Читать по-русски
Many media outlets cover disinformation campaigns in ways that inadvertently help the creators of fake news. What are they getting wrong, and what can be done about it?
This text was originally published by iStories. With the kind permission of its editorial team, we are republishing it on the Provereno website.
In May 2026, roughly three weeks before Armenia’s parliamentary elections, the Polish Russian-language TV channel Vot Tak published an article about a disinformation campaign targeting Armenian Prime Minister Nikol Pashinyan. Citing calculations by the Bot Blocker project, the article stated: “The pro-Kremlin bot network Matryoshka published 343 fake videos about Armenia,” noting that their dissemination had begun at least eight months before the election day. The piece was subsequently summarized by, among others, Meduza and Belarus’s largest independent outlet Zerkalo, while The Insider and Euronews cited it directly. The eye-catching figure of 343 was also repeated by Armenian and Ukrainian media.
At the very same time, I was preparing a major article for Provereno about disinformation surrounding Pashinyan ahead of the Armenian elections (which we eventually published in early June), and I found myself asking an obvious question: Where were all these fakes? Why hadn’t I seen them — or rather, why hadn’t I seen them in such numbers?
This is considering that I had been closely following the websites of Armenian and other international fact-checking organizations, hoping they might identify some fake story that we had missed during our routine monitoring of pro-Kremlin blogs and media. Such articles did appear. Not hundreds of them, of course, but I found three or four dozen fact-checks published over the previous several months. After compiling — and continually updating — a list, I systematically searched for each disinformation piece in Telegram, on other social media platforms, and on the websites of our regular “clients”, from Moskovsky Komsomolets to Mash. It turned out that the overwhelming majority of the fake stories examined by our colleagues either never appeared in Russian at all or attracted only hundreds, at most a few thousand views.
On the one hand, I was somewhat prepared for such a result. Last year, a colleague of mine wrote a longread about how Russian-language social media had been flooded with fake stories about Moldovan President Maia Sandu as she sought re-election, and how those stories closely resembled earlier disinformation campaigns targeting Ukrainian President Volodymyr Zelenskyy, both thematically and methodologically. As early as late summer and early autumn of 2025, we had already spotted the first Pashinyan-related fakes produced from the same template: he had supposedly purchased a $17 million mansion in Canada, while his wife, Anna Hakobyan, had allegedly siphoned off another $3.4 million from a charitable foundation intended to help sick children. The pattern was clear. With elections scheduled for 2026, we began systematically tracking future disinformation targeting the Armenian prime minister. We expected plenty of it. Yet, with very few exceptions, no posts comparable in virality emerged over the following months.
On the other hand, I began to wonder whether I was missing something. Perhaps all the channels that Provereno monitors daily simply did not publish fake stories about Pashinyan? I reopened the analyses produced by our foreign colleagues and reread them more carefully. For example, the Armenian project CivilNet debunked a fake Vogue cover claiming that Hakobyan’s outfits had been recognized as the most expensive among all Europe’s first ladies. According to the fact-check, the fake had been spread through five Telegram channels. In reality, one of those channels never published the fabricated cover at all. According to the TGStat service, the image appeared in only six channels altogether (the four correctly identified by CivilNet and two others it did not mention), where it accumulated a total of… roughly 5,500 views.
To put it mildly, that is not much.
It would have been logical to assume that colleagues had paid attention to this and other fakes because they first appeared in Russian-language blogs and then migrated into Armenian-language social media, where they became genuinely viral. But no. In the same article, CivilNet explicitly noted, “They generally did not appear in Armenian Telegram channels and had only a limited impact on Armenia’s media landscape”.
Perhaps, then, they were actively discussed by English-speaking users on X and other social media platforms? CivilNet also published an article on that topic, examining ten fake videos disguised as reports by foreign media outlets about Armenia. In most cases, each of the analyzed tweets had received around 100,000 views, which sounds impressive. However, as my colleagues at Provereno had previously observed, such figures are often manipulated: the number of likes and reposts not only fails to grow proportionally with view counts but sometimes actually decreases.
As a result, it is difficult to verify each example cited by CivilNet, since almost all of the tweets have since been deleted and the outlet did not provide links to archived copies. In the rare cases where the posts remain accessible, the hypothesis holds up. One such tweet, for example, accumulated 100,000 views, 193 reposts, and 573 likes in just three hours. Less than an hour later, the view count had barely changed, while reposts and likes had fallen to 179 and 570 respectively. Today, the tweet shows only 11 reposts and 88 likes. There are no comments, and archived copies suggest there never were any. To be fair, however, similar engagement statistics were observed on other tweets containing fake stories about Pashinyan that attracted the attention of fact-checkers from the American outlet Lead Stories.
And that was when I realized what the real problem was.
