A citation in AI search is the named web source a language model attributes a claim or answer to when generating a response. It functions as the model's stated evidence trail, linking output back to a specific URL.
Citation in plain English
A citation is the specific webpage an AI search engine credits when delivering an answer. If a shopper asks ChatGPT for the best running shoes for flat feet and the model references a Healthline article and a Runner's World review, those two URLs are the citations behind the response.
AI search engines retrieve a set of candidate pages through a search index or live web crawl, extract passages relevant to the prompt, and pass those passages to the language model as grounding context. When the model generates its answer, it attaches the source URLs of the passages it used. Citations appear as inline links, numbered footnotes, or a 'Sources' panel depending on the interface.
Done well, a citation points to a page that directly substantiates the exact claim made in the answer, with the cited passage matching the model's wording closely. Done poorly, the citation links to a page that mentions the topic loosely but does not support the specific claim, or points to a homepage instead of the relevant deep page.
Across major AI search interfaces, answers typically surface between 3 and 10 citations per response. Pages that get cited share consistent traits: clear factual statements, structured headings, named entities, and content that answers a specific question rather than burying the answer in narrative.
Why citation matters for ecommerce
Citations are the new referral channel for ecommerce. When a shopper asks an AI assistant which brand makes the most durable yoga mat or which retailer ships fastest to Canada, the cited sources shape both the answer and the click. Stores cited in product comparisons, buying guides, and category overviews get named in the response and earn qualified traffic from buyers already mid-decision. Stores absent from citations are invisible at the moment of recommendation, even if they rank on traditional Google. Getting cited requires content structured for extraction: direct answers, specs in tables, named products, and pages built around the exact questions shoppers ask.