Ads in AI Answers
Paid placements inside or beside AI-generated answers. Google shows ads in AI Overviews and AI Mode, Perplexity has tested sponsored follow-up questions, and OpenAI began testing ads in ChatGPT in 2026.
Glossary
Plain-language definitions of the terms used across AI search, generative and answer engine optimization, AI crawlers, technical SEO, content, authority and measurement.
Paid placements inside or beside AI-generated answers. Google shows ads in AI Overviews and AI Mode, Perplexity has tested sponsored follow-up questions, and OpenAI began testing ads in ChatGPT in 2026.
An open protocol announced by Google in September 2025 for AI agents to make payments on a user’s behalf with verifiable authorization.
A web browser with a built-in AI agent that can read and act on pages for the user. Examples include Perplexity’s Comet and OpenAI’s ChatGPT Atlas.
Shopping in which an AI agent discovers, compares and buys products for a user, often without the user visiting the merchant’s site.
An open standard from OpenAI and Stripe, announced in September 2025 with Instant Checkout in ChatGPT, that lets AI agents and merchants exchange product, checkout and payment data. In March 2026 OpenAI extended it to product discovery and let merchants use their own checkout; OpenAI now describes ACP as the open infrastructure for commerce in ChatGPT.
An AI system that carries out multi-step tasks for a user, such as researching options, filling in forms or making a purchase, rather than only answering a question.
Whether AI answers state a brand’s facts correctly: what it sells, prices, features, locations and who it serves.
A source an AI answer links to or credits for its information. A brand can be cited without being recommended, and recommended without its own site being cited.
A bot that fetches web pages for AI companies. Some collect training data, some build search indexes for AI answers, and some fetch a page live when a user asks about it. They are controlled separately in robots.txt.
An instance of a brand being named in an AI-generated answer, with or without a link to its site. Mentions shape what buyers hear even when nothing is clicked.
Google Search’s conversational, AI-first mode, opened to all US users in May 2025. It answers complex questions using query fan-out and supports follow-up questions.
AI-generated summaries at the top of some Google search results, with links to sources. Launched in the US in May 2024, evolving from the earlier Search Generative Experience (SGE).
When an AI answer suggests a brand, product or provider as a good choice for the user’s need, often in a shortlist or ranked list. For commercial questions this is the outcome that matters most.
Website visits that arrive from links in AI assistants, such as chatgpt.com, perplexity.ai or gemini.google.com, measurable in analytics tools like Google Analytics 4.
Search experiences in which an AI model writes an answer instead of, or alongside, a list of links. Examples include ChatGPT search, Google AI Overviews and AI Mode, Perplexity and Microsoft Copilot.
The umbrella term for work that improves a brand’s presence in AI-powered search and answer experiences. It covers GEO, AEO and the technical, content and authority work behind them.
A broad label used in two ways: optimizing for AI-driven search results (the same work as GEO and AEO), or using AI tools to do SEO tasks such as research, writing and auditing. Check which meaning is intended.
A brand’s share of all brand mentions, or recommendations, in AI answers to a defined set of questions, compared with competitors.
The small set of brands an AI answer names when a buyer asks which products or providers to consider. Being on it is the AI-era equivalent of making a vendor shortlist.
How often and how prominently a brand appears in AI-generated answers for the questions its buyers ask: whether it is mentioned, where it is placed, how it is described and whether its pages are cited.
A summary metric that combines how often and how prominently a brand appears across tracked AI answers. Each tool defines it differently, so compare scores only within one tool.
The clickable text of a link. Descriptive, natural anchor text tells search engines what the linked page is about.
A system that responds to a question with a direct answer rather than a ranked list of pages. Perplexity describes itself this way; ChatGPT search and Google AI Mode work similarly.
Structuring content so search engines and AI assistants can extract it as a direct answer, in featured snippets, voice answers and AI-generated responses. Often used interchangeably with GEO; AEO predates generative AI and grew out of featured-snippet and voice-search optimization.
The tendency of AI assistants to give different answers, and cite different sources, to the same question asked at different times. It is why AI visibility is measured across repeated runs rather than one check.
Putting the direct answer to a question in the first sentence or two of a section, then adding detail. It makes content easy to extract for snippets and AI answers.
A robots.txt token that lets sites opt out of having content crawled by Applebot used to train Apple’s AI models. It does not affect Applebot’s use for Siri and Spotlight results.
A conversion in which a channel played a part earlier in the journey but was not the last touch before the conversion.
Assigning credit for conversions and revenue to the channels that influenced them. For AI search, referral visits can be attributed; the influence of answers that sent no click usually cannot be measured directly.
