# AI Search Engine (/glossary/ai-search-engine) An AI search engine is a search tool that answers your question in writing instead of returning a list of links. It uses a large language model to understand the question, retrieves relevant pages from a search index, and writes one answer that links to the sources it drew on. Perplexity, ChatGPT, and Google AI Overviews all work this way. You will also see it called an artificial intelligence search engine, an AI-powered search engine, an answer engine, or a generative search engine. The names all describe a search engine that writes the answer. The practical difference is who does the reading. A traditional search engine ranks pages and leaves you to open and compare them. An AI search engine reads them for you and hands back a conclusion, usually naming a handful of brands. If your brand is not in that answer, a buyer who stops at the answer never sees you. This entry covers the engine itself: what it is, how it works, and which products count as one. The [AI search](/glossary/ai-search) entry covers the wider shift these engines are part of and what it means for your business. ## How an AI Search Engine Works An AI search engine works in three steps. It reads the question, retrieves pages that could answer it, and writes an answer from those pages. 1. The model reads the question. It interprets a full, conversational question ("which project management tool suits a remote team?") for what it asks, the way a person would. 2. The engine retrieves pages. Google describes its own version as [retrieval-augmented generation, also called grounding](https://developers.google.com/search/docs/fundamentals/ai-optimization-guide): its core Search ranking systems pull relevant, up-to-date pages from the Search index. AI Overviews and AI Mode can also split one question into several related searches, which Google calls [query fan-out](https://developers.google.com/search/docs/appearance/ai-features). 3. The model writes the answer from what those pages say and links to the ones that support it. Retrieval is what makes it a search engine. A model answering from training data alone cannot see newer pages or show where an answer came from. The [2020 paper that introduced retrieval-augmented generation](https://arxiv.org/abs/2005.11401) named both as open problems. Where the pages come from depends on the engine. Google uses its Search index. OpenAI and Perplexity run search crawlers of their own, [OAI-SearchBot](https://platform.openai.com/docs/bots) and [PerplexityBot](https://docs.perplexity.ai/docs/resources/perplexity-crawlers), and both tell site owners who want to appear in search results to allow them. How an engine picks among the pages it retrieves is a longer story. The guide to [how LLMs choose sources](/guides/how-llms-choose-sources) follows it step by step, with what each company documents about its own system. ## AI Search Engine vs. Traditional Search Engine A traditional search engine ranks pages and shows you a list. An AI search engine reads the pages and writes the answer. Both start from a search index, so the difference is what happens after retrieval. That changes what visibility means: * **Result format:** a list of links versus one written answer with a few source links. * **The question:** a few keywords versus a full question, with follow-ups in the same conversation. * **Clicks:** the answer often settles the question on the spot, so fewer searches end in a visit. The entry on the [Google AI Overview](/glossary/ai-overview) has the click data. * **Competition:** on a results page, pages compete for positions. In an AI answer, they compete to be one of the few sources the answer draws on and links. * **Winners:** a results page lists many organic links, while an AI answer usually names a handful of brands. Ranking first on Google no longer guarantees you appear when an AI search engine answers the same question. Even Google's own AI Mode often cites pages outside its top 10, as the [Google AI Mode SEO guide](/guides/google-ai-mode-seo) sets out. Using AI is not what separates the two. Google [launched RankBrain in 2015 and BERT in 2019](https://blog.google/products/search/how-ai-powers-great-search-results/), both AI systems that help it understand what a search means and rank the results. The new part of an AI search engine is that the AI writes the answer. ## AI Search Engine vs. AI Chatbot An AI chatbot is an AI search engine only when it retrieves current pages and cites them. AI on its own is not a search engine: a language model writes from what it learned in training, and a search engine finds pages. An AI search engine does both. The test is retrieval and citation, and it applies to any product. A chatbot answering from its training data alone is not searching. Google AI Overviews searches without being a chatbot at all. The same assistant can pass the test on one question and fail it on the next, because it decides each time whether to search. Anthropic's documentation, for example, says [Claude determines when to search based on the prompt](https://platform.claude.com/docs/en/agents-and-tools/tool-use/web-search-tool). For your brand, that decides where there is anything to win. An answer written from memory links to no one. An answer written from retrieved pages carries citations, and your page can be one of them. ## Types and Examples of AI Search Engines AI search engines come in three types: * **Answer engines:** products built around search, where the written answer is the main result. * **AI features in a traditional search engine:** a written answer above an ordinary results page, or a conversational mode beside it. * **AI assistants with web search:** chat assistants that search the web when a question needs current information. These are the main examples, each with its own audience and behavior: * **ChatGPT:** OpenAI's assistant has search features built in, and its search answers draw on sites found by OpenAI's own search crawler. * **Perplexity:** built around live web search, with answers that link the pages behind them. * **Google AI Overviews:** an AI summary shown above the regular results in Google Search, with links to explore further, and a form of [search generative experience](/glossary/search-generative-experience). * **Google AI Mode:** Google's conversational search mode for follow-up questions and complex comparisons. * **Gemini:** Google's