In September 2023, as Azerbaijani forces launched their offensive in Nagorno-Karabakh, forcing over 100,000 Armenians from their homes, another crisis was unfolding online. Armenians trying to document events in real time began noticing something unusual: posts disappearing, accounts restricted, views sharply reduced. Terms like “Artsakh,” “Armenian Genocide,” and even “genocide” itself appeared to trigger automated moderation systems. At the very moment global attention and awareness were most needed, Armenian voices were being muted.
This was not an isolated incident or a technical error. It points to a broader structural issue at the intersection of platform governance, financial incentives, and geopolitical pressure. This raises serious concerns for journalists, organizations, and anyone invested in free _expression_ in the Armenian world.
Not a Glitch
Evidence suggesting disproportionate moderation of Armenian-related content has accumulated over time. During the 2022–2023 blockade of the Lachin Corridor, the Facebook account of Artsakh’s President Arayik Harutyunyan, was suspended on August 26, 2023. According
to CyberHUB-AM, an Armenian NGO that tracks digital rights violations, the suspension resulted from a sustained campaign of complaints filed by Azerbaijani users, a pattern the organization had previously documented in its reporting on platform abuse during the 2020 War. This action effectively cut off a key communication channel at a critical moment, when information from the ground was already scarce. According to Armenpress, the office of President Arayik Harutyunyan confirmed that his Facebook account had been subjected to a sustained campaign of coordinated complaints over several months, which progressively restricted its reach, limiting the visibility of posts and disabling certain functions, until the page was rendered completely inaccessible on August 26.
The issue extends beyond isolated accounts. According to Freedom House’s Freedom on the Net 2021 report on Armenia, Facebook banned the accounts of numerous Armenian opinion leaders and media outlets for violating its terms of service. Rights activists noted that the decisions lacked transparency, and some of those banned were accused of hate speech without publicly available evidence to support the claims. The same report documented that during the 2020 War, Armenian media outlets’ YouTube channels were taken down as a direct result of mass reporting campaigns coordinated by users from Azerbaijan and Turkey.
Some journalists report experiencing “shadowbanning.” Content technically remains visible but receives minimal distribution, effectively preventing it from reaching any meaningful audience. Even the use of the word “Artsakh” appears, in some cases, to trigger automated content moderation systems.
The experience is not theoretical for those managing accounts for Armenian organizations. “We have had to learn to work around the platform’s invisible boundaries,” says Stephen Colangelo, the social media coordinator at AGBU. “Words like ‘genocide’ trigger sensitivity filters on social media platforms. Posts about Artsakh repeatedly see reduced visibility, not because they violate any policy, but simply because of the subject matter.” The result, he explains, is a form of self-censorship born of necessity. “You reword things, you soften the language, you adapt, because if you push too far, your reach drops, your account gets restricted, or worse, you get suspended entirely.”
Mariam Khaloyan, the director of Congressional Relations at the Armenian Assembly of America, observed similar dynamics. “During the 2020 War, it became increasingly difficult to effectively disseminate information related to Armenia and Artsakh. While content was being shared widely, I observed that its reach did not align with typical engagement patterns, particularly when posts were informational in nature or called for action.” She adds: “I have observed a consistent pattern. Content that is strictly informational—rather than provocative or algorithmically engaging—tends to receive disproportionately low visibility and engagement across platforms.”
Her experience reflects a larger reality of how platforms operate. As Tauhid Zaman, professor at Yale, explains, “TikTok and these other platforms select the content they show you. They can promote anything, demote anything. That means they can shift opinions any way they want.”
Incentive Driven
Social media platforms may present themselves as open public spaces, but they are, first and foremost, advertising businesses. These financial relationships influence which voices get amplified and which get quietly pushed aside.
According to Meta’s own published business case studies—of which there are multiple featuring Turkish Airlines—the carrier has run large-scale, targeted advertising campaigns across Facebook and Instagram for over a decade. As Reuters Events reported in an interview with Turkish Airlines’ Interactive Marketing manager, the airline manages over 20 local Facebook and Twitter accounts as part of a philosophy of being “globally yours as a local,” making it one of the most intensively involved airlines in social media activity globally.
Azerbaijan’s influence on Western institutions, meanwhile, has been documented through investigative reporting. The Organized Crime and Corruption Reporting Project (OCCRP) uncovered what became known as the “Azerbaijani Laundromat”—a $2.9 billion laundering scheme used, in part, to cultivate influence across political and media institutions in Europe and beyond. According to OCCRP’s investigation, the money was used to pay off Western politicians and silence criticism of Azerbaijan’s human rights record abroad. The same investigations described efforts to channel government narratives through third-party voices— academics, think tanks, and commentators—to create the appearance of independent analysis where there was none. The real-world consequences of this scheme have since been confirmed by courts: as OCCRP reported and Deutsche Welle covered, a Munich court convicted former German Bundestag member Eduard Lintner of bribing public officials on behalf of Azerbaijan—the first time a German lawmaker was found guilty of such an offense.
