AI search engines don’t rank pages the way Google traditionally does. Instead of returning a list of links, they generate a synthesized answer from multiple sources, then cite some of those sources inline. Whether your brand appears in that answer depends on a completely different set of signals than whether you rank on page one.
The mechanics vary across platforms, but the underlying pattern is consistent: a query comes in, the engine retrieves relevant content from its index, a large language model synthesizes a response, and the final output includes citations that point back to source pages. Getting cited is the new “ranking.” Understanding how each engine makes that choice is what separates brands that appear in AI answers from those that don’t.
This guide breaks down how Google AI Overviews, ChatGPT Search, and Perplexity actually work under the hood, what signals influence citation decisions, and what this means for your content and SEO strategy on both the traditional Google and AI visibility fronts.
How Google AI Overviews work
Google AI Overviews appear when Google’s systems determine that a generative summary adds value beyond what a traditional results page would deliver, typically for complex, multi-part queries where synthesizing information from several sources saves the user a round-trip. They don’t appear for every query.
The core mechanism is what Google calls “query fan-out.” Instead of matching a single query to a document, Google issues multiple related sub-queries across subtopics and data sources in parallel, then synthesizes the results into a single response. This is why AI Overviews often surface a wider diversity of sources than standard results for the same query, as documented in Google’s AI Overviews developer documentation.
To appear in an AI Overview, your page must already be indexed and eligible to appear in standard Google Search results with a snippet. Google’s documentation is explicit: “There are no additional requirements to appear in AI Overviews or AI Mode, nor other special optimizations necessary.” No special schema is required. However, content controls like nosnippet, max-snippet, and noindex can block a page from being included in AI features.
What does influence selection is content quality. Google’s documentation describes “advanced models” that identify supporting web pages during response generation, and the broader quality signals that have always mattered for traditional search (relevance, E-E-A-T, technical accessibility) are the same signals used to determine which pages become supporting citations in an Overview.
How Perplexity works
Perplexity operates as a retrieval-augmented generation (RAG) system. When a query arrives, Perplexity runs a real-time web search to retrieve relevant documents, passes those documents as context to a large language model (the specific model varies by query type and subscription tier), and generates a synthesized answer with numbered citations pointing back to the source pages.
Unlike traditional search, Perplexity re-crawls and retrieves content at query time rather than relying solely on a pre-built index. This makes freshness more important than in Google’s traditional results. A page that published useful, specific information recently has a reasonable chance of being retrieved and cited even without an established backlink profile.
Perplexity’s source selection favors content that directly answers the question being asked. Pages with clear structure, direct answers in the opening paragraphs, and specific factual claims tend to be extracted and cited. Content that buries the answer in preamble or relies on generic framing gets passed over. The pattern that emerges across cited sources is a preference for natural, substantive content over pages structured primarily around traditional search optimization signals.
You can explore Perplexity’s stated approach to sources at fokal.com/ai-seo/how-perplexity-chooses-sources/.
How ChatGPT Search works
ChatGPT Search, OpenAI’s search feature integrated into ChatGPT, retrieves live web content to supplement the model’s training knowledge. Where Google is retrofitting an existing search infrastructure with LLM capabilities, OpenAI has built its retrieval system from the ground up, optimized for how a language model uses retrieved documents rather than for human browsing.
The key architectural difference, documented in Fokal’s ChatGPT vs Google analysis, is intent. Google’s retrieval is optimized for efficiency at massive scale, using techniques like nested embeddings and query routing. OpenAI’s retrieval is optimized for enabling autonomous agent workflows and comprehensive answer synthesis.
In practice, ChatGPT Search favors content from established publishers and sites that carry strong traditional SEO authority signals. It also has publisher partnerships that give certain news and content organizations preferential treatment. This means that unlike Perplexity, where a newer site with direct, high-quality content can get cited quickly, ChatGPT Search tends to favor brands with existing authority and third-party mentions on reputable sites.
The shared infrastructure: why Google SEO still matters for AI visibility
A critical fact that gets lost in discussions of AI-specific optimization: all major AI search platforms use web crawlers and search indexes as their primary data source. Perplexity runs real-time web searches. Google AI Overviews draw from Google’s index. ChatGPT Search retrieves live web content to supplement the model’s training knowledge. If your content isn’t indexed, isn’t accessible to crawlers, and doesn’t rank or appear in standard search results, you won’t be cited in AI answers either.
