How to test whether citations really support their attached claims—and whether claims that need evidence receive it.
Fidelity asks whether attribution is correct
A citation can be prominent yet point to a relevant page that does not contain the specific claim, only partially supports it, contradicts it, or is incorrectly assigned after several sources were combined.
Citation presence is not citation support.
Precision and recall diagnose different failures
Citation precision is the share of displayed citations that correctly support their attached propositions. Citation recall is the share of verifiable claims that receive adequate citation support.
An answer can show a small set of highly accurate citations while leaving many other factual claims unsupported—or cite nearly every claim using weak evidence.
Groundedness requires claim-level evidence tracing
For each factual claim, identify its associated evidence and classify support as fully supported, partially supported, unsupported, or contradicted. This is more precise than merely checking whether a link appears nearby.
Citation aesthetics can create an illusion of reliability
Users may treat numbered citations as trust signals, but a source can play navigational, evidentiary, or attribution roles. A page related to the topic is not automatically evidence for a specific proposition.
Visible sourcing should therefore be audited at the claim level.
Questions about this topic
What is citation fidelity?+
It is the accuracy of the relationship between a generated claim and the source attributed as its evidence.
What is citation precision?+
It is the proportion of displayed citations that correctly support the claims to which they are attached.
What is citation recall?+
It is the proportion of verifiable claims that receive adequate citation support.
Can an answer have high precision but low recall?+
Yes. Its displayed citations may be correct while many other factual claims remain uncited.
What is groundedness?+
It is the degree to which generated claims can be traced to and supported by retrieved or provided evidence.
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
- 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
- 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