standalone assistant, a separate product from the AI features in Search, though [both run on Gemini models](https://search.google/ai-in-search/). * **Claude:** Anthropic's assistant, which [added web search in 2025](https://claude.com/blog/web-search) and cites the pages it uses. Perplexity is the answer engine in that list. AI Overviews and AI Mode are features of Google Search, and each has an entry of its own: [AI Overview](/glossary/ai-overview) and [AI Mode](/glossary/ai-mode). Two more belong on the list. Microsoft describes [Copilot Search in Bing](https://blogs.bing.com/search/April-2025/Introducing-Copilot-Search-in-Bing) as a blend of traditional and generative search with clearly cited sources. Brave Search answers a question with [a short summary and references to its sources](https://search.brave.com/help/ai). They do not always name the same brands for the same question, so visibility has to be measured on each. The [ChatGPT vs Perplexity vs Claude comparison](/guides/chatgpt-vs-perplexity-vs-claude) covers which to prioritize, and the [best AI search tools guide](/guides/best-ai-search-tools) compares the platforms for measuring your visibility across them. ### The second meaning: AI site search The phrase has a second meaning in business software. There, an AI search engine is the search on a company's own website, store, or documents, built to match the meaning of a search as well as its keywords. Vendors such as Algolia and Coveo sell it. Algolia describes its product as [combining keyword and semantic techniques](https://www.algolia.com/products/ai-search) on top of a company's own search data. This entry is about the public engines people use to research and buy. ## How Accurate AI Search Engines Are AI search engines get things wrong, and a source link under an answer does not prove the answer is right. Two independent tests put numbers on it: * In March 2025, the Tow Center for Digital Journalism asked eight AI search tools to find the source of news excerpts. They [gave incorrect answers to more than 60% of queries](https://www.cjr.org/tow_center/we-compared-eight-ai-search-engines-theyre-all-bad-at-citing-news.php). * In October 2025, a study coordinated by the European Broadcasting Union and led by the BBC reviewed more than 3,000 responses about the news from ChatGPT, Copilot, Gemini, and Perplexity. Its journalists found that [45% had at least one significant issue](https://www.bbc.co.uk/mediacentre/2025/new-ebu-research-ai-assistants-news-content), and 31% had serious sourcing problems. Both tests asked about news, so the figures are not an error rate for every question. They do show that an answer with links still needs checking. For your brand, that is a reason to read the answers themselves. An AI search engine can name you and still describe your product wrongly, or credit a claim to the wrong page. A link to your page inside an answer is an [AI citation](/glossary/ai-citation), and that entry covers what to do when one is wrong. ## Why AI Search Engines Matter for Your Brand When an AI search engine answers a buying question, the brands it names make the shortlist and the rest do not. B2B buyers are in those conversations too: [Forrester reports that 94% of business buyers](https://www.forrester.com/blogs/state-of-business-buying-2026/) use AI during their buying process. That makes [AI search visibility](/glossary/ai-search-visibility) a metric in its own right: how often, and how prominently, an AI search engine names your brand. RankZero tracks it daily across ChatGPT, Perplexity, Gemini, and Google AI Overviews, with Claude on request. You see where you are cited and where a competitor is chosen instead. To start earning those citations, the [AI search guide](/guides/ai-search) and the [complete AI SEO guide](/guides/ai-seo) cover the how. What makes a page worth citing, and how to build for it, is the work of [answer engine optimization](/glossary/answer-engine-optimization). To see which AI search engines recommend your competitors in your place, [get your AI Search Audit](/services). ## Frequently Asked Questions An AI search engine uses a large language model to understand a question and retrieve relevant web pages. It then writes one direct answer with links to its sources, rather than listing ranked pages. A classic Google results page shows you links to open and evaluate yourself. AI search engines, including Google's own AI Overviews and AI Mode, give you the answer with the shortlist already chosen. When it searches the web, yes. ChatGPT is a chatbot first, and it works as an AI search engine when it retrieves current pages and cites them in its answer. When it answers from its training data alone, it is not searching, and the answer links to no sources. People use AI search engines for questions that would otherwise take several searches: researching a topic, comparing products or suppliers, and getting a current summary with sources they can check. Follow-up questions are part of the use. Google AI Mode, ChatGPT, Perplexity, and Claude all let you refine the question in the same conversation instead of starting a new search. They name brands from the pages they retrieve for the question, so the choice starts with which pages get retrieved. From there, the engine narrows those pages down to the passages it reads and cites. The guide on [how AI engines pick the pages they cite](/guides/how-llms-choose-sources) walks through each step. It depends on the job. Perplexity is built around web search and links the sources behind its answers, which suits research. ChatGPT and Claude suit questions you want to refine over a conversation. Google AI Overviews needs no new tool at all, because it appears inside ordinary Google results. For brands, the more useful question is which AI search engines your buyers use, because that is where your visibility needs measuring first. No. They build on it. Google says its generative AI features are rooted in its core Search ranking and quality systems. ChatGPT and Perplexity use search crawlers of their own, so a page they can crawl and index is still the starting point. The difference is the target: traditional SEO optimizes for ranked positions, while [AI search optimization](/glossary/ai-search-optimization) optimizes for citations inside AI answers. Most brands now need both. The [complete AI SEO guide](/guides/ai-seo) covers how they fit together.