Researchers have analyzed a recent AI-driven influence campaign and developed a tool to identify its generation pipeline. They created a corpus of 2,646 propagandist articles from the campaign and compared them to human-written mainstream press articles from the same period. The analysis revealed that the propaganda articles were more vague, subjective, and negative, with fewer sources cited. The researchers also found prompt instruction leaks on some of the websites, which suggested the involvement of specific AI models. This work contributes to the field of propaganda forensics and has implications for detecting and attributing AI-driven influence campaigns. --- Why it matters: This research matters because it provides a new tool for identifying and analyzing AI-driven influence campaigns, which can be used by researchers, policymakers, and organizations to detect and prevent such campaigns. It also highlights the importance of understanding the generation pipeline behind propaganda efforts to effectively counter them. Source: https://arxiv.org/abs/2608.15746