A new preprint on arXiv investigates how ChatGPT Web selects which retrieved webpages to show as citations. The abstract notes that the system retrieves webpages to answer user queries, but only a subset appears in the final citation list. Understanding this gap is the paper's central concern.
The authors frame the work as important for generative engine optimization (GEO), the practice of improving content visibility in AI-generated answers. However, the abstract does not disclose specific methods or findings, so the details of the analysis remain unclear from the source alone.
As a preprint, the paper has not yet undergone peer review. Readers interested in the full methodology and results would need to consult the arXiv listing directly.