This is the dual angle that makes Google SEO and AI citation optimization complementary rather than separate strategies. The ai-seo-research hub covers this relationship in depth. The short version: traditional search indexing is the prerequisite for AI citation. A page that isn’t crawlable, isn’t indexed, and doesn’t appear in standard search results cannot be retrieved and cited by any of the major AI platforms. The funnel starts with discoverability.
The factors that have always driven Google rankings: strong topical authority, fast and accessible pages, clear content structure, authoritative backlinks, and schema markup that matches visible content, are the same factors that increase your probability of appearing in AI summaries. From the consensus between Google and Microsoft on AI search, both agree that E-E-A-T signals and genuine expertise matter more now, not less, and that there are no new “AI-specific” tricks that substitute for substance.
What signals influence AI citation decisions
Across all three major platforms, certain patterns show up consistently in what content gets cited:
Direct answers at the top. AI systems extract content for synthesis. A page that puts the answer in the first paragraph is more likely to have that passage used than a page that builds to its answer over several sections. This is also why Google’s AI Overviews developer documentation describes the query fan-out technique: the model is pulling targeted passages, not reading whole pages.
Specific claims with named entities. Vague generalizations (“many experts believe”) are hard for a language model to attribute. Specific claims with named sources, named products, or named companies are easier to extract and verify. Content that reads like a well-researched brief gets cited more than content that hedges everything.
Technical accessibility. Your robots.txt must allow crawling by AI engines (Googlebot, PerplexityBot, GPTBot). If these crawlers are blocked, your content doesn’t enter the retrieval pool. A newer signal to consider is llms.txt, a markdown file at your domain root that curates your most important content for language models. Per the specification at llmstxt.org, it functions as a structured guide helping models understand your site’s content quickly without parsing full HTML.
Third-party mentions. For ChatGPT in particular, appearing on third-party review sites, comparison pages, industry publications, and Wikipedia significantly increases citation likelihood. AI engines treat mentions on authoritative external sites as a trust signal in the same way that traditional SEO treats backlinks.
Freshness for time-sensitive queries. Perplexity especially prioritizes recently-updated content for queries where the answer is likely to have changed. Keeping core pages updated matters more now than it did when content could coast on accumulated backlinks for years.
The practical difference between ranking and being cited
In traditional search, success means your page appears in position one and a user clicks through. In AI search, success means your content is extracted and synthesized into an answer, and your brand name appears in the citation. These are different outcomes with different downstream effects.
Ranking gets you traffic when users click. Being cited in an AI answer can build brand awareness and authority even when users don’t click through, because your brand name appears in the answer alongside the claim. This is sometimes called zero-click visibility, and it’s why zero-click search is a growing concern for businesses that have built their model around organic traffic volume.
The strategic implication: optimize for extraction, not just ranking. Write in a way that allows a language model to pull a useful sentence or paragraph from your page and drop it into a synthesized response. Clear headings that map to questions, direct answers in the opening of each section, specific factual claims, and clean HTML structure all contribute to extractability.
How to optimize for both Google and AI search simultaneously
The good news is that the overlap is large. The tactics that make content extractable by AI systems, specifically direct answers, clear structure, strong entities, and topical authority, are the same tactics that have always helped with featured snippets and Google’s answer box. The ai-seo strategy hub lays out the combined approach in detail.
The areas that are genuinely new:
Crawler access for AI bots. Review your robots.txt to confirm GPTBot, PerplexityBot, and Google-Extended are not blocked. Fokal’s AI crawler access guide covers the specific directives.
llms.txt implementation. A curated markdown file at your domain root that takes roughly an hour to write and can meaningfully improve how models understand your site. Per the llmstxt.org specification, the file uses a simple structure: an H1 with your site name, an optional blockquote summary, and H2-organized sections of links with descriptions pointing to your most important content.
Answer-first content structure. Restructure key pages so the direct answer to the implied question appears in the first 50-100 words, before any context or qualification. This mirrors how AI systems extract passages: they look for the most direct, citable statement of the answer.
Tracking AI visibility alongside rankings. Traditional GSC rank tracking tells you nothing about whether you’re being cited in AI answers. You need separate monitoring to understand your AI visibility. Fokal tracks AI citations across ChatGPT, Perplexity, and Google AI Overviews so you can see where you appear and where you don’t.
Understanding how AI rankings differ from traditional SEO rankings is the foundation for building a content strategy that works across both surfaces. The brands that will hold ground through the shift to AI-dominated search are the ones treating Google SEO and AI citation optimization as the same job, not two separate workstreams.