Evidence that content was written by someone qualified: named authors, credentials, firsthand experience and a consistent presence on the topic.
Where a brand typically appears among the brands an AI answer lists. Being named first carries more weight than being named fifth.
A link from another website to yours. Links from relevant, reputable sites remain a core signal for search engines.
Microsoft’s free tool for site owners, covering indexing, search performance and IndexNow submissions for Bing, which also supplies Microsoft Copilot.
Microsoft’s crawler for the Bing index, which also supplies web results to Microsoft Copilot.
Any reference to a brand on a third-party site, review platform, forum or social network, linked or not. Mentions across trusted sources shape how AI models describe a brand.
How often people search for a brand by name. It reflects awareness that the brand built elsewhere, including in AI answers that never sent a click.
A crawler operated by ByteDance, the parent company of TikTok, associated with collecting data for its AI products.
The preferred version of a page, declared with a rel="canonical" link, when the same content is reachable at several URLs.
The crawler of Common Crawl, a nonprofit that publishes free web archives. Its data is a common source of LLM training data.
OpenAI’s AI assistant and the most widely used one. It answers from its trained knowledge and, with search, from live web results it cites.
OpenAI’s web browser with ChatGPT built in, launched in October 2025, including an agent mode that can act on web pages.
ChatGPT’s web search capability, launched in October 2024. It retrieves current web pages and shows them as cited sources in the answer.
Product results in ChatGPT, with images, prices, reviews and links to buy, shown for shopping questions. OpenAI says these results are not ads.
The user agent OpenAI uses when ChatGPT visits a page on a user’s behalf, for example to read a link the user shared or during an agent task.
Splitting documents into smaller passages for retrieval. AI systems often retrieve and cite a passage, not a whole page, so clear self-contained sections are easier to use.
The percentage of tracked AI answers that cite a brand’s own pages as sources.
How much the sources an AI engine cites for the same question change from run to run or after model updates.
A web page an AI engine draws on and credits in an answer. Cited sources for a brand’s category often include review sites, publishers, Reddit threads and competitors’ pages, not only the brand’s own site.
Anthropic’s AI assistant. Claude added web search in 2025 and cites the pages it uses.
Anthropic’s crawler for improving search results in Claude. Blocking it can reduce a site’s visibility in Claude’s search answers.
The user agent Anthropic uses when Claude fetches a page in response to a user’s request.
Anthropic’s crawler for collecting content that may be used to train Claude models.
The percentage of people who click a result after seeing it. AI Overviews and other answer features tend to reduce click-through rates for the results below them.
Two brands or pages being referenced together by a third source, which signals that they are related.
A brand appearing near particular topics or competitor names across the web. It helps models associate the brand with a category.
Perplexity’s AI web browser, launched in 2025, with an assistant that can read pages and carry out tasks for the user.
Pages that compare products or providers side by side, such as “X vs Y” or “alternatives to X”. Buyers and AI answers rely on them for commercial questions.
Comparing a brand’s AI visibility, share of voice and citations with named competitors on the same questions and engines.
A gradual decline in a page’s traffic or visibility as information ages or competitors publish better pages.
How current a page’s information is. Keeping facts, prices and examples up to date matters most for time-sensitive topics.
The pace at which a team publishes and updates content. Consistency matters more than volume.
The maximum amount of text, measured in tokens, a model can consider at once, including the question and any retrieved pages.
Searching through a natural-language dialogue with follow-up questions, rather than one keyword query. AI assistants keep context across turns, so later answers depend on earlier ones.
Google’s three page-experience metrics: Largest Contentful Paint (loading), Interaction to Next Paint (responsiveness) and Cumulative Layout Shift (visual stability).
The number of pages a crawler will fetch from a site in a given time. It matters mainly for large sites, where wasted crawls on duplicate or low-value URLs can delay important pages.
How many pages an AI platform crawls for each visit it sends back to the site. Cloudflare began publishing these ratios in 2025 to show how much AI companies take compared with what they return.
How easily search engine and AI crawlers can discover and fetch a site’s pages, through links, sitemaps, fast responses and permissive robots.txt rules.
How much page content moves unexpectedly while loading. A score of 0.1 or less is good.
Visits influenced by AI answers that analytics cannot attribute to them, for example when a user reads an answer and later types the brand’s URL or searches its name.
An assistant mode, offered by ChatGPT, Gemini, Perplexity and others, that browses many sources over several minutes and writes a long, cited report. It reads far more pages per question than a normal answer.
