{"slug":"answer-engine","term":"Answer engine","full":null,"aliases":[],"definition":"An answer engine responds to a query with a composed answer rather than a list of links, drawing on retrieved sources and usually citing some of them. ChatGPT, Perplexity, and Google's AI Overviews and AI Mode are the prominent examples.","body":["The distinction from a search engine is architectural: a search engine ranks documents and hands the reader a choice; an answer engine reads the documents and hands the reader a conclusion. The line blurs in practice — the major search engines now embed answer layers directly in their results, and the standalone assistants increasingly search the live web before answering.","For anyone who publishes on the web, the change is economic before it is technical. Visibility in a ranked list produced clicks; visibility in a composed answer produces citations, brand mentions, and — disputed, and centrally so — some fraction of the clicks it replaced.","The wire covers answer engines as its core beat: what they ship, what they cite, and what they change."],"entities":["chatgpt","perplexity","ai-overviews","ai-mode"],"siblings":["aeo","geo","rag","zero-click-search"],"items":["2026-07-23-ai-mode-and-chatgpt-test-bottom-anchored-citations"],"qa":[],"table":null,"updated":"2026-07-25","url":"https://anythingengineoptimization.com/glossary/answer-engine/"}
{"slug":"aeo","term":"AEO","full":"Answer engine optimization","aliases":["AI engine optimization"],"definition":"AEO — answer engine optimization — is the practice of making content visible, retrievable, and citable inside answer engines: the systems that answer questions directly instead of returning links. It spans machine-readable structure, entity clarity, question-shaped content, and crawler access policy.","body":["In practice AEO work concentrates on four questions. Can the engines read you — crawler access, clean markup, structured data? Can they retrieve you — content organized around the questions people actually ask? Will they trust you — sourcing, authorship, and consistency an engine can verify? And will they cite you — the difference between informing an answer and being credited in one?","The acronym itself is unsettled: AEO standardly expands to answer engine optimization, but AI engine optimization is in genuine circulation for the same practice, and both expansions describe the same work. The discipline keeps renaming itself — SEO begat AEO, GEO, and LLMEO, and the terms are used near-interchangeably, each emphasizing a different corner of the same shift. This publication's name, Anything Engine Optimization, is a position on that churn: the surfaces multiply — search engines, answer engines, shopping engines, agents — and the category is whatever ships next.","Measurement is the discipline's current frontier: citation share is emerging as the visibility metric, and the engines themselves have begun shipping first-party AI-performance data."],"entities":["llms-txt"],"siblings":["geo","llmeo","seo","answer-engine","citation-share","aeo-vs-geo"],"items":["2026-07-20-search-console-adds-generative-ai-performance-reports","2026-07-20-bing-webmaster-tools-previews-ai-visibility-insights-including"],"qa":[{"q":"What is AI engine optimization?","a":"Another name for the same practice as AEO: making content visible, retrievable, and citable inside AI answer surfaces. Answer engine optimization is the standard expansion of AEO; AI engine optimization is in genuine circulation for identical work."},{"q":"What does AEO stand for?","a":"Standardly, answer engine optimization. The expansion AI engine optimization is also in genuine circulation, and both describe the same practice — making content visible, retrievable, and citable in AI answer surfaces."}],"table":null,"updated":"2026-08-11","url":"https://anythingengineoptimization.com/glossary/aeo/"}
{"slug":"geo","term":"GEO","full":"Generative engine optimization","aliases":[],"definition":"GEO — generative engine optimization — is the practice of optimizing content for visibility in generative engines: AI systems that compose answers from retrieved sources. The term entered use through a 2023 research paper of the same name and remains the discipline's preferred name in academic work.","body":["GEO and AEO describe the same territory from different angles — GEO from the research literature, AEO from the practitioner side — and the wire treats them as near-synonyms with different accents.","The research framing is worth knowing because it is unusually honest about the problem's shape: recent survey work describes GEO not as a ranking task but as a stochastic, partially observable pipeline spanning retrieval, citation, prominence, and user behavior. You are optimizing a system you can only see into partway, whose outputs vary between runs.","The literature also studies GEO's adversarial twin: the same techniques that win citations can be used to poison retrieved evidence, and benchmarks now test fact-checking systems against exactly that."],"entities":[],"siblings":["aeo","llmeo","answer-engine","rag"],"items":["2026-07-20-arxiv-survey-maps-generative-engine-optimization-research-2023-2026","2026-07-24-new-benchmark-tests-fact-checkers-against-geo-style-evidence-poisoning"],"qa":[],"table":null,"updated":"2026-07-25","url":"https://anythingengineoptimization.com/glossary/geo/"}
{"slug":"llmeo","term":"LLMEO","full":"Large language model engine optimization","aliases":[],"definition":"LLMEO — large language model engine optimization — is the same discipline as AEO and GEO under yet another name: making content that LLM-based systems retrieve, trust, and cite. It is the least standardized of the acronyms and is often folded into the other two.","body":["The proliferation of names is itself a fact about the field: the practice is young enough that no term has won, and vendors, researchers, and practitioners each coin their own. LLMEO's emphasis falls on the model — the LLM as the thing being optimized for — where AEO emphasizes the product surface and GEO the generative architecture.","The wire's editorial position is to cover the category under all its names and let the market pick the word. The lanes here carry all four acronyms for exactly that reason."],"entities":[],"siblings":["aeo","geo","seo"],"items":[],"qa":[],"table":null,"updated":"2026-07-25","url":"https://anythingengineoptimization.com/glossary/llmeo/"}
{"slug":"seo","term":"SEO","full":"Search engine optimization","aliases":[],"definition":"SEO — search engine optimization — is the three-decade-old practice of earning visibility in ranked search results. It is the parent discipline of AEO and GEO, and its center of gravity is moving into them as search engines embed AI answer layers above the rankings SEO was built for.","body":["What carries over from SEO is the base layer: crawlability, site structure, structured data, and demonstrable authority still decide whether an engine — ranked or generative — can find and trust a site at all.","What changes is the objective. Classic SEO optimized a document's position in a list; the AI surfaces reward being retrieved into an answer and cited by it. Rank was a stable, inspectable outcome; citation is probabilistic and varies with the model underneath — which is why model swaps inside the engines are news for this discipline in a way algorithm updates once were."],"entities":["google","bing"],"siblings":["aeo","geo","zero-click-search"],"items":["2026-07-23-google-search-now-runs-the-latest-gemini-3-5-flash-lite"],"qa":[],"table":null,"updated":"2026-07-25","url":"https://anythingengineoptimization.com/glossary/seo/"}
