When propaganda no longer needs direct readers
Traditional propaganda normally tries to persuade people through newspapers, television or social media. Some newer operations add another layer by creating large networks of websites and articles that can surface in search engines, retrieval systems and the online sources used around AI products. NewsGuard describes Pravda as a pro-Kremlin network and says it published roughly 3.6 million articles in 2024. In the organization’s testing, major AI systems repeated some false narratives laundered through the network.
The important change is the intermediary target. Operators do not necessarily need millions of people to visit an obscure site directly; they may try to make claims sufficiently abundant and searchable that aggregators, retrieval tools or web-connected AI systems encounter them. That does not mean language models are automatically “poisoned” by repetition alone, because systems use different ranking, filtering and safety mechanisms. But flooding the information environment can raise the cost of verification and make a false claim appear more established than it really is.