The Missing Numbers
If you open any Provereno fact-check, you will notice that immediately after the headline and lead comes a highlighted section in italics on a yellow background. In one or two paragraphs, the author demonstrates that the story under examination became viral in the Russian-language segment of the Internet. The section includes links to media outlets and blogs that spread the claim. In the latter case, authors are generally expected to specify how many views a particular post received.
The purpose of this “yellow section”, as we call it in our editorial team, is to show — and indeed to prove — that we are writing about something genuinely viral and significant: something our readers, or at least their relatives, colleagues, or neighbors, could realistically have encountered online, in the press, or on television.
Our fact-checkers usually bring these links with their pitches, and editors heavily rely on them when deciding whether a story is worth pursuing. For a long time, the unofficial threshold for approval was at least 500,000 views on Telegram, generated by posts on large, established Telegram channels with plausible numbers of reactions and comments. There was even a suggestion to raise the bar to one million so as not to spread our resources too thin.
In the context of growing restrictions on Telegram and the rise of Max — a platform to which pro-Kremlin bloggers have been actively migrating their audiences — that threshold has understandably been lowered. Yet we still try to keep it relatively high. Even so, readers continue to leave comments such as “Where do you even find this stuff? I’ve never seen any of the fakes you write about”.
And this is precisely what is missing: information about how many people actually saw the falsehoods being described or debunked. It is absent not only in the Vot Tak article mentioned at the beginning, but also in the pieces that echoed it in other media outlets, and in many similar publications, including, much to my regret, those produced by fellow fact-checking organizations. The claim that there were 343 fake claims about Armenia certainly sounds striking, but how many real people actually saw them? How many potential voters could they realistically have influenced? Of course, some share of any view count will always come from bots, but the overall scale can still be estimated with reasonable accuracy. In this case, however, the impression is that we are talking about a combined audience of, at most, tens — and perhaps hundreds — of thousands of Internet users, not millions.
The same approach can be seen in many other reports based on the findings of the Bot Blocker project. Its research is frequently cited by outlets such as The Insider. Since the beginning of the year alone, the publication has released at least 12 articles about disinformation campaigns uncovered by Bot Blocker — covering topics regarding Epstein, Ukrainian athletes at the Winter Olympics in Italy, the elections in Hungary, and so on. Yet none of those articles specified how many views the identified fake stories had actually received. In recent years, a similar approach has appeared in publications from Novaya Gazeta Europe, Kholod, Verstka, and a number of foreign media outlets as well.
Another outlet that regularly reports on Bot Blocker’s findings is Agentstvo. Previously, like many of its fellow agencies, it did not provide information about the audience reach of the fake stories it covered, but it later changed that practice. Between the beginning of the year and the end of May, Agentstvo’s Telegram channel published six such posts. In the overwhelming majority of those cases, the outlet not only included view statistics but also pointed out that the figures appeared to have been artificially inflated. In June, however, Agentstvo published a post about fake stories identified by Bot Blocker concerning the opening of the FIFA World Cup and, this time, provided no information about their potential audience.
Doppelgängers and Matryoshkas
Headlines — or at least the articles themselves — gain additional impact from references to various Kremlin-linked networks involved in creating and disseminating fake stories. Anyone who follows independent Russian-language media with any regularity has probably heard the names Doppelgänger and Matryoshka. Over time, the two have become so intertwined in retellings that they are often conflated.
Doppelgänger came first. The name was coined by the EUvsDisinfo project in the autumn of 2022. According to researchers, the campaign involved creating clone websites that mimicked well-known media outlets from various countries, differing only in their domain names. These sites were used to legitimize fake stories. The appearance of disinformation on websites resembling genuine news outlets was intended to make it more convincing. Later, the label “Doppelgänger” came to encompass the creation of websites for media organizations that did not exist at all, spreading fake stories through fraudulent social media profiles (a tactic described, for example, by the U.S. Treasury Department when explaining sanctions against Kremlin-linked media managers), the abuse of social media advertising mechanisms, and more.
The name Matryoshka, meanwhile, was coined by the Bot Blocker project itself in early 2024. In a post on X, it described the network as “an entirely new influence campaign targeting fact-checkers and reputable media organizations worldwide”. Bots operating within the Matryoshka network sent emails to newsrooms or tagged media accounts on social platforms, urging them to fact-check particular pieces of disinformation that had been created by other — or sometimes the same — bots within the network. Hence its alternative name: Operation Overload. Over time, the term also began to be used more broadly when speaking of the techniques for planting and spreading fake stories.
But are these campaigns really as effective as the media portray them?
Back in 2024, Thomas Rid, a professor at Johns Hopkins University, raised precisely this question in an article for Foreign Affairs. He analyzed traffic statistics for websites associated with the Doppelgänger network, identified through leaks and intelligence reports. The results revealed that between November 2023 and August 2024, roughly 800 clone websites attracted a combined total of only 700,000 views. In other words, each site was visited, on average, about 1,000 times over nearly an entire year.
