How a rewrite can improve citation after context inclusion while reducing retrieval, reranking, and total visibility.
One stage can improve while the whole system worsens
A rewrite designed to increase citation after a page is already in context can change the document in ways that reduce initial retrieval or reranking. This does not mean content optimization is harmful; it means the evaluation boundary matters.
Martinez highlights SAGEO Arena because it evaluates retrieval, reranking, and citation together instead of fixing context in advance.
Conditional gain is not total gain
Total citation probability depends on both retrieval and citation after retrieval. If retrieval probability falls from 60% to 30% while conditional citation rises from 40% to 60%, total probability falls from 24% to 18%.
Better once included does not mean better overall.
A downstream benefit can be smaller than the upstream penalty it creates.
SAGEO illustrates the upstream penalty
In Martinez’s summary, body-only optimization in SAGEO Arena reduced average top-20 retrieval presence by about 9%, post-reranking top-10 presence by about 16%, and final citation by about 6%.
These results belong to that benchmark rather than every engine, but they demonstrate why live selection cannot be assumed to remain constant.
Evaluate each stage and the final outcome
Where technically possible, measure retrieval presence, reranked position, top-k inclusion, context allocation, citation, and answer influence separately. Compare treated and control conditions with repeated runs instead of relying on a single fixed-context result.
Questions about this topic
How can a citation-focused rewrite reduce total citations?+
It can improve citation conditional on context while making the page less likely to be retrieved, reranked highly, or included in top-k.
What is a conditional gain?+
It is improvement measured after an upstream event—such as retrieval or context inclusion—has already been satisfied.
What does SAGEO add to fixed-context research?+
It restores upstream retrieval and reranking, allowing evaluations to detect penalties that fixed-context experiments cannot observe.
Do SAGEO percentages apply to every platform?+
No. They are benchmark-specific evidence of a possible mechanism, not universal performance guarantees.
References
Sources are listed in APA 7 style. Preprints are identified as such and should not be treated as peer-reviewed findings unless separately published.
- Martinez, O. (2026). Optimizing visibility in generative engines: A critical survey of generative engine optimization (2023–2026) [Preprint]. arXiv. https://doi.org/10.48550/arXiv.2607.14035
- Aggarwal, P., Murahari, V., Rajpurohit, T., Kalyan, A., & Narasimhan, K. (2024). GEO: Generative engine optimization. In Proceedings of the 30th ACM SIGKDD Conference on Knowledge Discovery and Data Mining (pp. 5–16). Association for Computing Machinery. https://doi.org/10.1145/3637528.3671900
- Zhang, K., He, X., & Yao, J. (2026). From citation selection to citation absorption: A measurement framework for generative engine optimization across AI search platforms [Preprint]. arXiv. https://doi.org/10.48550/arXiv.2604.25707