# AI Citation (/glossary/ai-citation) An AI citation is a link back to your page that an AI system places inside the answer it writes, crediting your page as the source of something it said. It is something your brand receives, not something you add. When ChatGPT, Perplexity, or Google AI Overviews compose an answer, they credit a short list of sources and stop. Your brand is either inside that list at the moment the reader is deciding, or it is absent from the answer entirely. ## What Counts as an AI Citation Two different things happen when an AI answer references a brand, and only one of them is a citation. * **A mention** names your brand in the answer text with no link. It builds awareness, but it sends no traffic and gives the reader nothing to open. * **A citation** names your brand as a source and links back to your page. It builds awareness and sends a reader who already understands how you solve their problem. Both count toward [AI search visibility](/glossary/ai-search-visibility), and both are worth having. The citation is the stronger of the two, because it is the AI system telling the reader that your page is worth opening in full. Keep the three terms straight. The citation is the event: one source credit, in one answer. How often you earn one across the questions your buyers ask is your AI search visibility. The work of earning more of them is [LLM optimization](/glossary/llm-optimization). ## How an AI Citation Differs From a Traditional One A footnote is added by a writer to support a claim they have already made. An AI citation is issued by the model itself, and it sits in the reader's line of sight instead of at the bottom of the page. * **Who adds it:** a human writer or editor adds a traditional citation. A model selects an AI citation. * **What it is for:** an academic citation establishes integrity and avoids plagiarism. An AI citation establishes the credibility of the answer itself. * **Where it appears:** a reference list sits at the end of a document. An AI citation appears in the answer card, next to the sentence it supports. * **How it is chosen:** an author picks references by judgment. A model picks sources by signal, weighing authority, freshness, structure, and corroboration. * **What is at stake:** a missing footnote is an editorial problem. A missing AI citation removes your brand from the answer entirely. ## Where AI Citations Appear Every [AI search engine](/glossary/ai-search-engine) presents citations differently, so the same page can be credited on one platform and ignored on another. * **[Google AI Overviews](/glossary/ai-overview)** show linked sources beside the generated summary at the top of the results page, above the organic listings you may already rank in. * **[Perplexity](/guides/perplexity-seo)** displays numbered source links inline against the claims it makes, which makes it the most literal implementation of the idea. * **ChatGPT** shows inline references when it uses live web retrieval, and those references often shape a shortlist the reader never checks anywhere else. Because the platforms select sources on different signals, a brand rarely holds the same citation rate on all of them. Each one has to be looked at separately. ## How AI Systems Choose Which Sources to Cite AI systems weigh trust before keywords. Five signals recur across platforms. **Relevance to the question.** Retrieval matches semantic intent, not exact terms. A page that covers your topic broadly loses to a page that answers the specific question that was asked. **Authority and E-E-A-T signals.** Experience, expertise, authoritativeness, and trust still apply. Named authors with real credentials, a credible backlink profile, and coverage on respected third-party sites all raise your odds. Search Engine Land's [analysis of 8,000 AI citations](https://searchengineland.com/how-to-get-cited-by-ai-seo-insights-from-8000-ai-citations-455284) points the same way. **Freshness.** Retrieval pipelines favor recently updated pages, because stale data is the fastest route to a wrong answer. **Clarity and structure.** Direct answers, short paragraphs, clean headings, and FAQ blocks make a passage easy to extract and quote. Unstructured prose forces the model to infer structure, and it frequently picks the competitor who did not make it guess. Building pages for extraction on purpose is called [content engineering](/glossary/content-engineering). **Consensus across sources.** A claim several trusted domains repeat carries more weight than one that appears only on your own site. A single unsupported assertion rarely earns the credit. Acting on these five in order is the subject of the [complete AI SEO guide](/guides/ai-seo) and the [guide to improving AI search visibility](/guides/improve-ai-search-visibility). ## When the Citation Is Wrong AI systems sometimes invent sources, and they sometimes cite a real source inaccurately. Both create business risk. * **Lost trust:** a reader who catches one fabricated detail discounts everything around it, including the parts about you that were correct. * **Compliance exposure:** in a regulated category, a false claim attributed to your brand is a problem you own even though you did not write it. * **Misdescription:** a model that has assembled a fragmented picture of your brand will cite you for the wrong category, the wrong price point, or the wrong use case. The response is the same in all three cases. Publish clear, verifiable information, describe yourself the same way everywhere it appears, and check what AI systems currently say about you rather than assuming it is right. A citation leaves almost no trace in ordinary reporting, so checking means reading the answers themselves, which is what the [AI search tracking guide](/guides/ai-search-tracking) sets up. ## Frequently Asked Questions An AI citation is a linked source credit inside an AI-generated answer. The model names your page as where a statement came from and links to it. You do not place it. The model chooses it while it is writing, out of the pages it retrieved for that question, which makes a citation a decision taken at answer time rather than something you can publish. It appears beside the sentence it supports, not in a reference list at the end. A mention names your brand in an AI answer with no link. A citation names your brand as a source and links back to your content. Both count toward your AI search visibility, and a mention is worth having. Only the citation gives the reader a route through to you. A mention says the model knows about you. A citation says it trusts one specific page of yours enough to send its reader there. They weigh trust before keywords, across five recurring signals. Relevance comes first, because retrieval matches the meaning of the question rather than the words in it, so a page that answers the question directly beats one that covers the topic broadly. Authority and E-E-A-T signals follow, then freshness, then how cleanly the passage is structured for extraction, and finally whether other trusted sources say the same thing. No single signal wins on its own. A page that is fresh but vague, or authoritative but off the question, still loses the slot. It becomes your problem regardless of who wrote it. AI systems can fabricate a source outright or cite a real one inaccurately, and a reader who catches one wrong detail discounts the accurate parts alongside it. In a regulated category, a false claim attributed to you carries the exposure of one you published yourself. The fix starts upstream. Publish verifiable material and describe your product the same way everywhere it appears, because a fragmented picture of your brand is what produces an inaccurate citation. Then check rather than assume. Monitoring what AI systems currently say about you is the only way to catch a wrong citation before a prospect does.