The picture is similar on social media. A detailed article published two months later by Novaya Gazeta Europe discussed both Doppelgänger and Matryoshka and included screenshots of tweets containing fake stories. In many cases, engagement statistics had been cropped out. Among those where they remained visible, not a single fake had attracted even 3,000 views.
“My own analysis of the recently leaked documents shows that although they included no less than 24,375 links to fake Bild articles and 7,111 Daily Mail stories, the vast majority of these URLs received little to no engagement. <…> The major press coverage that Doppelganger received, especially in Germany and the United States, means that far more people likely read the secondary coverage of the exposed forgery campaigns than ever viewed the primary disinformation,” Rid concluded.
Matryoshka tells a similar story.
At one point, a representative of Bot Blocker discussed with Provereno their observations and the hypothesis that pro-Kremlin bots were attempting to burden fact-checkers with unnecessary work by encouraging them to investigate fake stories generated by the same network. We were skeptical.
How long does it take to read such a message? One minute.
How long does it take to click the link, see that the post — whether fake or not — has attracted only a few hundred or a few thousand views, and conclude that it unlikely deserves serious attention? Another minute.
How, exactly, is this supposed to significantly distract a professional newsroom from its work? That remains entirely unclear.
It is worth noting that Provereno never received such messages, despite maintaining both a feedback form on our website and a Telegram bot through which readers regularly submit ideas for future fact-checks. Why Matryoshka ignored us remains a mystery.
Perhaps because it primarily targeted foreign agencies and did not regard us as such. Perhaps because our “yellow sections” made it clear how popular a post would have to be before it attracted our attention— and it was obvious that the network lacked the resources needed to achieve organically viral reach.
Yet some of our colleagues not only received such messages but also published lengthy articles based on them, describing how pro-Kremlin bots were disrupting their work.
Missing the Bigger Picture
The way disinformation campaigns of recent years have been covered — both in independent Russian-language media and in international outlets — is, frankly, disheartening.
Yes, Kremlin-linked actors are producing more fake stories than ever before.
Yes, they routinely use clone websites to legitimize those stories and bots on X and tiny Telegram channels to distribute them. Based on Provereno’s experience, many of those channels seem to exist almost exclusively for that purpose.
Yes, a sprawling, flexible disinformation ecosystem has emerged — one capable of adapting to different audiences and objectives. And yes, advances in artificial intelligence are making the production of such content increasingly easy.
Yet the media often cover this phenomenon in a very strange way.
On the one hand, many colleagues fail to see the forest for the trees. Writing in general terms about dozens or even hundreds of fake stories, they often omit a crucial fact: unlike major news organizations, these fakes have probably — and in some cases certainly — never reached a substantial real-world audience and likely never will.
Meanwhile, genuinely viral and demonstrably influential falsehoods frequently go unnoticed.
For example, how many outlets reported on the fake story claiming that Ukrainian children had been forced to donate blood for wounded Swedish mercenaries allegedly killed in a Russian strike on Poltava in September 2024? A single post by Mash on that topic received more than 1.2 million views, and Vladimir Solovyov discussed it on Russian national television.
Or consider the 2025 fake claiming that Volodymyr Zelenskyy had purchased a luxury apartment in Dubai for his mother. The story was amplified by Russia’s largest Telegram channels and by state-controlled media outlets ranging from Lenta.ru to Moskovsky Komsomolets.
And there are dozens, if not hundreds, of similar examples in Provereno’s archive.
A fair criticism would be that we, as a specialized fact-checking outlet, are partly to blame. Bot Blocker, judging by the articles based on its findings, actively approaches journalists at major independent media organizations and offers them material for stories. We have not built an equally effective network of contacts.
Still, that is not the main problem.
The real issue is that creators of pro-Kremlin disinformation and media outlets that write about them — without specializing in disinformation themselves — appear to have formed a vicious cycle without even realizing it.
The disinformation producers receive contracts to create and spread fake news. They launch it on social media with little concern for actual impact beyond meeting numerical KPIs. If a fake goes viral, great. If it does not, engagement statistics can always be manipulated in reports.
Then major media organizations publish stories about dozens of fake narratives, thereby fulfilling their own KPIs related to reach and monetization.
The disinformation producers return to their clients and say, “Look, the media are writing about us. Give us more funding”.
And the cycle repeats.
Dozens of fake stories become hundreds. One network with a catchy name gives way to another.
Eventually, foundations and donors enter the picture, allocating grants specifically for reports about Matryoshka, Doppelgänger, and similar operations.
And this spiral can continue indefinitely.
The one group largely missing from this entire system is the public itself — readers, viewers, and listeners.
Or rather, they enter the picture mainly through coverage of those disinformation campaigns by major media outlets.
That is backwards. And it should not be this way.
Translated with the assistance of AI and reviewed by Provereno.
Cover photo: Pixabay
Если вы обнаружили орфографическую или грамматическую ошибку, пожалуйста, сообщите нам об этом, выделив текст с ошибкой и нажав Ctrl+Enter.
