured data written with the schema.org vocabulary, describing a page as an article, a product, an FAQ, an organisation or another recognised type. It makes a page's purpose explicit rather than inferred.JSON-LDJSON-LD is the recommended format for structured data: a block of JSON in the page head describing the page using schema.org types. It is preferred because it sits apart from the visible markup and is easy to maintain.FAQ schemaFAQ schema is structured data marking a page's questions and answers as an FAQPage. It makes question-and-answer content explicit, which suits both rich results and AI systems looking for a direct answer to lift.Canonical URLA canonical URL tells search engines which version of a page is the one to index when the same content is reachable at several addresses. It consolidates signals onto a single address instead of splitting them.hreflanghreflang is markup that tells search engines which language and region each version of a page serves. A valid cluster names every version including itself, so readers are sent to the right one.Topical authorityTopical authority is how thoroughly and credibly a site covers a subject area. Depth across a subject, rather than a single strong page, is what makes a site a dependable source on it.Source authoritySource authority is how much weight answer engines give a particular website when answering questions in a category. High-authority sources are cited repeatedly, and the brands they name benefit accordingly.Content freshnessContent freshness is how recently a page was published or meaningfully updated. Answer engines favour current sources on fast-moving subjects, and a visible update date helps both readers and machines judge it.Answer-first contentAnswer-first content states the answer in the opening lines, before context or background. It suits AI answers because engines lift self-contained passages, and a passage that answers the question outright is the easiest to take.
Prompts and buyers
PromptA prompt is the question or instruction someone gives an AI assistant. In AEO, prompts take the place of keywords: they are whole questions, usually carrying context, and that context shapes which brands get named.Prompt trackingPrompt tracking is monitoring a fixed set of buyer questions across answer engines over time, recording which brands each engine names, in what order, and which sources it cites.Branded promptA branded prompt names a specific company or product — asking whether a brand is any good, or how it compares with a named rival. These answers shape the impression of a brand already under consideration.Unbranded promptAn unbranded prompt describes a need without naming any company — asking for the best option for a particular situation. These are where new brands enter a buyer's consideration set.Comparison promptA comparison prompt asks an assistant to weigh two or more named options against each other. The answer usually sets out differences attribute by attribute, and often carries a recommendation.Buyer intentBuyer intent is what someone is trying to achieve with a question — exploring a category, comparing options, resolving a concern, or ready to act. Intent determines how much a mention in that answer is worth.Funnel stageA funnel stage is where a question sits in the buying journey — discovery, comparison, objection or decision. Being named at the decision stage is worth considerably more than being named at discovery.Conversational searchConversational search is looking for information through a dialogue, asking follow-up questions rather than issuing a new query each time. The assistant carries context forward, narrowing toward a recommendation.
Ready to Dominate AI Search?
Get started now
Allswap