{"slug":"ai-crawlers","term":"AI crawlers","full":null,"aliases":[],"definition":"AI crawlers are the bots AI companies run to gather web content — for model training, for building search indexes, or for live retrieval when answering a query. Major user-agents include GPTBot and OAI-SearchBot (OpenAI), ClaudeBot and Claude-SearchBot (Anthropic), Google-Extended (Google), and PerplexityBot (Perplexity).","body":["The three purposes matter more than the names, because they carry different trade-offs. Blocking a training crawler withholds content from future models. Blocking an index or retrieval crawler removes a site from the answers an engine gives today — invisibility, not protection. Sites deciding access policy in robots.txt are deciding, user-agent by user-agent, which of those bargains to take.","The user-agent strings themselves are unstable: companies rename crawlers as products rebrand, and a robots.txt rule pointed at yesterday's string silently stops applying. Crawler identity is access policy, which is why renames are news on this wire.","This site's own policy is in its robots.txt: the AI crawlers are welcome here. Cite us; we cite our sources."],"entities":["openai","anthropic","google","perplexity"],"siblings":["rag","aeo"],"items":["2026-07-20-google-renames-notebooklm-crawler-user-agent-to-google-gemininotebook"],"qa":[{"q":"What is an AI crawler?","a":"A bot an AI company runs to gather web content — for model training, for building a search index, or for live retrieval when answering a query. GPTBot, ClaudeBot, Google-Extended, and PerplexityBot are among the major user-agents."},{"q":"Should I block AI crawlers?","a":"The three crawler purposes carry different trade-offs: blocking a training bot withholds content from future models, while blocking an index or retrieval bot removes you from today's answers. The decision is per-bot, not all-or-nothing."}],"table":{"caption":"Named user-agents and robots.txt tokens, by what blocking each one actually costs. Two of these are not crawlers at all — Google-Extended and Applebot-Extended are directives about use, so blocking them withholds content from training without removing you from any answer.","head":["User-agent / token","Operator","Purpose","Blocking it means"],"rows":[["GPTBot","OpenAI","Model training","Out of future training data"],["OAI-SearchBot","OpenAI","Search index","Out of ChatGPT search results"],["ChatGPT-User","OpenAI","Live retrieval","Not fetched when a user asks"],["ClaudeBot","Anthropic","Model training","Out of future training data"],["Claude-SearchBot","Anthropic","Search index","Out of Claude's search results"],["Claude-User","Anthropic","Live retrieval","Not fetched when a user asks"],["Google-Extended","Google","Token, not a crawler — governs Gemini training and grounding","Out of Gemini training and grounding; Search unaffected"],["Google-GeminiNotebook","Google","NotebookLM retrieval","Not readable in NotebookLM"],["PerplexityBot","Perplexity","Search index","Out of Perplexity's index"],["Perplexity-User","Perplexity","Live retrieval","Not fetched when a user asks"],["Applebot-Extended","Apple","Token, not a crawler — governs Apple model training","Out of Apple model training; Siri and Spotlight unaffected"],["Amazonbot","Amazon","Search and assistant","Out of Amazon's assistant answers"],["meta-externalagent","Meta","Model training","Out of Meta model training"],["Bytespider","ByteDance","Model training","Out of ByteDance training data"],["CCBot","Common Crawl","Open dataset","Out of a corpus many models train on"]]},"updated":"2026-08-01","url":"https://anythingengineoptimization.com/glossary/ai-crawlers/"}
{"slug":"citation-share","term":"Citation share","full":null,"aliases":[],"definition":"Citation share is the fraction of an answer engine's citations that go to a given domain or brand across a set of queries — the emerging visibility metric of AI search, where being cited replaces being ranked.","body":["The metric is measured two ways. Third-party tools sample queries at scale and count who the engines cite — the approach behind the published rankings of most-cited domains. And the engines themselves have begun shipping first-party numbers, which replaces estimation with measurement wherever it is available.","Its caveats are structural. Citation share is engine-specific: the same content earns different shares in different engines. It is model-sensitive: results shift when the engine swaps the model underneath. And methodology varies by measurer — query sets, geographies, and counting rules differ, so numbers from different sources rarely reconcile cleanly."],"entities":["ahrefs","semrush","bing"],"siblings":["aeo","geo","zero-click-search","ai-visibility","ghost-citation"],"items":["2026-07-23-ahrefs-ranks-the-50-most-cited-domains-across-four-ai-surfaces","2026-07-20-semrush-study-most-chatgpt-topics-have-no-dominant-brand"],"qa":[{"q":"What is citation share?","a":"Citation share is the fraction of an answer engine's citations that go to a given domain or brand across a set of queries — the emerging visibility metric of AI search."},{"q":"How is citation share measured?","a":"Two ways: third-party tools sample queries at scale and count who the engines cite, and Bing Webmaster Tools reports it first-party in preview. Numbers from different measurers rarely reconcile — methodology, query sets, and the engine's model all move them."},{"q":"What is a citation share metric?","a":"A per-engine percentage: of all sources an AI engine cited for a query set, the share belonging to one domain. It is observational — a citation is not a click."}],"table":null,"updated":"2026-08-01","url":"https://anythingengineoptimization.com/glossary/citation-share/"}
{"slug":"rag","term":"RAG","full":"Retrieval-augmented generation","aliases":[],"definition":"RAG — retrieval-augmented generation — is the architecture behind most answer engines: retrieve documents relevant to the query, then have a language model compose its answer from them, grounding the response in sources it can cite.","body":["The two stages define the optimization problem. Retrieval decides whether content is in the room when the answer is written; generation decides whether it gets quoted and credited. Content that is never retrieved can never be cited, which makes retrievability — clear structure, self-contained passages, unambiguous entities — the precondition for everything else in AEO and GEO.","Grounding is the architecture's promise and its pressure point: an answer assembled from retrieved sources can cite them, but it inherits whatever the retrieval brought in — which is what makes evidence quality, and evidence poisoning, live research problems."],"entities":[],"siblings":["answer-engine","geo","ai-crawlers","grounding"],"items":["2026-07-29-pillarbase-study-15-7-million-ai-mode-citations-show-what-google","2026-07-24-new-benchmark-tests-fact-checkers-against-geo-style-evidence-poisoning","2026-08-01-study-ties-ai-answer-trust-directly-to-the-news-brands-it-cites"],"qa":[],"table":null,"updated":"2026-07-25","url":"https://anythingengineoptimization.com/glossary/rag/"}
{"slug":"zero-click-search","term":"Zero-click search","full":null,"aliases":[],"definition":"A zero-click search ends without a click to any website: the query was answered on the results surface itself. AI answer layers intensify the pattern, which makes citation presence and brand visibility the remaining prize on those queries.","body":["The term predates AI answers — featured snippets and knowledge panels produced zero-click results for years — but composed answers extend it from simple lookups to complex questions that once required visiting several pages.","How much traffic the answer surfaces actually remove is the category's central dispute. The engines assert that AI features still send substantial clicks; third-party studies keep finding click-through declines; and the first-party data that would settle it is published selectively. The wire tracks both sides of that ledger as news."],"entities":["google"],"siblings":["citation-share","answer-engine","seo"],"items":["2026-07-20-google-claims-ai-search-sends-billions-of-clicks-weekly-without-the"],"qa":[],"table":null,"updated":"2026-08-01","url":"https://anythingengineoptimization.com/glossary/zero-click-search/"}
