Study Measures How Easily Public Posts Enter AI-Search Citations

Summary

The paper examines whether ordinary users can place content into AI-search citations by publishing on domains that platforms frequently prefer. Unlike traditional search, AI search selects sources and citations before generating an answer, creating a potential security problem when source choice is concentrated. The researchers built a framework combining citation mapping across platforms, tests of publication barriers, and marker-controlled publication experiments. Across 10 platforms, they analyzed 17,211 citation instances covering 6,356 domains; the top 20 domains accounted for 20.5% to 70.8% of citations on individual platforms. Of 22 publication platforms linked to cited sources, 15 had low or medium barriers for both account creation and posting. In controlled experiments, ordinary posts caused 8 of 10 platforms to cite a fabricated concept within seven days, and one article on a highly preferred platform had more citation impact than more than 20 matched posts on low-preference platforms. A $14 GEO purchase generated 13 public posts, and one platform cited content containing the researchers’ markers within an hour. The findings indicate that source selection and publication access can materially affect AI-search outputs, although the study measures observed citation behavior rather than every platform’s underlying selection process.