A Chinese AI lab whose open-weight models, including the R1 reasoning model released in January 2025, power its own assistant and are used by other products.
Creating newsworthy stories, data and expert commentary to earn coverage and links from publishers.
A third-party score of a site’s link strength: Moz calls it Domain Authority (DA), Ahrefs Domain Rating (DR) and Semrush Authority Score. They estimate strength; Google does not use them.
The same or very similar content at more than one URL. Search engines pick one version to show, which may not be the one you want.
How long a visitor stays on a page before returning to the search results. Google has not confirmed it as a ranking signal; it is best used as a measure of whether content meets the visitor’s need.
Experience, Expertise, Authoritativeness and Trustworthiness: the qualities Google’s Search Quality Rater Guidelines use to judge content quality. Trust is described as the most important.
Coverage a brand earns rather than pays for or owns: press articles, reviews, analyst reports, community discussion. It makes up much of what AI answers cite.
A numerical representation of text that captures its meaning, so passages with similar meaning sit close together. Retrieval systems use embeddings to find relevant content.
A distinct, identifiable thing, such as a company, person, product or place, that search engines and AI models recognize and connect to facts about it.
Making a brand clearly recognizable as an entity: consistent names and descriptions across the web, structured data with sameAs links, and presence in sources such as Wikidata.
A highlighted answer excerpted from a web page and shown above Google’s organic results.
Further training an existing model on a specific dataset to change its behavior or specialize it.
A search or assistant system that uses a generative AI model to compose its answer, often citing the web sources it used.
The practice of improving how a brand and its content appear in answers generated by AI systems such as ChatGPT, Gemini, Google AI Overviews and Perplexity: whether the brand is mentioned, recommended and cited as a source. The term was introduced in a 2023 research paper of the same name.
Google’s web analytics platform. AI referral traffic can be separated in GA4 by the referring source, such as chatgpt.com.
Google’s free listing for local businesses, shown in Maps and local results, with hours, location, photos and reviews.
Google’s family of AI models and its consumer assistant app. Gemini models also power Google’s AI Overviews and AI Mode in Search.
Google’s platform for merchants to submit product data for Shopping results, ads and AI shopping experiences in Search.
Google’s free tool for site owners, showing indexing status, search queries, impressions, clicks and technical issues.
A robots.txt token, not a separate crawler, that lets sites choose whether Google may use their content to train and ground Gemini models. Google says it does not affect inclusion or ranking in Google Search, including AI Overviews.
Google’s main web crawler. It feeds the index used by Google Search, AI Overviews and AI Mode.
OpenAI’s crawler for collecting content that may be used to train its models. Blocking it does not remove a site from ChatGPT search results.
xAI’s AI assistant, integrated with X (formerly Twitter). It can draw on posts on X as well as the web.
Tying a model’s answer to specific retrieved sources rather than to its memory alone. Grounded answers are what produce citations.
A confident but false statement produced by an AI model, such as a wrong price, feature or fact about a brand.
Help A Reporter Out, a service that connects journalists with expert sources. It launched in 2008, was shut down in December 2024 while operating as Connectively, and was revived under the HARO name by Featured.com in April 2025.
Google’s guidance that content should be written for people first, not to rank. The separate helpful content system was folded into Google’s core ranking systems in March 2024.
An attribute that tells search engines which language and region each version of a page is for.
How many times a page appeared in search results. In Google Search Console, appearances in AI Overviews and AI Mode are counted within the standard Search performance data.
Storing a crawled page in a search engine’s index so it can be returned in results. A page that is not indexed cannot appear in search or in AI answers that draw on that index.
A protocol that lets sites notify participating search engines, including Bing, the moment a URL is added, updated or deleted. Google does not use it.
How much new information a page adds beyond what other pages on the topic already say, such as original data, firsthand experience or a new analysis. A Google patent describes scoring it.
A ChatGPT feature, launched in September 2025, that let users buy from participating merchants inside the conversation. In March 2026 OpenAI let merchants use their own checkout, and its merchant page now says it is moving away from a standalone Instant Checkout experience, with purchases completed on merchant-owned websites or apps.
How quickly a page responds visually to user interactions. It replaced First Input Delay as a Core Web Vital in March 2024; 200 milliseconds or less is good.
Links between pages on the same site. They help crawlers find pages and show how topics relate.
Running a page’s JavaScript to produce its final content. Googlebot renders JavaScript, but many AI crawlers read only the initial HTML, so content or links that appear only after JavaScript runs may be invisible to them.
The format Google recommends for structured data: a script block of JSON that describes the page without changing its visible content.