{"slug":"ai-visibility","term":"AI visibility","full":null,"aliases":[],"definition":"AI visibility is how findable, retrievable, and citable a brand or site is across AI surfaces — answer engines, chatbots, and the AI features inside search engines. It is the discipline's umbrella measure: citation share, brand mentions in answers, and AI referral traffic are all ways of counting it.","body":["The metric family splits into first-party and third-party. First-party: the engines' own consoles — Bing Webmaster Tools reports AI citations directly, and Search Console reports impressions inside Google's AI features. Third-party: tools that sample queries at scale and record which brands and domains the engines surface. The two rarely agree precisely, and neither is wrong — they count different windows on the same behavior.","What makes AI visibility harder than classic rank tracking: it is engine-specific, model-sensitive (numbers move when the engine swaps its underlying model), and probabilistic — the same query can produce different answers on different runs. Visibility work in this discipline means moving a distribution, not holding a position."],"entities":["bing","search-console","ahrefs","semrush"],"siblings":["citation-share","aeo","geo"],"items":["2026-07-20-bing-webmaster-tools-previews-ai-visibility-insights-including","2026-07-20-search-console-adds-generative-ai-performance-reports"],"qa":[{"q":"What is AI visibility?","a":"How findable, retrievable, and citable a brand or site is across AI answer surfaces — measured through citation share, brand mentions in answers, and AI referral traffic."},{"q":"How do I check my AI visibility on Bing?","a":"Bing Webmaster Tools' AI Performance report shows citations of your site across Microsoft Copilots and partners first-party, including Citation Share — in preview since June 2026."},{"q":"What does the AI Performance report in Bing Webmaster Tools show?","a":"Clicks and impressions from Microsoft's AI surfaces, extended in preview with Intents, Topics, Citation Share and Compare — Microsoft's move from AI traffic counts toward citation-level data."},{"q":"Does Google Search Console report AI visibility?","a":"Search Console reports impressions and clicks from Google's AI features through its generative-AI performance reports, but does not break out which pages an AI answer cited."}],"table":null,"updated":"2026-08-08","url":"https://anythingengineoptimization.com/glossary/ai-visibility/"}
{"slug":"aeo-vs-geo","term":"AEO vs. GEO vs. LLMEO vs. SEO (2026)","full":null,"aliases":[],"definition":"Four names for substantially the same practice, each emphasizing a different corner: AEO the answer surfaces, GEO the generative architecture (the research literature's term), LLMEO the model doing the answering, and SEO the three-decade-old parent discipline of ranked search. The work — retrievable, trustworthy, citable content — is shared.","body":["The distinctions that hold up: SEO optimizes a document's position in a ranked list, and its center of gravity — crawlability, structure, authority — remains the base layer every other acronym builds on. AEO and GEO both describe optimizing for composed answers; AEO is the practitioner's word, GEO the academy's, entering use through a 2023 research paper. LLMEO is the least standardized and usually folds into the other two.","The honest answer to which term is correct: the market has not decided, and may never — the surfaces keep multiplying faster than the vocabulary settles. This publication's name, Anything Engine Optimization, is a position on exactly that churn."],"entities":[],"siblings":["aeo","geo","llmeo","seo"],"items":["2026-07-20-arxiv-survey-maps-generative-engine-optimization-research-2023-2026"],"qa":[{"q":"Is GEO the same as AEO?","a":"Near enough that the terms are used interchangeably: GEO comes from the research literature, AEO from practice, and both describe making content visible and citable in AI-generated answers."},{"q":"What is the difference between SEO and GEO?","a":"SEO optimizes position in a ranked list of links; GEO optimizes retrieval and citation inside a composed AI answer. SEO's fundamentals remain the base layer GEO builds on."}],"table":null,"updated":"2026-08-03","url":"https://anythingengineoptimization.com/glossary/aeo-vs-geo/"}
{"slug":"llms-txt","term":"llms.txt","full":null,"aliases":[],"definition":"llms.txt is a proposed standard: a plain-text file at a site's root that indexes what the site offers to AI systems, with an optional llms-full.txt carrying the content itself. Google's documentation says no such file is needed to appear in Search or its AI features.","body":["The proposal is cheap by construction — one file, plain text, no server work — and tooling has made it cheaper still: Ahrefs shipped a free generator in July 2026, built on its own analysis of llms.txt adoption across 137,000 domains. Low cost is most of the argument for publishing it.","Google has answered its half of the open question on the record. Its \"AI features and your website\" documentation states that you don't need to create new machine readable files, AI text files, or markup to appear in those features. For Google Search and the AI features inside it, the file does nothing. That is a documented answer, not an inference.","What the record does not contain is the other half. No statement from OpenAI, Anthropic, Perplexity or Microsoft about whether their systems read llms.txt appears on this wire, and no study here ties publishing the file to citations anywhere. The adjacent move is better evidenced: publishers have begun serving agent-readable markdown versions of pages outright, and Time is selling ads inside its own."],"entities":["llms-txt","google","ahrefs"],"siblings":["aeo","ai-crawlers","agentic-search","seo"],"items":["2026-07-24-ahrefs-launches-a-free-llms-txt-generator-tool","2026-08-01-google-guidance-machine-readable-ai-files-do-nothing-for-search","2026-08-01-time-starts-selling-ads-inside-its-ai-agent-readable-markdown-pages"],"qa":[{"q":"Does Google use llms.txt?","a":"No. Google's \"AI features and your website\" documentation states that you don't need to create machine readable files, AI text files, or markup to appear in Search or its AI features."},{"q":"What is an llms.txt file?","a":"A proposed machine-readable index of what a site offers to AI systems, placed at the site root, with llms-full.txt carrying the content itself. Ahrefs publishes a free generator for it, built on an analysis of adoption across 137,000 domains."},{"q":"Do any AI engines read llms.txt?","a":"The record answers only for Google, which says its Search and AI features don't need the file. No statement from OpenAI, Anthropic, Perplexity or Microsoft about whether their systems read it appears on this wire."}],"table":null,"updated":"2026-08-08","url":"https://anythingengineoptimization.com/glossary/llms-txt/"}
{"slug":"ai-overviews","term":"AI Overviews","full":"Google AI Overviews","aliases":["AIO"],"definition":"AI Overviews are Google's AI-generated answers, composed from web sources and shown above the traditional results. They are the largest AI answer surface on the web: a Similarweb analysis put them on 43% of Google searches as of July 2026, up from 15% a year earlier.","body":["Coverage at that scale makes the citation mix consequential. Ahrefs' July 2026 rankings put YouTube first by mention share in AI Overviews at 21.1%, then Reddit (18.5%), Facebook (10.7%), Google itself (7.1%) and Instagram (5.8%) — a concentration in large user-generated and Google-hosted properties. Writesonic measured AI Overviews omitting the source brand in 41% of citations, so citation counts and named visibility are not the same number.","The surface also keeps annexing adjacent real estate. A Top Stories carousel now runs inside AI Overviews, appearing on roughly 15.5% of U.S. and 17.5% of U.K. news searches — which is why Search Console's AI opt-out has a cost analysts flag but Google has not confirmed: excluding content from AI Overviews may exclude it from Top Stories there too. In the UK, the CMA has made an opt-out mandatory, covering training, fine-tuning and grounding alike.","Provenance is a live question. Google's support documentation states that signed-in users' search interactions help improve Search's generative AI models. And when a contested race-based answer surfaced, Google said the system was picking up particular web pages for that specific search — the operator attributing an answer to indexed pages rather than the model."],"entities":["ai-overviews","google","search-console","ahrefs"],"siblings":["ai-mode","answer-engine","citation-share","grounding","zero-click-search","ai-overviews-opt-out"],"items":["2026-07-27-ai-overviews-now-appear-in-43-of-google-searches-up-from-15","2026-07-23-ahrefs-ranks-the-50-most-cited-domains-across-four-ai-surfaces","2026-07-20-top-stories-carousel-now-live-inside-google-ai-overviews","2026-07-28-google-s-ai-opt-out-setting-may-cost-publishers-top-stories-placement","2026-07-24-google-help-page-spells-out-how-signed-in-history-trains-ai-overviews","2026-08-01-reddit-ceo-says-ai-overviews-still-haven-t-found-a-win-win"],"qa":[{"q":"What are Google AI Overviews?","a":"Google's AI-generated answers, composed from web sources and shown above the traditional results, with a selection of those sources linked."},{"q":"How often do AI Overviews appear in Google search results?","a":"In 43% of Google searches as of July 2026, up from 15% a year earlier, per a Similarweb analysis."},{"q":"Which websites do AI Overviews cite most?","a":"By mention share in Ahrefs' July 2026 rankings: YouTube (21.1%), Reddit (18.5%), Facebook (10.7%), Google itself (7.1%) and Instagram (5.8%)."},{"q":"Can publishers opt out of AI Overviews?","a":"Search Console carries an AI opt-out, and the UK's CMA has ordered Google to let publishers exclude content from AI training, fine-tuning and grounding. One analysis finds using the opt-out may also cost Top Stories placement inside AI Overviews; Google has not confirmed the link."}],"table":null,"updated":"2026-08-08","url":"https://anythingengineoptimization.com/glossary/ai-overviews/"}
{"slug":"ai-mode","term":"AI Mode","full":"Google AI Mode","aliases":[],"definition":"AI Mode is Google's conversational search surface: a full AI-generated answer with citations, ads and app integrations, running alongside classic results. Datos and SparkToro measured it at 0.13% of U.S. search-related visits in Q2 2026 — heavily instrumented, lightly used.","body":["The gap between instrumentation and use is the surface's defining fact. Similarweb counted AI Mode visits rising from 126 million in June 2025 to 279 million in May 2026; Datos and SparkToro put it at 0.13% of U.S. search-related visits (0.29% in the EU/UK) and describe U.S. usage as plateaued or dipped. Underneath it, Google says Gemini 3.5 Flash-Lite is rolling out in Search — and the answer layer keeps becoming an action layer, with Instacart, Canva and YouTube Music connected to it.","What it cites is unusually well documented. Google itself is the second most-cited domain in AI Mode answers, on Google-hosted Business Profile and Product Knowledge Panel cards. A Pillarbase analysis of 15.7 million citations found the passages reused most average 117 words, answer in the first sentence about 80% of the time, are self-contained about 85% of the time, and come 76% of the time from pages already ranking first organically. Writesonic measured AI Mode omitting the source brand in 49% of citations.","The mix moves. Over the 30 days ending July 15, 2026, AI Mode's publisher mentions fell nearly in half while mentions of retailers rose sharply across 15 of 20 verticals, per Kevin Indig's Growth Memo analysis — even as AI Mode's total mention volume contracted 12.9%. Ads, meanwhile, appear on roughly one in three commercial-keyword queries, with advertisers rarely among the cited sources."],"entities":["ai-mode","google","gemini"],"siblings":["ai-overviews","answer-engine","citation-share","agentic-search","grounding","ai-overviews-opt-out"],"items":["2026-08-03-zero-click-google-searches-hit-a-record-high-ai-mode-use-stays-flat","2026-07-23-google-is-ai-mode-s-second-most-cited-domain-new-data-shows","2026-07-29-pillarbase-study-15-7-million-ai-mode-citations-show-what-google","2026-07-23-ads-appear-on-a-third-of-commercial-ai-mode-queries-analysis-finds","2026-07-20-google-ai-mode-adds-instacart-canva-and-youtube-music-integrations"],"qa":[{"q":"What is Google AI Mode?","a":"Google's conversational search surface: a full AI-generated answer with citations, ads and connected-app integrations, running alongside classic search results."},{"q":"How many people use Google AI Mode?","a":"Datos and SparkToro put it at 0.13% of U.S. search-related visits in Q2 2026 (0.29% in the EU/UK), with U.S. usage plateaued or dipped. Similarweb separately counted AI Mode visits rising from 126 million in June 2025 to 279 million in May 2026."},{"q":"What kind of content does AI Mode cite repeatedly?","a":"Per a Pillarbase study of 15.7 million citations, the most-reused passages average 117 words, answer in the first sentence about 80% of the time, and are self-contained about 85% of the time; 76% of passages reused 100 or more times came from pages already ranking first organically."}],"table":null,"updated":"2026-08-08","url":"https://anythingengineoptimization.com/glossary/ai-mode/"}
{"slug":"agentic-search","term":"Agentic search","full":null,"aliases":["Agentic browsing"],"definition":"Agentic search is the step past answering: an AI system that browses, acts and transacts on a user's behalf rather than returning an answer for the user to act on. It reads a site as an agent, not a reader, which makes machine readability, actionability and access policy the optimization surface.","body":["The record shows the capability arriving across every major engine at once. Google says Gemini 3.5 Flash-Lite is rolling out in Search to power agentic search experiences; Gemini Spark's Chrome integration uses saved passwords and logged-in accounts to research and begin bookings — flights, apartment viewings — before handing payment steps back to the user, now in more than 160 additional countries but not the EEA, UK, Switzerland or Nigeria. Perplexity ships an agentic browser, Comet. OpenAI is testing a ChatGPT ad format where the click opens a Business Agent conversation instead of the advertiser's website.","The publishing side has a vocabulary forming around it. A July 2026 arXiv paper proposes three design principles for agent-ready websites — machine readability, actionability, decision reliability. Time has converted pages to stripped-down markdown so agents can read them, and sells ads inside those pages. Cloudflare's Agent Readiness tool audits the same territory: robots.txt and sitemap rules, content machine-readability, and agent-callable APIs.","Access is the unresolved half. IAB Australia's crawler matrix treats live AI agents as their own category, subject to the same four verdicts as any other crawler — allow, allow with conditions, require licensing, or block. And the Ninth Circuit has given the question its first appellate test, reversing an injunction that barred Comet from Amazon.com on the ground that the user, not Perplexity, directs the Assistant's access."],"entities":["gemini","google","openai","chatgpt","perplexity"],"siblings":["answer-engine","ai-crawlers","llms-txt","aeo","ai-mode"],"items":["2026-07-20-researchers-propose-framework-for-agent-ready-websites","2026-08-01-gemini-spark-s-chrome-shopping-agent-expands-to-160-countries-skips","2026-07-31-openai-tests-chatgpt-ad-format-that-opens-an-agent-chat-not-a-website","2026-08-01-time-starts-selling-ads-inside-its-ai-agent-readable-markdown-pages","2026-08-05-9th-circuit-reverses-perplexity-cfaa-injunction-in-amazon-suit"],"qa":[{"q":"What is agentic search?","a":"An AI system that browses, acts and transacts on a user's behalf rather than returning an answer the user acts on — reading sites as an agent rather than as a reader."},{"q":"What makes a website agent-ready?","a":"A 2026 arXiv paper proposes three design principles: machine readability, actionability, and decision reliability. Cloudflare's Agent Readiness tool audits the same ground — robots.txt and sitemap rules, content machine-readability, and agent-callable APIs."},{"q":"Are AI shopping agents allowed to use my site?","a":"Unsettled. The Ninth Circuit reversed an injunction barring Perplexity's Comet browser from Amazon.com, holding the user directs the Assistant's access, so Perplexity is unlikely to violate the Computer Fraud and Abuse Act. IAB Australia's guidance sorts live AI agents into allow, allow with conditions, require licensing, or block."}],"table":null,"updated":"2026-08-08","url":"https://anythingengineoptimization.com/glossary/agentic-search/"}
