Dozens of mainstream U.S. outlets have unwittingly amplified a sprawling pro-Kremlin disinformation network that has links from hundreds of other sites by referencing its stories in their own reporting, new research suggests. The study characterizes the network, which it calls Pravda, as a series of more than 200 interconnected sites that publish millions of articles in multiple languages — many with narratives that echo Russian state interests.
ISD found close to 900 websites spanning the political spectrum that have cited articles produced by the Pravda network in the last year, including at least 300 English-language sites.
They were found to:
- Treat Pravda content as credible (just over 80% of citations)
- Properly flag it as part of a coordinated information operation (fewer than 5%)
- Reference its Russian origin without sufficient context, or place links in areas like comment sections (some 15%)
What the Report Says About U.S. Outlets and Their Citations
The ISD report calls out respected outlets, including The Atlantic, Politico, Forbes, and the Denver Gazette, in a list of the offenders that referenced Pravda-linked journalism without enough caveats. It also mentions mainstream opinion and commentary sites including The Gateway Pundit and Jacobin. Other outlets, such as The Washington Post, Newsweek, Fortune, and the Des Moines Register, at times identified the sites as Russian but did not report that they were part of a disinformation network, according to the analysis.
There’s a pattern of what these researchers call “link laundering,” in which bogus claims on fringe websites are semi-legitimized through reference in mainstream reportage, newsletters, or blog posts — and then further laundered through sharing by aggregators and social network accounts. Over time, those links can lend credibility to dubious narratives in search results and naturalize them into the public conversation.
How the Pravda Network Works to Spread Content Online
ISD describes “The Pravda Hacking” as a high-volume content mill optimized for reach, not necessarily for accuracy. The network has promoted over 6 million pieces and frequently shares new ones every few seconds. Its infrastructure does seem to be optimized for search and social discovery: with mirrored domains, repeated headlines, and syndication into multiple languages, odds increase that a narrative ends up in front of distinct audiences under different brand identities.
The network’s reporting often promotes pro-Kremlin frames on geopolitical issues, including Ukraine and Western policies. The amount of material and the distribution tactics do make manual editorial checks difficult, particularly for smaller newsrooms with fewer fact-checking resources or larger sites that depend heavily on freelancers, wire copy, or automated linking tools, ISD notes.
Narratives Are Being Recycled by AI Tools
Another set of studies by NewsGuard found that popular AI chatbots read and even repeat Pravda-influenced claims. In its testing, four of the 10 systems republished claims linked to the network, such as one false report that members of Ukraine’s Azov Battalion burned effigies in Poland of a former United States president. Some chatbots also echoed unverified claims that there was corruption among some Ukrainian military officials. NewsGuard refers to this as “AI grooming,” where the mere existence of misinformation makes it more likely that generative systems will treat it like formal training data.
The AI contagion illustrates how link practices in journalism can have outsize implications. When mainstream sites link to dodgy sources — even accidentally — the signals can reverberate through search indexes, knowledge graphs, and training corpora, making mis- and disinformation difficult to unwind down the line.
Why Link Hygiene Matters in the Time of Coronavirus
The risk for editors is two-pronged: the trust of an audience and algorithmic amplification. Veteran disinformation researchers like those at organizations such as the Atlantic Council’s DFRLab and the EU’s East StratCom Task Force have long warned that state-aligned actors exploit openings within the editorial process — user-generated comments, op-eds, syndicated feeds, automated link widgets — to slip in narratives that can later be cited as “reported elsewhere.”
The practical steps are easy if not always comfortable: treat unknown domains as unverified until vetted; add explicit context when linking to state-linked properties; use nofollow attributes on untrusted sources; and periodically audit archives to correct or annotate problematic links. Newsrooms in the business of consuming wire copy or managing hundreds of contributors should weigh automated link scrubbing, domain blocklists, and clear guidelines for citing adversarial state media.
The Bigger Picture on Links, Search, and AI Training Data
ISD’s findings also add weight to the idea that the fight over information integrity is less about one site, more about infrastructure: how links travel, how search ranks them, and how AI systems learn from them. While high-volume networks can produce authority through repetition, every unexamined citation is thus a force multiplier.
The lesson here for publishers is obvious: linking is editorial. It’s often a matter of a few seconds doing source verification, reading an extra line of context, or left- or right-leaning policies that treat sketchy domains more cautiously — and it can be exactly the difference between amplifying propaganda and tearing at its threads.