When several pages on one site target the same query, splitting signals so none performs as well as one strong page would.
Finding the terms and questions people search for, with their volume and intent, to decide what content to create. In AI search it extends to the longer, conversational questions buyers ask assistants.
The date after which a model has no training data. Without search, a model cannot know about anything newer, such as a recent product launch or pricing change.
A database of entities and the relationships between them. Google’s Knowledge Graph supplies facts for knowledge panels and helps it understand what pages are about.
The information box Google shows for a recognized entity, such as a company or person, drawn from its Knowledge Graph.
An AI model trained on very large amounts of text to understand and generate language. GPT, Claude, Gemini and Llama are LLM families; they power AI assistants and AI search.
How long the largest visible element of a page takes to render. Google considers 2.5 seconds or less good.
The rate at which a site gains backlinks over time. Sudden unnatural spikes can look like manipulation.
An article organized as a list, such as “best tools for X”. Lists of options are a frequent source for AI answers to “what should I use” questions.
Optimizing content and brand signals so large language models describe and recommend a brand accurately, both from what they learned in training and from what they retrieve live. Also called LLM SEO.
A companion to llms.txt that contains a site’s documentation or key content in full, as one markdown file, so an AI tool can load it in one request.
A proposed standard, published in September 2024, for a markdown file at /llms.txt that gives AI systems a concise summary of a site and links to its key pages. Adoption by AI providers is still limited and uneven.
Optimizing for location-based searches, such as “near me” queries and map results, through business profiles, reviews and consistent business details.
A specific, lower-volume search phrase, often several words long. AI assistant prompts tend to be longer and more specific than typical search queries.
The percentage of tracked AI answers that mention a brand.
Meta’s AI assistant, available in WhatsApp, Instagram, Facebook and Messenger and as a standalone app.
Meta’s crawler for collecting web content, including for training its AI models.
Microsoft’s AI assistant, built into Bing, Windows, Edge and Microsoft 365. Its web answers draw on the Bing index.
Google’s practice of crawling and evaluating the mobile version of a page for indexing and ranking.
An open standard introduced by Anthropic in November 2024 for connecting AI assistants to external tools and data. A company can offer an MCP server so assistants can query its products or services directly.
Able to work with more than one type of input or output, such as text, images, audio and video.
Keeping a business’s name, address and phone number identical across its website, profiles and directories, so search and AI systems can match them to one entity.
An open project from Microsoft, announced in May 2025, that helps websites offer a natural-language interface to their content for people and AI agents, building on Schema.org data.
A robots meta tag or HTTP header that tells search engines not to include a page in their index.
A robots directive that stops Google from showing a text snippet for a page, including in AI Overviews and AI Mode. The data-nosnippet attribute does the same for part of a page, and max-snippet limits snippet length.
OpenAI’s crawler for ChatGPT search. Sites that block it may not appear as cited sources in ChatGPT search answers.
Data, surveys, tests or analysis a company produces itself. Unique, quotable findings give other publishers and AI answers a reason to reference the source.
What a model learned during training and stores in its weights, as opposed to information it retrieves at answer time. It shapes how a model describes a brand when it does not search.
Writing each section so it stands on its own, with a clear heading and a complete answer, because search and AI systems often rank and cite individual passages rather than whole pages.
A Cloudflare program, announced in July 2025, that lets site owners charge AI crawlers for access, allow them free or block them.
A Google results feature listing related questions with expandable answers. It is a useful source of the questions buyers ask around a topic.
An AI answer engine that searches the web for each question and shows numbered citations to its sources.
The user agent Perplexity uses when it visits a page to answer a user’s question in real time.
Perplexity’s crawler for indexing pages so they can appear and be cited in Perplexity answers.
A comprehensive page on a broad topic that links to more detailed pages on its subtopics.
A structured file of a merchant’s products, prices and availability, submitted to platforms such as Google Merchant Center and to AI shopping programs.
Generating many pages from templates and data, such as one page per location or integration. It works when each page is genuinely useful; otherwise it risks being treated as scaled content abuse.
The question or instruction a user types into an AI assistant. In AI visibility work, prompts are the buyer questions tracked to see which brands the answers mention and recommend.
Writing instructions and context so an AI model produces the desired output.
Running a fixed set of buyer questions through AI engines on a schedule and recording which brands each answer mentions, recommends and cites.
An estimate of how often people ask AI assistants a question or topic. AI platforms do not publish query data, so these figures are modeled estimates.
A technique in which an AI search system splits one question into many related sub-queries, runs them in parallel and combines the results into one answer. Google has described using it in AI Mode, which is why pages can be cited for questions they never target directly.
