Practical Detection Performance of GEO and Targeted-Publication Signals

Determine the detection rate and false-positive rate in practice of campaign-level signals for identifying paid GEO or targeted publication, including account history, posting timing, cross-platform duplication, source-domain clustering, abrupt account-topic shifts, and citation timing.

Background

The paper proposes several campaign-level indicators that AI-search platforms, publication platforms, and auditors could use to prioritize potentially coordinated or paid publication activity. These include account history, synchronized or tightly grouped posting times, near-duplicate content across platforms, source-domain concentration, abrupt changes in account topics, and rapid citation of newly published pages.

The authors explicitly characterize these defenses as unevaluated design considerations and note that the empirical performance of the proposed detection approach has not been established. In particular, practical deployment requires measuring both how often the signals identify genuine GEO or targeted-publication campaigns and how often they incorrectly flag legitimate activity.