Summary
Current ways to measure generative content in AI answer engines look only at single questions and answers, missing the bigger picture. The authors show that in a conversation, what the AI says early on strongly influences later questions and answers, causing repeated references to the same sources—a phenomenon they call conversational capture. They create a mathematical framework to measure how this happens over the whole chat, proving that benefits grow much more during longer conversations. This helps better understand and design AI systems that interact over multiple turns.
What this means in practice
- •For chatbot developers: Design chatbots that improve long conversations by tracking and optimizing how early answers influence later questions and source citations.
- •For search engine teams: Develop multi-turn answer ranking systems that consider cumulative citation effects across query sequences to enhance user engagement metrics.
A theory result. No direct application yet.
Authors
Junwei Yu, Jieyu Zhou, Mufeng Yang, Yepeng Ding, Hiroyuki Sato
Abstract
Generative Engine Optimization (GEO) shapes content to increase its likelihood of being cited by answer engines built on retrieval-augmented large language models. GEO is typically evaluated as a single-turn property: for a fixed query, an evaluator measures a source's visibility in one answer. We argue that the single answer is an inadequate unit of analysis. Human-agent information seeking forms a closed loop: the agent's answer changes the user's beliefs and therefore the next question, which in turn determines what the agent retrieves. We introduce conversational capture, a phenomenon in which a source cited early becomes substantially more likely to be cited again. Capture operates through a machine-side channel, history-conditioned retrieval, and a human-side channel, follow-up questions directed toward the captured source. We formalize the interaction as a two-layer closed-loop system and derive trajectory-level constructs: cumulative conversational visibility; a direct/feedback decomposition of trajectory gain; a nested split of the feedback term into machine-side and human-side channels; a capture coefficient; a compounding ratio; and a misranking diagnostic. Using reinforcement-process (Pólya-urn) theory, we prove that the feedback term is zero under single-turn evaluation and that GEO's cumulative payoff grows superlinearly with conversation length while capture develops. A model-derived illustration shows that the feedback term can exceed the direct term, the compounding ratio exceeds two within ten turns, and single-turn and trajectory rankings agree only weakly (Kendall's $τ= 0.4$). We connect the human channel to information foraging, trust calibration, and Bayesian persuasion, and discuss design implications for answer engines.