{"slug":"grounding","term":"Grounding","full":null,"aliases":["Grounded generation"],"definition":"Grounding is the step that ties a generated answer to retrieved source documents, so the response reflects — and can cite — specific material rather than the model's parameters alone. The UK's CMA treats it as a distinct use of publisher content, separate from training and fine-tuning.","body":["Grounding is what makes citation possible, and it is also what makes an answer inherit whatever the retrieval brought in. Google demonstrated both halves in one statement: asked about an AI Overview that answered near-identical queries differently, Google said its system was picking up particular web pages for that specific search and that the behavior was not present for all queries — the engine locating the answer in grounded sources rather than in the model. The failure mode has a research literature: the GPE benchmark simulates generative-engine-optimization attacks by controlling evidence sources and poisoning ratios, and finds verification systems degrade under poisoning in ways clean-data evaluation never reveals.","The regulatory record is what turned grounding from an architecture term into a controllable use. The UK CMA's Publisher Conduct Requirement forces Google to let publishers opt out of having content used for AI training, fine-tuning, and the grounding behind AI Overviews and AI Mode answers — three separate permissions, named separately. Google must also add clearer attribution links and supply publishers with impressions, clicks and click-through rate through a platform such as Search Console.","Grounding is where the visibility metrics are counted. Bing Webmaster Tools' AI Performance reports added Citation Share in preview in June 2026, and grounded answers are the only answers with citations to share. There is a demand-side reason to care which sources ground an answer: an AOP/Ipsos survey of 1,000 UK adults found trust in an AI answer ran above 90% when it cited a fully trusted news source, and about 25% for a neutral one."],"entities":["google","ai-overviews","ai-mode","bing"],"siblings":["rag","citation-share","answer-engine","ai-visibility"],"items":["2026-08-01-uk-s-cma-orders-google-to-give-publishers-ai-opt-out-and-attribution","2026-07-30-google-says-ai-overview-s-race-based-answer-came-from-source-pages","2026-07-24-new-benchmark-tests-fact-checkers-against-geo-style-evidence-poisoning","2026-07-20-bing-webmaster-tools-previews-ai-visibility-insights-including","2026-08-01-study-ties-ai-answer-trust-directly-to-the-news-brands-it-cites"],"qa":[{"q":"What does grounding mean in AI search?","a":"Composing an answer from retrieved source documents rather than from the model's parameters alone — the step that lets an answer cite specific pages, and that makes the answer only as good as the sources retrieved."},{"q":"What is a grounding query?","a":"A query the engine answered by retrieving and citing sources — the unit AI-visibility reporting counts citation share against. Bing Webmaster Tools added Citation Share to its AI Performance reports in preview in June 2026."},{"q":"Can publishers opt out of having their content used for grounding?","a":"In the UK, yes. The CMA's Publisher Conduct Requirement requires Google to let publishers exclude content from AI training, fine-tuning, and the grounding behind AI Overviews and AI Mode."}],"table":null,"updated":"2026-08-08","url":"https://anythingengineoptimization.com/glossary/grounding/"}
{"slug":"query-fan-out","term":"Query fan-out","full":null,"aliases":["Fan-out queries","Query expansion"],"definition":"Query fan-out is the step where an answer engine turns one user question into its own set of background search queries and retrieves against those, rather than against the words the user typed. What the engine puts in those queries shapes the answer more than the pages it later fetches.","body":["Fan-out sits between the question and the retrieval, which is why it is the hardest layer in this discipline to influence and the least measured. A publisher can control what a page says and whether a crawler can read it. Nobody outside the engine controls which queries the engine decides to run.","The published measurements are all on ChatGPT, and they are consistent about the size of the effect. An independent analysis of raw traffic from a ChatGPT Plus account found the model injecting brand names the user never typed into its own background queries in 21 of 27 tested conversations. Brands ChatGPT named in its own query reached the final answer 68.9% of the time, against 2.1% for brands that appeared only on retrieved pages — a roughly 33x gap.","The same analysis is the clearest published answer to a question practitioners ask constantly: how much retrieval converts to citation. Just 110 of 3,554 retrieved pages — 3.1% — earned a citation at all, and the rate fell from 5.2% at the first-listed domain to 0.3% from the sixth position on. Being fetched is not being cited.","Fan-out rules change without announcement, and the change is visible in the citation data before anyone explains it. Promptwatch attributed the August 2026 collapse in Reddit's ChatGPT citations to an August 8 shift in how the engine assembles these queries: use of the `site:` operator in fan-out queries rose from 0.37% to 16.8%, roughly forty-six-fold, which it reads as the engine asking named sites directly rather than searching the open web. Reddit's share of ChatGPT citations fell 86.4% in four days without Reddit changing anything.","Google describes a fan-out technique behind AI Mode, and the term is in common practitioner use for it. No comparable measurement of Google's query-writing step has been published, so the numbers above should not be carried across to AI Overviews or AI Mode — the mechanism is named on both sides, the evidence is not."],"entities":["chatgpt","openai","ai-mode"],"siblings":["grounding","rag","ai-mode","citation-share","ai-visibility"],"items":["2026-08-30-query-fan-out-decides-ai-citation-before-a-page-is-ever-read","2026-08-18-chatgpt-s-own-search-queries-name-brands-users-never-typed","2026-08-27-reddit-s-citation-drop-is-chatgpt-only-86-there-11-on-ai-overviews","2026-07-30-study-finds-ai-models-search-more-often-for-brands-they-already-know"],"qa":[{"q":"What is query fan-out?","a":"The step where an answer engine turns one user question into its own set of background search queries and retrieves against those instead of the words the user typed. Measurements published so far are all on ChatGPT, which injected brand names the user never typed into its own queries in 21 of 27 tested conversations."},{"q":"Why do AI engines search for things I didn't ask?","a":"Because the engine reformulates the question before retrieving, adding terms and brand names it associates with the topic. That reformulation is where much of the outcome is set: brands ChatGPT named in its own query reached the final answer 68.9% of the time versus 2.1% for brands found only on retrieved pages."},{"q":"What percentage of retrieved pages actually get cited?","a":"3.1% in the one published measurement — 110 of 3,554 pages across 27 ChatGPT conversations — falling from 5.2% at the first-listed domain to 0.3% from the sixth position on. Retrieval and citation are different outcomes and the gap between them is large."},{"q":"Can you optimize for query fan-out?","a":"Not directly, and nothing published shows a page-level lever that reaches it. The adjacent finding is that models search for brands they already know 3.2 times more often than unfamiliar ones, which points at brand familiarity rather than on-page work — suggestive, and not the same measurement."}],"table":null,"updated":"2026-08-30","url":"https://anythingengineoptimization.com/glossary/query-fan-out/"}
