Answers are grounded in this paper and cite their sources.
Many consumers now put a product question to a generative AI system and read the answer it writes, where they once scanned a page of ranked links. If that answer names three brands, a fourth may never be considered. Firms have started to practice generative engine optimization (GEO): changing content so that answer engines mention the brand, cite its pages, and describe it well. Marketing research has a detailed account of search engine optimization and paid search, but it does not explain how a brand gets into a machine-written answer or what being there does to choice. We define answer presence as a construct with four dimensions: inclusion, prominence, framing, and attribution. We then argue that a brand can reach an answer by two routes. In the parametric pathway the model draws on what it learned in training. In the retrieval pathway it reads documents fetched when the question is asked. The two routes reward different investments. Using research on consideration sets, search costs, persuasion knowledge, and brand signaling, together with the first computer science studies of GEO, we derive nine propositions. Among them: large brands have a wider lead in answers drawn from memory than in answers drawn from retrieval; third-party coverage matters more for evaluative questions than for factual ones; and answer presence can raise choice without raising visits to the brand's website. We close with implications for managers and regulators and a research agenda.
No takes yet. Share an insight, caveat, or question.
K et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: