GeoAeoAleGeoAeoAle · Platforms
What official sources say about search and citation in Google, ChatGPT, Perplexity, and other systems.
Published now: 13 · growing collection
Microsoft Learn describes generative answers over public websites in Copilot Studio. It is useful evidence about public-web configuration and its constraints, but it does not prove that any page will be shown or cited in consumer Bing or Copilot. Measure access, retrieval, citation, and referral separately.
Read articleBrave Search is a search service with its own independent index. Its AI Answers produce a concise summary over search results and expose source links, while Ask Brave is designed for longer dialogue and follow-up questions. The documentation confirms source presentation and modes, but does not reveal the formula selecting a particular URL.
Read articleChatGPT Search is a web-connected answer mode in ChatGPT. OpenAI distinguishes OAI-SearchBot for search, GPTBot for training, and ChatGPT-User for user-triggered fetching. Allowing a crawler enables access, but does not guarantee retrieval, inclusion, ranking, or a citation.
Read articleAnthropic's official documentation confirms Claude web search and sources shown in responses. It does not disclose a universal ranking formula, the complete retrieval pipeline, or a rule that every accessible page will be selected and cited. Evaluate Claude through preserved answers and links, while measuring site access separately.
Read articleDuckDuckGo Search Assist is an optional feature in search results: it scans the web, generates a brief answer, and always links it to one or two sources. Duck.ai is a separate anonymized chat experience using models from multiple providers. Search Assist and Duck.ai should not be measured as one system.
Read articleGemini may show a Sources panel with links, but that panel proves only which sources were displayed for one answer. It does not reveal the internal crawler, ranking, or retrieval completeness. Measure page access, displayed links, claim-level citation, and visits as separate events.
Read articleGoogle AI Search is a set of generative features inside Google Search. A site should be crawlable and indexable through normal Search systems, but Google does not promise a separate path into an AI answer. Search Console and server logs expose parts of the process, not a complete citation ledger.
Read articleX Help describes Grok as a product that may decide to search public X posts and conduct a real-time web search. That confirms a web/X search surface, but not its index, weights, candidate set, or the reason for a particular citation.
Read articleKagi combines its Teclis and TinyGem indexes, external search and vertical sources, user controls, and AI surfaces. Assistant can use web access through Kagi Search, but the web toggle, lens, and answer link are observable conditions of a query—not a promise of citation.
Read articlePerplexity documents a crawler called PerplexityBot, but a request from that bot proves only that a page was delivered. It does not prove that the document entered an index, was retrieved for a particular question, appeared in an answer, or produced a visit. Technical access, retrieval, citation, and referral therefore need separate measurements.
Read articleYandex describes Neuro/Yandex AI as a mode that breaks down complex questions, finds relevant web materials, synthesizes an answer, and adds source links. Official documentation confirms source-linked web discovery, but not internal ranking or citation of any particular page.
Read articleYandex documentation helps verify crawler access, statuses, and index state, but does not reveal Alice’s internal source selection. Crawling is a readiness signal, not proof of AI citation.
Read articleYou.com publishes several distinct surfaces: consumer agents, Research/ARI, and web/news search APIs. They share web data and citations, but their capabilities must not be conflated: API annotations, a highlighted Research citation, and a consumer answer are different audit artifacts.
Read article