{"slug":"citation-drift","term":"Citation drift","full":null,"aliases":["Citation decay","Source rotation"],"definition":"Citation drift is the rotation of which sources an AI engine cites for the same prompt over time, with the prompt and the page both unchanged. It is not malfunction — it is how retrieval-based answers behave, and it is why AI visibility is a distribution to move rather than a position to hold.","body":["The largest public measurement re-sampled 82,619 prompts weekly for 17 weeks — December 17, 2025 to April 8, 2026 — producing 1,548,213 snapshots across six countries and three surfaces. It found Google replacing 56% of the sources in its AI answers every week and ChatGPT Search as much as 74%.","The structure matters more than the headline rate. For 86% of prompts there is a stable core of a few domains, with everything outside that core rotating at 89% per week. On Google AI Overviews specifically, 53% of prompts saw no source change at all across the full 17 weeks. Aggregate churn and per-prompt permanence are both real: they describe the rotating tail and the persistent core of the same answer.","Two consequences follow directly. AI Overviews and AI Mode cite different domains 83% of the time, so a citation-share figure measured on one Google surface is not a reading of the other. And only 1.4% of cited news articles stayed in the citation set permanently, which makes an AI citation behave, for a publisher, less like an asset that accrues than a placement that expires. Brands fare better: 43% of brand queries kept the brand's own domain present in all 17 weeks, while the co-citations beside it rotated at 70% per week.","Drift has more than one cause, and they are worth separating because only some are addressable. The engine's preferences can shift — AI Mode's publisher mentions fell nearly by half in thirty days while retailer mentions tripled across 15 of 20 verticals. Its retrieval method can change — Reddit's 86.4% four-day fall in ChatGPT citations traced to a change in how the engine assembles its background queries. And the model underneath can be swapped at any time, which the glossary's citation-share entry has flagged since it was written. None of those is something a page can be optimized against.","Measurement is arriving. Profound shipped a feature called Citation Decay in August 2026 that tracks week-over-week counts for every cited URL rather than reporting a point-in-time total — the longitudinal view this research argues is the only honest one."],"entities":["chatgpt","ai-overviews","ai-mode"],"siblings":["citation-share","ai-visibility","query-fan-out","grounding","ghost-citation"],"items":["2026-09-01-sistrix-chatgpt-replaces-74-of-its-cited-sources-every-week","2026-08-16-profound-adds-citation-lifespan-tracking","2026-08-06-ai-mode-citation-mix-shifts-from-publishers-toward-retail-products","2026-08-27-reddit-s-citation-drop-is-chatgpt-only-86-there-11-on-ai-overviews"],"qa":[{"q":"What is citation drift?","a":"The rotation of which sources an AI engine cites for the same prompt over time, with neither the prompt nor the content changing. SISTRIX measured Google replacing 56% of its cited sources weekly and ChatGPT as much as 74%."},{"q":"How stable are AI search citations?","a":"Two things at once. 86% of prompts have a stable core of a few domains, while everything outside that core rotates at 89% per week — and on AI Overviews, 53% of prompts saw no source change across 17 weeks. Aggregate churn and per-prompt permanence are both true of the same answer."},{"q":"Do Google AI Overviews and AI Mode cite the same sources?","a":"Mostly not — they cite different domains 83% of the time. A visibility number measured on one is not a reading of the other, even though both are Google."},{"q":"How long does an AI citation last?","a":"For news, usually not long: only 1.4% of cited news articles remained permanently across 17 weeks. Brand domains persist better, present in all 17 weeks for 43% of brand queries, though the sources cited alongside them rotated at 70% per week."},{"q":"What is citation decay?","a":"The same phenomenon seen from one page's side: the week-over-week fall in how often a URL is cited after it enters an AI answer. Profound ships it as a report named Citation Decay, launched August 13, 2026, which tracks week-over-week citation counts for every cited URL instead of a point-in-time total. SISTRIX's figures are the base rate it is measured against — 56% of Google's cited sources and up to 74% of ChatGPT's replaced each week."}],"table":{"caption":"Citation drift over 17 weeks — SISTRIX, 82,619 prompts, 1,548,213 snapshots, six countries, published 2026-05-01. Vendor-published and not replicated.","head":["Measure","Finding"],"rows":[["Weekly source replacement — ChatGPT Search","74%"],["Weekly source replacement — Google","56%"],["Prompts with a stable core of a few domains","86%"],["Weekly rotation outside that core","89%"],["AI Overviews prompts with no source change in 17 weeks","53%"],["AI Overviews and AI Mode citing different domains","83%"],["Cited news articles that stayed permanently","1.4%"],["Brand queries keeping the brand's own domain all 17 weeks","43%"]]},"updated":"2026-09-11","url":"https://anythingengineoptimization.com/glossary/citation-drift/"}
{"slug":"coati","term":"Coati","full":null,"aliases":[],"definition":"Coati is a name Google's VP of Search used once, at a conference in November 2022, describing it as an update that replaced the Panda algorithm. Google has published nothing about it since, and it appears in no Google documentation — including the ranking-systems guide Google maintains for exactly this purpose.","body":["Everything published about Coati traces to a single source: Hyung-Jin Kim, Google's VP of Search, speaking at the SMX Next keynote on November 15, 2022. Kim said “Panda has been subsequently replaced by Coati,” and described it as an update to Panda. It was a remark in a keynote, not an announcement, and it remains the entire primary record.","Google's own “guide to Google Search ranking systems” is the page where Google lists what it is willing to name — 17 active systems including BERT, RankBrain, MUM, neural matching, the passage ranking system and the reviews system, plus four retired ones. It was last updated 2025-12-10. Coati appears in neither list, and no other Google documentation contains the word.","Panda is on that page. Google describes it as a system for surfacing high-quality original content, announced in 2011, which “evolved and became part of our core ranking systems in 2015.” That is a different account from Kim's. The written record says Panda was absorbed into core ranking and names no successor; the spoken record says Panda was replaced by something called Coati. Both come from Google. Only one of them is maintained.","None of this makes Coati fictional. Kim was speaking about systems he oversees, and a company is under no obligation to document every internal name. The precise statement is narrower and more useful: Coati has been named by Google once, out loud, four years ago, and never in writing.","The reason it belongs on this beat rather than in an SEO history is what those conditions do to an answer engine. A plausible name, an authoritative-sounding origin, years of third-party explainers describing its behaviour and offering optimisation advice for it, and nothing authoritative to contradict them: an engine synthesising the open web finds near-unanimity and no dissent. Coati is a live test of what answer engines do with a term that has heavy secondary coverage and no primary source."],"entities":["google"],"siblings":["seo","answer-engine","ai-overviews"],"items":["2026-09-01-google-coati-rests-on-one-2022-sentence-and-appears-in-no-google"],"qa":[{"q":"What is the Google Coati algorithm?","a":"A name Google's VP of Search used once, at a conference, in 2022. Hyung-Jin Kim told the SMX Next keynote on November 15, 2022 that “Panda has been subsequently replaced by Coati,” describing Coati as an update to Panda. Google has published nothing about it since."},{"q":"Is Coati in Google's official ranking systems documentation?","a":"No. Google's own guide to Google Search ranking systems, last updated 2025-12-10, names 17 active systems and 4 retired ones. Coati is not among them, and no other Google documentation names it."},{"q":"Did Coati replace Panda?","a":"Google's spoken and written accounts differ. Kim said in 2022 that Panda was replaced by Coati. Google's ranking-systems guide says Panda was announced in 2011 and evolved into the core ranking systems in 2015 — absorbed, with no successor named."}],"table":null,"updated":"2026-09-06","url":"https://anythingengineoptimization.com/glossary/coati/"}
