A citation gap is when a buyer asks an answer engine a question you should win and someone else (or no useful source) gets named or linked instead. It is useful as the plain diagnosis for “we are invisible in AI,” because it points to a measurable miss rather than a vague branding problem.
Whilst gaps are often blamed on “the algorithm,” in practice they usually trace to weaker answer pages, fuzzy entities, or missing proof. For example, the model recommends a rival with a sharp comparison page while your site buries the claim under a homepage hero. We often recommend logging gaps against a fixed prompt set, mapping each miss to a URL that should own the answer, then closing with clarity and consistency — not more unfocused blog volume.
Citation gaps are observed, not inferred: same prompt, same engines, record named / linked / absent and who appeared instead. Patterns matter more than single failures — one unlucky run is noise; repeated preference for a competitor is signal. Close gaps with pages that answer in the first screen, stable facts across site and profiles, and citable assets engines can quote. Re-measure after shipping; generative systems move, so gaps reopen.