How AI answer engines choose who to cite

Traditional search ranks ten links; an answer engine cites two or three sources. What decides which ones, and what a business can actually influence.

Sample article shipped with the build

A growing share of searches never reach a results page. Someone asks an assistant a question and gets a paragraph back with two or three sources attached. Those citations are the new first position, and they are not awarded the same way rankings are.

Citation is a retrieval problem, not a ranking one

An answer engine retrieves a small candidate set, then decides which sources support the claims it is about to make. A page that ranks fourth for a query may be cited when the top three are not, because it states a fact plainly enough to be quoted.

What this changes in practice

  • Structure beats length. A clear, self-contained answer near the top of a page is quotable; the same answer buried in paragraph nine is not.
  • Entity consistency matters more. If a business describes itself differently across its own pages, retrieval has a weaker signal to bind those pages to one subject.
  • Freshness carries different weight. Answer engines lean harder on recently updated sources for anything time-sensitive.

How we measure it

Search Growth probes answer engines directly for the queries a business cares about and records whether it was cited, with the citation itself as evidence — not an estimate of “AI visibility” derived from rankings.

This is a sample article included with the site build. Replace it with a real post before launch.

All articles