{"slug":"google-discover","term":"Google Discover","full":null,"aliases":["Discover feed"],"definition":"Google Discover is a personalised feed Google surfaces without a query, in the Google app and on Android home screens. For news publishers it now supplies more Google traffic than Search does: 67.51% against 27% in Q4 2025, across the 400+ publishers NewzDash tracks.","body":["The structural fact about Discover is that it has no queries in it. There is no keyword to research, no position to hold, and no query report to read. Traffic arrives because the feed decided to show something, not because anyone asked for it — which makes most of the search-optimisation toolkit inapplicable to the surface that now carries most of the traffic.","The reversal is recent. In 2023 Web Search sent 51% of news publishers' Google traffic and Discover sent 37%. By Q4 2025 those had crossed and separated: Discover 67.51%, Web Search 27%. NewzDash CEO John Shehata calls it “the Great Flip.”","Exposure to AI answers is a property of the beat rather than of news as a category, and this is the detail most often lost. In US data Shehata reports AI Overviews triggering on 3–5% of breaking and political news queries, around 18% of sports queries, and 40–50% of health queries. A single figure for “news” averages a political desk and a health desk into something true of neither.","What that leaves the practice called editorial SEO is not post-publication optimisation, which the economics no longer support at scale, but decisions made where a story is briefed: what the headline says, how the piece is structured, which surface it is built for. The name survives from a period when Search was the channel; the work has moved.","Treat the split as vendor-published and directionally reliable rather than settled. NewzDash sells news-SEO tooling and published the analysis itself, the traffic figures are Q4 2025, and the interview they come from was conducted in January 2026 and published 2026-08-31. Nobody has replicated the split independently. The direction is not in dispute — Search Console has reported Discover separately since it began to matter, and publisher-side accounts describe the same movement."],"entities":["google","search-console"],"siblings":["seo","zero-click-search","ai-overviews"],"items":["2026-09-01-google-discover-now-sends-news-publishers-more-than-twice-what-search"],"qa":[{"q":"Does Google Discover send more traffic than Search to news publishers?","a":"Yes, by more than two to one. Across 400+ news publishers NewzDash tracked, Discover supplied 67.51% of Google traffic in Q4 2025 against 27% for Web Search — a reversal from 2023, when Search sent 51% and Discover 37%."},{"q":"How do you optimise for Google Discover?","a":"Not with keywords, because Discover has no queries in it. There is no position to hold and no query report to read, so the levers are the ones set at the story-brief stage — headline, structure, and which surface the piece is built for."},{"q":"How often do AI Overviews appear on news queries?","a":"It depends heavily on the section. In US data John Shehata reports breaking and political news triggering AI Overviews on 3–5% of queries, sports around 18%, and health 40–50%. Exposure is a property of the beat, not of news as a category."}],"table":{"caption":"Share of news publishers' Google traffic by surface — NewzDash, 400+ publishers","head":["Surface","2023","Q4 2025"],"rows":[["Google Discover","37%","67.51%"],["Web Search","51%","27%"]]},"updated":"2026-09-06","url":"https://anythingengineoptimization.com/glossary/google-discover/"}
{"slug":"bing-ai-performance","term":"AI Performance (Bing Webmaster Tools)","full":null,"aliases":["Bing AI Performance report","Bing AI visibility insights"],"definition":"AI Performance is the Bing Webmaster Tools report that shows how a site's content appears in AI-generated answers across Microsoft Copilot, Bing and select partner AI experiences. Since June 2026 it has carried four preview capabilities — Intents, Topics, Citation Share and Compare — which make it the only first-party report on this record that measures citations rather than impressions.","body":["Microsoft announced the four capabilities on June 16, 2026, in preview. Intents classifies the grounding queries behind AI answers into categories such as Informational, Commercial, Navigational, Learn and Solve, Research, Creation and Local. Topics groups related grounding queries into thematic clusters. Citation Share reports the percentage of citations attributed to a site out of all citations shown across all sites for the same grounding query. Compare overlays a previous period on the current view so citation activity can be read over time.","The report is organised around grounding queries — the searches the engine runs in the background to assemble an answer — rather than the prompts people type. That is why a site can hold citation share for questions it never ranked for, and why the Intents and Topics views exist: they describe demand at the layer where the engine actually retrieves.","The contrast with Google's console is the reason the report matters. Search Console's generative-AI performance report, introduced in June 2026 and available to all sites since August 31, shows impressions across AI Overviews and AI Mode without click data, and names no query. Bing's report names the query, its intent and the site's share of citations for it. Third-party tools estimate all three by sampling prompts at scale; this is the one place an engine publishes them itself.","What the record does not hold: any comparison of Bing's first-party numbers with third-party estimates for the same site, any measure of Bing's own share of AI answers, and any first-party equivalent from OpenAI, Anthropic or Perplexity."],"entities":["bing","search-console"],"siblings":["citation-share","ai-visibility","grounding","query-fan-out"],"items":["2026-07-20-bing-webmaster-tools-previews-ai-visibility-insights-including","2026-08-31-search-console-s-ai-overviews-opt-out-control-rolls-out-to-all-sites","2026-07-20-search-console-adds-generative-ai-performance-reports"],"qa":[{"q":"What is AI Performance in Bing Webmaster Tools?","a":"The report that shows how a site's content appears in AI-generated answers across Microsoft Copilot, Bing and select partner AI experiences. On June 16, 2026 Microsoft added Intents, Topics, Citation Share and Compare to it, in preview."},{"q":"Does Bing Webmaster Tools show AI citations?","a":"Yes — it is the only first-party console on this record that does. Citation Share reports the percentage of citations a site receives out of all citations shown for the same grounding query. Search Console's generative-AI report gives impressions across AI Overviews and AI Mode without click data and without the query."},{"q":"What is a grounding query?","a":"The search an AI engine runs in the background to assemble an answer, as distinct from the prompt a person typed. Bing's AI Performance report classifies grounding queries by intent and groups them by topic; citation share is calculated per grounding query."}],"table":{"caption":"The four AI Performance capabilities — Microsoft, June 16, 2026, in preview.","head":["Capability","What it reports"],"rows":[["Intents","Grounding queries classified into categories: Informational, Commercial, Navigational, Learn and Solve, Research, Creation, Local, and more"],["Topics","Related grounding queries grouped into thematic clusters"],["Citation Share","Citations attributed to the site as a percentage of all citations shown across all sites for the same grounding query"],["Compare","A previous period overlaid on the current reporting view"]]},"updated":"2026-09-11","url":"https://anythingengineoptimization.com/glossary/bing-ai-performance/"}