A model that works through a problem in steps before answering, and may search several times along the way. Examples include OpenAI’s o-series and DeepSeek R1.
An online discussion platform that is frequently cited in AI answers and Google results. Reddit licensed its data to Google and to OpenAI in 2024.
A second step in retrieval that reorders candidate passages by relevance before the model uses them, deciding which sources make it into the answer.
A method in which an AI system first retrieves relevant documents, for example from a web search, and then writes its answer from them. It is how AI search engines produce current, cited answers.
Sites such as G2, Capterra, Trustpilot and Google reviews where customers rate products and providers. They are frequent sources for AI answers to “best” and “alternatives” questions.
Enhanced search listings, such as review stars, product prices or recipes, that Google can show when a page has eligible structured data. Google stopped showing HowTo rich results in September 2023, limited FAQ rich results to authoritative government and health sites the same year, and stopped showing FAQ rich results entirely on May 7, 2026.
A file at a site’s root that tells crawlers which paths they may fetch, with rules per user agent. It is how sites allow or block individual AI crawlers. Compliance is voluntary.
A Schema.org property that links an entity on your site to its official profiles elsewhere, such as LinkedIn or Wikidata, confirming they are the same entity.
A Google spam policy, introduced in March 2024, against producing many pages mainly to manipulate rankings, whether written by people or AI.
The shared vocabulary for structured data, founded in 2011 by Google, Microsoft, Yahoo and Yandex.
Improving a site’s visibility in search engine results through technical health, relevant content and authority. Most AI search systems retrieve from web indexes, so SEO remains the foundation that GEO and AEO build on.
The idea that discovery now happens across many surfaces, including Google, AI assistants, YouTube, TikTok, Reddit, Amazon and app stores, so optimization should cover each place buyers search rather than one search engine.
Google’s 2023 experiment with AI-generated answers in Search, offered through Search Labs. It became AI Overviews in 2024.
The goal behind a query: to learn something, find a specific site, compare options or buy. Content that matches the intent is more likely to rank and to be used in answers.
Search that matches the meaning and intent of a query rather than its exact words.
Optimizing for topics, entities and meaning rather than individual keywords, so content covers a subject the way people and search systems understand it.
Whether AI answers describe a brand positively, neutrally or negatively, and in what terms.
Reviewing server or CDN logs to see which bots, including AI crawlers, request which pages and how often. It is the most direct evidence of how AI systems access a site.
Generating a page’s full HTML on the server before sending it, so crawlers that do not run JavaScript still see the content and links.
A Google spam policy, also called parasite SEO, against publishing third-party content on a reputable site to exploit its ranking signals.
Reviewing which websites AI answers cite for a set of questions, to find the publishers, review sites and communities that shape the answers in a category.
Machine-readable labels added to a page, usually in JSON-LD using the Schema.org vocabulary, that state what the page describes: an organization, product, article, FAQ and so on.
Instructions an AI product gives its model before the user’s message, setting tone, rules and behavior. Users usually cannot see it.
The unit of text a model reads and writes, usually a word or part of a word. Context limits and usage costs are measured in tokens.
A group of related pages organized around a central pillar page and linked to each other, covering a subject in depth.
The depth and breadth with which a site covers a subject, built through many connected, high-quality pages on it.
The text and other content a model learns from, typically including large web crawls such as Common Crawl, books, code and licensed data.
A brand reference on another site without a hyperlink. It still contributes to awareness and to how AI systems associate a brand with a topic.
The name a browser or bot sends with each request to identify itself, such as GPTBot or Googlebot. robots.txt rules and server logs are organized by user agent.
Content created by users rather than brands, such as Reddit threads, forum answers, reviews and videos. AI answers often cite it for real-world opinions.
Finding content by comparing embeddings rather than matching keywords, which surfaces passages that mean the same thing in different words.
A free, structured knowledge base run by the Wikimedia Foundation. Search engines and AI systems use it as a source of facts about entities.
The free encyclopedia and one of the most-used sources in LLM training data and AI answers. Inclusion requires independent notability under its own editorial rules.
A file listing a site’s URLs, often with last-modified dates, to help crawlers find and prioritize pages.
“Your Money or Your Life”: topics that can affect a person’s health, finances, safety or well-being, which Google holds to higher quality standards.
A search where the user’s question is answered on the results page or in the AI answer, so they never click through to a website. AI Overviews and assistants increase the share of zero-click searches.
Run your buyer questions through Genmark AI and see which brands each engine names, cites and recommends.