How models synthesize retrieved evidence—and why a visible citation is only a proxy for source use.
Generation is synthesis, not copying
Once context has been selected, the model may summarize, compare, paraphrase, combine, restructure, or omit information. Several sources can support one sentence, while one source can influence multiple parts of an answer.
This many-to-many relationship is why traditional rank and simple citation counts cannot describe all source influence. The foundational GEO study introduced word-based and position-adjusted measures to estimate the answer space a source receives.
Platforms expose sources differently
Citation rendering is the interface layer that connects generated claims to source URLs or documents. Platforms may use inline links, numbered markers, source cards, expandable panels, or a list of consulted pages.
The interface determines what researchers can observe. A visible citation indicates selection for attribution, but it does not reveal the full internal contribution of a source.
Selection is not absorption
Zhang, He, and Yao distinguish citation selection from citation absorption. Selection asks whether a source is shown. Absorption asks how deeply its facts, language, evidence, or structure appear to shape the answer.
A source can be visible as a link while contributing little to the response—or influence the answer without prominent attribution.
Questions about this topic
Does a generative model copy retrieved pages?+
Not necessarily. It can summarize, compare, paraphrase, combine, or omit information from several sources when constructing a response.
What is citation rendering?+
It is the interface method used to connect generated claims with source URLs, such as inline citations, numbered markers, cards, or source panels.
Does a visible citation prove that a source shaped the answer?+
It is evidence of visible attribution, but only a proxy for underlying use. Citation selection and the depth of source absorption are distinct.
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.
- 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
- 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