{"slug":"ghost-citation","term":"Ghost citation","full":null,"aliases":["Uncited use","Unattributed citation"],"definition":"A ghost citation is an AI answer that draws on a source's content without naming the brand or site behind it. The material is used; the credit, and any link, is missing. The term comes from a Writesonic study of roughly 16 million brand appearances across seven answer engines over 30 days, which found the practice on every one of them and most often on Perplexity.","body":["Writesonic's measurement put the ghost-citation rate at 52% on Perplexity, 49% on Google AI Mode, 41% on Google AI Overviews, 37% on ChatGPT, 25% on Gemini, 22% on Grok and 19% on Microsoft Copilot. Two caveats travel with those numbers: the study's author is Writesonic's chief executive, and the company describes the figures as a snapshot that establishes nothing about clicks, brand recall or conversions.","The measure matters because AI-visibility tooling counts citations — a source the engine names or links — and a ghost citation is by definition absent from that count. A brand can be doing the work of the answer and register as invisible, which makes any citation-share figure a floor on how much an engine uses a source rather than a measure of it. It is also why crawl-to-referral ratios, which are observable from a server log, are a useful second instrument: the crawl is visible even when the credit is not.","The rates are model settings, not fixed properties of an engine. In September 2026, SEOs reported that the day-old Gemini 3.8 Flash showed almost no sources at all in AI Mode — the Google surface with the higher ghost rate — a reminder that attribution behaviour changes when the model does, and that a rate measured in July describes July."],"entities":["perplexity","ai-mode","ai-overviews","chatgpt"],"siblings":["citation-share","citation-drift","ai-visibility","grounding"],"items":["2026-07-29-writesonic-study-finds-ai-engines-often-skip-naming-source-brands","2026-09-03-gemini-3-8-flash-cites-fewer-sources-in-google-ai-mode","2026-08-11-study-ai-assistants-favor-brands-with-better-sounding-names"],"qa":[{"q":"What is a ghost citation?","a":"An AI answer that uses a source's content without naming the brand or site it came from. Writesonic coined the term in a July 2026 study of about 16 million brand appearances across seven engines."},{"q":"Which AI engine has the most ghost citations?","a":"Perplexity, at 52% of the appearances Writesonic measured, followed by Google AI Mode at 49% and AI Overviews at 41%. Microsoft Copilot had the fewest at 19%. One vendor's 30-day snapshot; on this record no second measurer has published a comparable rate."},{"q":"Do ghost citations show up in AI visibility tools?","a":"Not as citations — the tools count sources the engine names or links, and a ghost citation is neither. Treat a citation-share figure as a floor on how much an engine uses a source, not a measure of it."}],"table":{"caption":"Ghost-citation rate by engine — Writesonic, roughly 16 million brand appearances over 30 days, reported 2026-07-29. Vendor-published, authored by the vendor's CEO, not replicated.","head":["Engine","Answers using a source without naming it"],"rows":[["Perplexity","52%"],["Google AI Mode","49%"],["Google AI Overviews","41%"],["ChatGPT","37%"],["Gemini","25%"],["Grok","22%"],["Microsoft Copilot","19%"]]},"updated":"2026-09-12","url":"https://anythingengineoptimization.com/glossary/ghost-citation/"}
{"slug":"ai-overviews-opt-out","term":"AI Overviews opt-out","full":null,"aliases":["Search generative AI control","AI features opt-out"],"definition":"The AI Overviews opt-out is a Google Search Console setting — Google's name for it is the Search generative AI control — that removes a site from AI Overviews, AI Mode and the generative features in Discover without touching its ordinary search results. Google's documentation says the setting is not used as a ranking or inclusion signal elsewhere in Search, that exclusion takes one to two days, and that a site which opts out receives no traffic or impressions from those features.","body":["It is the only control that separates the AI features from Search proper. Before it, every lever took something else with it: nosnippet, data-nosnippet and max-snippet limit what Search may display but shorten the ordinary result snippet too; noindex removes the page from Search altogether; Google-Extended governs training and grounding for Gemini, not appearance in Search's own AI features.","Google shipped it to a subset of Search Console accounts first — this wire found it on August 27, 2026 — and its documentation recorded a rollout to all websites worldwide on August 31. Search Console's separate AI performance reports, which show impressions in AI Overviews and AI Mode without click data, reached global availability the same day.","The price is stated plainly and is not small: no traffic or impressions from features that appeared on 43% of Google searches in July 2026, up from 15% a year earlier, per Similarweb. There is also an unconfirmed second cost. Google renders Top Stories carousels inside AI Overviews on roughly 15.5% of U.S. and 17.5% of U.K. news searches, per an analysis by NewzDash founder John Shehata, so a news publisher who opts out may lose the Top Stories placement drawn within them. Shehata called the link high-confidence; Google has not addressed it.","Two things it does not do. It does not cover training: Google's documentation sends publishers to Google-Extended for that, which makes the control a partial answer to the UK Competition and Markets Authority's Publisher Conduct Requirement of June 3, 2026 — a binding order that Google let publishers opt out of content used for training, fine-tuning and the grounding behind AI Overviews and AI Mode, with nine months to comply. And Google's own guidance page for site owners, \"AI features and your website,\" did not mention it: checked August 27, that page named nosnippet 18 times and the control not once."],"entities":["ai-overviews","ai-mode","search-console"],"siblings":["ai-overviews","ai-mode","google-discover","ai-crawlers","llms-txt"],"items":["2026-08-27-what-google-s-ai-overviews-opt-out-does-and-two-things-it-does-not","2026-08-31-search-console-s-ai-overviews-opt-out-control-rolls-out-to-all-sites","2026-07-28-google-s-ai-opt-out-setting-may-cost-publishers-top-stories-placement","2026-08-01-uk-s-cma-orders-google-to-give-publishers-ai-opt-out-and-attribution"],"qa":[{"q":"How do I opt out of Google AI Overviews?","a":"In Google Search Console, under Settings, the Search generative AI control removes a site from AI Overviews, AI Mode and Discover's generative features. Google says exclusion takes one to two days. It rolled out to all sites worldwide on August 31, 2026."},{"q":"Does opting out of AI Overviews hurt rankings?","a":"Google's documentation says the control is not used as a ranking or inclusion signal affecting other parts of Search. The stated cost is the traffic and impressions from the AI features themselves. One unconfirmed side effect: Top Stories carousels are rendered inside AI Overviews on roughly 15.5% of U.S. news searches, per NewzDash, and may go with them."},{"q":"Does the AI Overviews opt-out stop Google training on my content?","a":"No. Google's documentation points publishers to Google-Extended for training and grounding. The UK CMA's June 2026 conduct requirement obliges Google to offer opt-outs for training, fine-tuning and grounding as well as appearance; the Search Console control covers appearance."},{"q":"What is the difference between the AI Overviews opt-out and nosnippet?","a":"nosnippet, data-nosnippet and max-snippet limit what Search displays anywhere, so they shorten the ordinary result snippet as well as the AI answer. The Search generative AI control removes the site from the AI features only and leaves the ordinary result as it was."}],"table":{"caption":"Google's controls over its AI features, as Google documents them and this wire reported them, August 2026.","head":["Control","Removes the site from","What else it costs"],"rows":[["Search generative AI control (Search Console)","AI Overviews, AI Mode, generative features in Discover","Traffic and impressions from those features; possibly Top Stories rendered inside AI Overviews (unconfirmed)"],["nosnippet / data-nosnippet / max-snippet","Text shown in the AI features","The ordinary result snippet too"],["noindex","Search entirely","All Search traffic"],["Google-Extended (robots.txt token)","Gemini training and grounding","Nothing in Search — and nothing in Search's AI features either"]]},"updated":"2026-09-12","url":"https://anythingengineoptimization.com/glossary/ai-overviews-opt-out/"}