👻 LLMs are creating "ghost personas" in the scientific community

Researchers have discovered that Claude, Gemini, and GPT models use specific sets of names that act as digital fingerprints. This has led to the emergence of fictional authors polluting scientific repositories (such as Zenodo), where more than 1,600 fake records with valid DOIs have been identified.

🌍 This poses a threat to scientific verification and risks "poisoning" future AI training datasets with synthetic content.

👤 The presence of correlating names can serve as a sign of synthetic content origin.

Source 1: https://arxiv.org/abs/2606.02184