# AI Search: What It Is and How It Reaches Your Store > What AI search is, how Google, ChatGPT, Perplexity and Copilot answer, and the path a question takes to your product page. URL: https://clione.ai/glossary/ai-search/ Lang: en "AI search" is the umbrella name for something you probably already do without calling it that: asking Google and reading a summary before the links, or asking ChatGPT which sneakers to buy and getting an answer with its sources. Each company calls it something different, but **the mechanism is the same**: a language model searches, reads and writes. For a store, what changes is what competes. In a list of results, your whole page competes for a position; in a written answer, pieces of pages compete for a spot among the sources, and the shopper can decide without having visited any of them. ## How it works Almost every AI search system does three things in a row. 1. **They retrieve pages.** Google explains it for its AI features: [retrieval-augmented generation (RAG)](/glossary/rag/), which it also calls [grounding](/glossary/grounding/), uses its usual ranking systems to bring relevant, up-to-date pages from its index, and then the system reviews the specific information on those pages to respond.[^google-ai-guide] 2. **They expand the question.** With [query fan-out](/glossary/query-fan-out/), the model generates several related queries at once to request more information and fetch more results. Google gives an example: for "how to fix a lawn that's full of weeds," it may also run "best herbicides for lawns" or "how to prevent weeds in lawn."[^google-ai-guide] 3. **They compose and cite.** With what was retrieved, the system writes an answer and shows prominent links to the pages that support it.[^google-ai-guide] Microsoft describes it from Bing: assistants such as [Copilot](/glossary/copilot/) break content into smaller pieces, evaluate them for authority and relevance and assemble them into a single answer, often **from several sources at once**. That is why, it says, in AI search it matters less to rank whole pages and more which fragments earn a place in the final answer.[^microsoft-ai-search] ### Who offers it and where the pages come from | Service | Where the pages come from | How your site gets in or out | | ----------------------------------------------------------------------------------------------- | ----------------------------------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------ | | Google: [AI Overviews](/glossary/ai-overviews/) and [Google AI Mode](/glossary/google-ai-mode/) | Google Search's index[^google-ai-guide] | Page indexed and eligible for a snippet; site included in the generative AI features in [Search Console](/glossary/google-search-console/)[^google-ai-guide] | | [ChatGPT Search](/glossary/chatgpt-search/) | ChatGPT's search features, where [OAI-SearchBot](/glossary/oai-searchbot/) surfaces websites[^openai-bots] | A site that blocks OAI-SearchBot does not show up in ChatGPT's search answers[^openai-bots] | | [Perplexity](/glossary/perplexity/) | Its search results, where [PerplexityBot](/glossary/perplexitybot/) surfaces and links websites[^perplexity-bots] | Perplexity recommends allowing PerplexityBot in [`robots.txt`](/glossary/robots-txt/) to appear in its results[^perplexity-bots] | | Copilot and Bing's summaries | Bing's index[^microsoft-ai-search] | [Bing Webmaster Tools](/glossary/bing-webmaster-tools/) reports how many times you are cited[^bing-ai-performance] | Google has no separate [crawler](/glossary/web-crawler/) for AI: the `robots.txt` directives for [Googlebot](/glossary/googlebot/) are the control, because AI is built into Search.[^google-ai-features] ChatGPT, by contrast, has **separate search and training agents**, and you can allow one and block the other.[^openai-bots] The entry on [GEO (Generative Engine Optimization)](/glossary/geo/) has the full table of agents. ### What changes for a store Google's answers can include product listings, product information and local business data, and Google says [Google Merchant Center](/glossary/google-merchant-center/) feeds help make your products visible in AI answers as well as in the rest of Search.[^google-ai-guide] The same guide insists you do not need to write for AI: these systems understand synonyms and the general sense of what someone is looking for, so **you do not have to cover every variant of a query** with a page.[^google-ai-guide] ## In a store The path a question takes to a [product page](/glossary/product-page/), in a **fictitious example**. Someone asks: "which stainless steel electric kettle shuts off by itself when it runs out of water?" | Step | What happens | Where your product page can drop out | | ------------ | --------------------------------------------------------------------------------------------------------------- | ---------------------------------------------------------------- | | 1. Queries | The system runs the question and several related ones: plastic-free kettles, auto shut-off, dry-boil protection | None: you do not decide anything here yet | | 2. Retrieval | It pulls pages from its index that answer one of those queries | The page is not indexed, or that service's crawler cannot get in | | 3. Reading | It takes the pieces that answer each subtopic | The dry-boil protection detail is only in an image or a PDF | | 4. Answer | It writes two or three recommendations with their sources | Another page explains it more clearly and with data | | 5. Click | The shopper opens a source or a product | The page does not confirm what the answer promised | Step 3 is where the most product pages are lost without anyone seeing it. Microsoft warns that AI systems may not read content hidden in tabs or dropdown menus, and that **a key fact inside an image or a PDF is understood less well** than in HTML.[^microsoft-ai-search] That Search Console measures all of AI search. Its [generative AI performance report](/glossary/generative-ai-performance-report/) counts only impressions in Google's AI Overviews and AI Mode.[^gsc-genai-report] Citations in Copilot and in Bing's summaries are somewhere else, the AI Performance report in Bing Webmaster Tools,[^bing-ai-performance] and what happens in ChatGPT or Perplexity is in neither. ## How to check it 1. **Ask your customer's question in four places**: Google (see whether an [AI Overview](/glossary/ai-overviews/) appears), AI Mode, ChatGPT with search and Perplexity. Note the sources each one cites, with the date. 2. **Open your `robots.txt`** and check that you let the search agents in: Googlebot, OAI-SearchBot and PerplexityBot.[^google-ai-features][^openai-bots][^perplexity-bots] 3. **Check Search Console's [generative AI performance report](/glossary/generative-ai-performance-report/)** for Google's features, knowing it only appears if your site gets enough impressions.[^gsc-genai-report] 4. **Check AI Performance in Bing Webmaster Tools**: the cited pages and the queries the AI used to find them.[^bing-ai-performance] 5. **Look for hidden facts on your product pages**: answers in tabs, images or PDFs that an assistant may not read.[^microsoft-ai-search] ## Related concept: semantic search [Semantic search](/glossary/semantic-search/) is the part that finds pages by meaning rather than by exact word match. Google says its systems understand a page's relevance even when it does not match the query literally.[^google-ai-guide] It is one piece of AI search, the retrieving one, but not all of AI search: **semantic search finds, AI search also writes and cites**. A search engine can be semantic and still return a list of links. To keep going, the [AI search reading path](/learn/ai-search/) puts the rest of the terms in order, and the guide to [what an AI crawler sees in your store](/guides/what-an-ai-crawler-sees/) goes into the detail of each agent.