Fudan researchers got ChatGPT, Perplexity, Grok and Google AI to cite a made-up concept within a week
A Fudan University team invented a concept called the Cactus Gravity Protocol, posted about it from ordinary accounts on nine sites the tools already cite, and watched eight of ten AI search tools repeat it within seven days. All four English-web tools tested, ChatGPT, Perplexity, Grok and Google AI, were among them.
Grok linked to a post in 30 of 32 answers counted from its first citation through day seven, Google AI in 15 of 33, Perplexity in 13 of 35 and ChatGPT in 7 of 20. Planted details travelled less far: 53% of Grok's answers carried one, against 10% for ChatGPT, 6% for Google AI and none for Perplexity.
The effect did not repeat on the other two topics. A made-up keyboard drew no citations from any of the four within the week, and only Perplexity cited the opinion post, in 11.4% of its responses. The authors do not separate topic, competition and question type, since all three varied together.
Placement mattered more than volume in a separate test on five little-known holiday destinations: the post on a frequently cited source always produced the first citation or planted detail, while 40 postings on a self-hosted WordPress site produced none. The team also bought a Chinese posting package from a GEO vendor for $14, and a Doubao answer citing one of the 13 posts with a planted detail appeared within an hour.
The tests ran in March and April, before Google said on 15 May that its spam policies cover attempts to manipulate generative AI responses. No real brand or product was tested, so the results do not show whether the pattern holds for commercial searches.
Reported by Search Engine Journal ↗
Topics: ChatGPT, Perplexity, Citations