# Pravda Network The Pravda network is a sprawling set of pro-Russian "news" websites — pravda-fr, pravda-de, pravda-en and dozens more — that publish almost nothing original. They republish Russian state media, pro-Kremlin Telegram channels and official statements, machine-translated and pushed out in huge volume. France's [[VIGINUM]] exposed it in February 2024 under the name **Portal Kombat**. Its significance changed in 2025, when researchers concluded that its main audience was not people. Its main audience is the AI systems that learn from the web. Researchers call this "LLM grooming": filling the internet with a narrative so that chatbots trained on, or searching, the web repeat it as fact. ## Operator VIGINUM and the Atlantic Council's [[Digital Forensic Research Lab]] trace the network to TigerWeb, an IT company in Russian-occupied Crimea, and its founder Yevgeny Shevchenko. On July 15, 2025, the European Union sanctioned Shevchenko and TigerWeb for spreading pro-Russian content aimed at Western countries. [[NewsGuard]] named Shevchenko its 2025 "Disinformer of the Year" on January 27, 2026, the year after it gave the title to [[Mark Dougan]]. Shevchenko declined to comment. ## Growth | Date | Size | Source | |---|---|---| | Feb 2024 | At least 193 sites, analyzed September–December 2023, targeting France, Germany, Austria, Switzerland, Poland, Spain, the UK and the U.S. | VIGINUM (reported by Euronews and Ukrainska Pravda) | | Mar 2025 | About 150 domains, aimed at 49 countries, in dozens of languages | NewsGuard | | Mid-Jan 2026 | 370 sites identified, 286 active, 49 languages | NewsGuard (via IDMO) | Counts vary because researchers count differently: domains versus subdomains, active versus dormant, and languages versus target countries (DFRLab reported more than 80 languages in April 2026). The direction of travel is not in dispute. The [[American Sunlight Project]] estimated about 3.6 million articles in 2024, a figure NewsGuard repeated. ## The AI finding - **February 2025.** The American Sunlight Project concluded the network appeared built to influence AI chatbots more than human readers, and coined the term "LLM grooming." - **March 2025.** NewsGuard tested 10 leading chatbots on 15 false narratives the network had spread. The chatbots repeated the narrative in **33.55%** of 450 responses, debunked it in 48.22%, and declined to answer in the rest. Seven of the 10 cited Pravda sites directly. - **March 2025.** DFRLab and [[CheckFirst]] found **1,907 links** to Pravda sites on 1,672 [[Wikipedia]] pages across 44 language editions, plus 153 X Community Notes citing Pravda domains. Wikipedia and Community Notes are both sources that AI systems trust. ## Why it matters for the KB The Pravda network is where the vault's Russian-influence material meets its AI material. Earlier operations ([[Doppelganger]], [[Storm-1516]], [[Operation Overload]]) need a human to see the fake and believe or share it. The Pravda model needs only a crawler. Volume replaces persuasion, which is the [[Firehose of Falsehood]] logic carried to machines. It exploits the same weakness as a [[Data Void]]: on a niche or fast-moving question, whoever has published the most text can become the answer. The leaked 2026 plans of the [[Social Design Agency]], which include self-populating "knowledge bases" built to outrank real sources in search, show the strategy spreading beyond one network. ## See Also - [[Large Language Model]] - [[Data Void]] - [[Social Design Agency]] - [[Operation Overload]] - [[NewsGuard]] - [[VIGINUM]] - [[Kremlin Disinformation]] - [[Disinformation Hub]]