The Ghost Couple: Correlated LLM Name Priors and Their Haunting of the Web and Academic Publishing
A new research paper reveals that large language models generate correlated "ghost" personas, such as Elena Vasquez and Marcus Chen, who appear across numerous AI-generated documents as experts or co-authors. These correlated name priors are specific to model families and versions, leaving detectable fingerprints. The study found these ghost authors have been used to create over 1,600 fake academic records on repositories like Zenodo, with fabricated publication dates, and have also formed synthetic research groups on platforms like ResearchGate. AI
IMPACT Reveals LLM-generated personas are used to create fake academic records, impacting scholarly integrity and trust in AI-generated content.