Introducing our new Pravda network dataset: 15 million articles in 47 languages

Our dataset on this pro-Kremlin Information Manipulation Set (IMS) has more than doubled: from over 7 million to more than 15 million articles, published across 88 websites in 47 languages. Here is what changed, what we share openly with the community, and what we make available to analysts who need to go further.

In February 2025, an investigation by the DFRLab and CheckFirst revealed how the Pravda network, a pro-Kremlin website ecosystem, expanded its influence in one year. We have been continuously monitoring this network, which kept publishing despite being linked to a sanctioned entity. Along the way, we’ve been gathering more data on how it operates: how much it publishes, in which languages, and which sources it copies.

What is the Pravda network?

The Pravda network is a Russian website ecosystem. It produces no original content, but republishes, in bulk, posts from Russian state media, pro-Kremlin Telegram channels and official websites. VIGINUM exposed it in February 2024 under the name “Portal Kombat” and linked it to TigerWeb, an IT company based in occupied Crimea. In July 2025, the Council of the EU sanctioned TigerWeb and its founder. We have been tracking the network, notably with the DFRLab:

In February 2025, we showed that it used 140 subdomains targeting more than 83 countries and regions, including in minority languages.
In March 2025, we found more than 1,900 links to Pravda sites in 44 language versions of Wikipedia, and 153 X Community Notes referencing them. None of the chatbots we tested warned users about the source.
In April 2025, we launched the Pravda dashboard.
In April 2026, we released the CheckFirst Import Connector for OpenCTI with Filigran.

Our data on the network also fed the thinking around the adoption of the IMS framework, together with EU DisinfoLab, VIGINUM, CASSINI, the German Federal Foreign Office, the EEAS and the DFRLab.

Other organisations have documented the network’s impact on AI. The American Sunlight Project and NewsGuard showed how its content reaches large language models. The DFRLab has since traced around 40,000 English-language Pravda articles into Common Crawl, one of the main sources of AI training data. Some researchers argue that chatbots cite such sites because credible information is missing on niche topics, not because of deliberate manipulation. Whichever reading prevails, the network keeps publishing, and growing.

From 7 to 15 million articles

Over the past weeks, we rebuilt our collection of the Pravda network from the ground up. We went back through the archives of every site, recovered articles our previous collection had missed, and made our database the single reference for everything we publish. We also cleaned the data and suppressed potential duplicates.

The result, as of 28 October 2026:

15,046,455 articles collected since June 2023;
88 websites and 47 languages;
6.8 million articles in 2025 alone, and already 5.2 million in 2026;
around 18,000 new articles a day over the last month, with a peak of 28,497 on 15 May 2025;

The largest sites are the English-language hub (2.4 million articles), followed by the Spanish, French, German and US editions (between 0.8 and 1.1 million each); the most republished sources are Telegram channels and Russian state media, starting with TASS and RIA Novosti.

Open data, for everyone

Our commitment has not changed: the community should not have to start from scratch every time.

portal-kombat.com shows, for the whole network and for each site, the number of articles over time, the most republished sources, topics, languages and publication hours. It is free, and now runs on the full dataset.
github.com/CheckFirstHQ/pravda-network publishes the same aggregated data as open files, one per site plus one for the whole network, updated at least twice a day. Researchers, journalists and data scientists can download them, track trends, or plug them into their own tools.

The full archive, for analysts: our OpenCTI connector

Some teams need a deeper level of granularity: each article, its source, its date, its links, and the ability to cross-reference them with their own investigations. For them, we offer the CheckFirst Pravda Network Feed.

The CheckFirst Import Connector brings the Pravda network straight into OpenCTI:

the full historical archive, now more than 15 million articles, each modelled as standard STIX 2.1 objects;
daily updates: new articles are detected and ingested every day;
the known domains and hosting infrastructure of the network, linked to the Portal Kombat campaign;
a ready-to-deploy connector that handles authentication, pagination, state persistence and batch ingestion for you;
technical documentation and email support.

The feed is available on request, as an annual subscription with tiered pricing depending on the type of organisation. Existing subscribers get the new archive at no extra cost. To find out more, write to us at info@checkfirst.network.

Fifteen million articles since June 2023. Around 18,000 more every day. Exposure did not stop the network, and neither did sanctions. The Pravda network bets on volume. We answer with data: open statistics for everyone who studies it, and a complete, structured archive for the analysts who need to go further.