Stealth Suppression
Beyond financial relationships, platform systems themselves can be exploited. According to Meta’s own Adversarial Threat Report, the company has formally acknowledged and documented coordinated networks of fake accounts engaged in mass-reporting campaigns targeting journalists, activists, and political figures. In Meta’s own words, “mass reporting activity” occurs when adversarial networks “coordinate to abuse our reporting systems to get accounts or content incorrectly taken down,” typically with the intention of silencing others. Critically, Meta’s reports have directly documented how, in the South Caucasus region, the company removed 589 Facebook accounts, 7,665 pages, and 437 Instagram accounts linked to the Youth Union of the New Azerbaijani Party, a network that posted content touching on “tensions between Armenia and Azerbaijan” and “Armenia’s actions during past escalations in Nagorno-Karabakh.”
Some journalists report experiencing “shadowbanning.” Content technically remains visible but receives minimal distribution, effectively preventing it from reaching any meaningful audience.
High Stakes
Conflicts unfold in real time across social media feeds, and the ability to document, share, and bear witness has become inseparable from the conflict itself. When one side’s accounts are suspended, their hashtags suppressed, and their posts quietly buried by automated systems, it is a form of erasure. Narratives shape how the world responds to crises, how governments decide to act, and ultimately, how history is written. Stories only survive if they are allowed to circulate. Suppressing an Armenian voice posting from Stepanakert in September 2023 is a decision about what future generations will know about what happened there. Platforms have become the arbiters of which voices reach the world during a conflict, and they are no longer neutral infrastructure.
The Armenian Assembly of America has witnessed similar patterns directly. “I have heard similar concerns from other organizations and community members. While largely anecdotal, these observations suggest a broader, shared experience across the community,” shared Khaloyan.
Bot Business
Armenians are not alone in raising these concerns. A December 2023 report by Human Rights Watch, Meta’s Broken Promises: Systemic Censorship of Palestine Content on Instagram and Facebook, documented widespread suppression of pro-Palestinian content across Meta platforms, including removals, account restrictions, and reduced visibility. Similar concerns have been raised by Uyghurs and Kurdish communities, pointing to a broader pattern of platform over-moderation affecting groups whose narratives often fall outside dominant media frameworks.
Compounding these challenges are the coordinated bot networks and disinformation campaigns, which actively distort online discourse. Bots rely on networks of automated or semi-automated accounts that amplify hostile narratives, flood comment sections, and artificially boost misleading or fabricated content. Investigations by CivilNet document how “bot” campaigns target Armenia on fake websites designed to mimic credible international media outlets. AI-generated videos and hundreds of anonymous accounts also work in concert to legitimize false claims.
Such infrastructure is not unique to actors targeting Armenia. Turkey has developed one of the most extensively documented state-linked troll and bot ecosystems in the world. Known as “AK Trolls,” these networks of fake accounts disseminate propaganda for the ruling Justice and Development Party (AKP) and target the opposition, operating through bots, compromised accounts, and coordinated retweets. In June 2020, Twitter suspended and removed 7,340 accounts attributed to the youth wing of Turkey’s ruling AKP for organized inauthentic activities. According to the Stanford Internet Observatory, the network consisted of batches of fabricated personalities created on the same day, alongside centrally managed compromised accounts. Research has also found that in 2019, fake accounts or bots generated an average of 26.7 percent of the daily top ten Twitter trends in Turkey. Turkey’s bot infrastructure has also been deployed internationally, including to legitimize its 2019 military offensive in northeastern Syria—demonstrating that these networks serve outward-facing geopolitical narratives directly relevant to the South Caucasus.
Azerbaijan operates in close alignment with this model. Researcher Karena Avedissian documents how Azerbaijan attempts to influence external audiences through troll farms, social media propaganda, Western PR companies, and funded academics to legitimize its narratives. Notably, Azerbaijani trolls on Twitter have repeatedly been ignored, while networks of other state actors’ bot armies have been taken down—a gap that, as Avedissian warns, allows autocratic regimes to use social media as a tool of disinformation, intimidation, and harassment abroad.
Together, these dynamics create a dual pressure: legitimate content can be limited through moderation or reduced visibility, while false or hostile narratives are simultaneously amplified. For smaller communities like Armenians, this imbalance can significantly shape what the world sees, and what it does not.
Content that is strictly informational—rather than provocative or algorithmically engaging—tends to receive disproportionately low visibility and engagement across platforms.
Algorithms Win
Algorithms reflect the priorities, incentives, and limitations of the institutions that design them. When those systems consistently disadvantage certain voices—particularly those documenting conflict or historical injustice—the consequences extend beyond individual posts or accounts. The issue is no longer about visibility. It is about who gets to participate in the global conversation and whose narratives are permitted to shape public understanding.
As Khaloyan emphasizes, the implications extend beyond visibility alone. “It is deeply concerning. Limiting the visibility of the word ‘genocide’ undermines efforts to educate the public and preserve history. Without the ability to share and discuss these realities openly, it becomes increasingly difficult to promote awareness and prevent future atrocities.”
When one side of a conflict can flood a platform with coordinated reports or purchase influence through advertising relationships, they do not need to win the argument. They simply need to ensure the other side is not heard. In an era where public opinion, international intervention, and even legal recognition increasingly follow the narrative that dominates online, the power to silence is the power to win.