FIMI Executive Report
Armenia Election 2026 Narrative Intelligence Report
Between November 2025 and June 2026, Repsense monitored the Armenian information environment continuously through the campaign for the 7 June parliamentary election. Across eight months, the project tracked 1.8 million media items and 6.42 million sentences in twelve platform categories and three languages, and measured them against twelve pairs of opposing narratives. This is the final report of that engagement.
The finding at a glance
A coordinated Russian influence operation, built to suppress pro-Western turnout – and defeated at the ballot box, even as its infrastructure remained in place.
The monitoring documented a coordinated Russian influence operation targeting the Armenian electorate. It was carried by a small, concentrated network, ran six interlocking hostile narratives through a four-layer distribution architecture, and followed a playbook Russia had already deployed in Moldova and Lithuania – including templates that instructed operators to substitute one country's name for another. The operation was designed not to persuade undecided voters but to suppress pro-Western turnout by demoralising the government's support base.
It ran on two tracks. The loud track – open economic coercion in the final weeks, from import bans to gas threats to the recall of Russia's ambassador – was visible and internationally reported. The quiet track – narrative infrastructure built steadily across the eight months – was never contested. The loud track failed. The quiet track is the one that mattered, and it did not switch off on election day.
Set against a nationally representative survey (n = 1,955), the monitoring showed that across all twelve narrative pairs the information environment amplified the hostile position far beyond what the public actually believed – by as much as 71 percentage points on some platforms. Yet opinion did not move. Armenians were exposed to the Russian argument at overwhelming volume, and rejected it.
The outcome
The operation, despite its scale and sophistication, did not achieve its objective at the ballot box. Its infrastructure, however, remains in place.
On 7 June, turnout reached 58.9% – the highest in any parliamentary election since 2017, and the first regularly scheduled vote since that year. Civil Contract, the pro-European incumbent, won 49.8% and retained its majority; the pro-Russian Strong Armenia bloc came second with 23.3%.
Executive summary
Russia ran its election-interference playbook in full. The information environment tilted Russia-aligned; the ballot went the other way.
The interference playbook ran in full – and the ballot went the other way
A familiar FIMI architecture was active throughout: Russian-state Telegram channels, Armenian-language relay networks and a coordinated TikTok clique, the same infrastructure pattern documented in recent Russian operations across Germany, Lithuania and Moldova. Yet Civil Contract still secured 49.8% on the highest turnout since 2017.
Two campaigns, two outcomes – one is still running
Russia combined public pressure with a quieter narrative campaign. The first generated headlines but failed to shift the result. The second – Telegram networks, TikTok amplification, coordinated narratives – continued operating long after the votes were counted.
Russian-origin framing, domestically amplified – and built to survive takedowns
Russian-state channels set the themes, Armenian Telegram channels adapted them, and TikTok accounts amplified them for younger audiences. Because the disinformation system relies on parallel replication rather than a single channel, removing individual accounts does little to disrupt it, making it hard to counter with traditional means.
The audience amplified the framing rather than checking it – but the public held its line
Platform narratives were consistently more extreme than public opinion, often by wide margins. Yet the electorate proved more resilient than the information environment suggested. The online conversation leaned heavily Russia-aligned; the vote did not.
How we measured
Two measurements ran in parallel: which narratives dominated the information environment, and how far that environment diverged from what the public actually believed.
The Repsense Havel platform assessed media content against twelve pairs of opposing narrative statements – for example, "Russia is the guarantor of Armenia's security" versus "Armenia should diversify its partnerships." Each item was scored for alignment with one pole or the other, producing a measurable picture of which narratives dominated, on which platforms, and at what volume. In parallel, the monitoring grouped content into propagation clusters – distinct chains of related material moving across platforms – to trace how specific events spread in real time. The engagement identified 11,896 such clusters.
The monitoring data was also compared against a nationally representative pre-election survey (n = 1,955, adjusted to the Armenian adult population). This comparison – between what the information environment delivered and what the public actually believed – produced the engagement's central analytical findings.
01Same playbook, new country
Russia runs foreign information operations from a standardized playbook, transposing proven narratives and infrastructure across targets with minimal localization.
Armenia carries the same constellation of FIMI indicators seen in earlier Russian operations. Russia runs information operations from a standardised playbook, transposing proven narratives and infrastructure across targets with minimal localisation.
Cross-platform propagation here is slow and editorial – downstream platforms reproduce the upstream script in parallel rather than forwarding it – which is precisely what makes the structure hard to disrupt. The 7 June result confirms the most important parallel of all: as in Moldova, the operation shaped the information space without capturing the vote.
02A four-layer architecture
The Armenian information environment carries a consistently Russia-aligned narrative supply, organised across four distinct layers.
Russian-state Telegram
~50 Armenian-language Telegram
13-account TikTok clique
Government handles
Every top-30 directed edge originates in the Russian-state framing layer. Russia supplies the framing; the Armenian outlets do the fast, operational amplification. Because the layers reproduce the framing in parallel rather than relaying a live feed, there is no single chokepoint to close.
The propagation network
03Six narratives that lock together
Six interlocking hostile narratives formed the backbone of the operation. All grew through the campaign; none declined. The two democracy-targeting narratives grew fastest – the attack was aimed at the legitimacy of democratic governance itself.
The stack was self-reinforcing by design. If democracy is a facade, elections are rigged; if the West uses Armenia, NGOs are foreign agents; if Russia is the only guarantor, diversification is national suicide. Contesting any single narrative left the rest intact – the stack had to be challenged as a system, and it never was. The "democracy is a facade" narrative led the entire field at a 3:1 ratio over its counter-narrative, widening every week.
04The manipulation gap
Set against the pre-election survey (n = 1,955), the information environment amplified the hostile pole across all twelve narrative pairs – by 28 to 71 percentage points on TikTok above what the public actually believed.
The largest divergence fell on the question that defined the election – Russia versus diversification – where the public leaned decisively toward diversification (−0.70 on a ±2 scale, the strongest lean in the entire dataset) while the media environment pushed the opposite way (+0.35). On TikTok, 95% of accounts carried the same stance stack; a user scrolling the platform saw what looked like near-universal agreement, while the public held the opposite view on most issues. The "strong leader" narrative reached 73.7 million contacts against 19.5 million for "the country's future is democratic" – a 4:1 reach advantage. Opinion did not follow the volume.
Thirteen accounts, one ecosystem
A thirteen-account coordinated clique anchored every major anti-government narrative on Armenian TikTok, cross-posting as a functionally single operation. The tightest pair, araratnews.am and civic.am, shared 316 near-duplicate sentences across 8 narratives at a 3.6-hour median lag (p < 10⁻⁴⁵).
The debunking trap
Counter-messaging failed for a structural reason visible in cluster #803, the anti-Pashinyan chant "Qez tesnem, cavd tanem". Three single-day surges shared the same spike pattern – two opposition-originated, the third the Speaker of the National Assembly rebutting it. The platform could not distinguish defence from amplification: the Speaker's 20-sentence rebuttal counted, volumetrically, as 20 more sentences of anti-Pashinyan amplification.
Contesting narratives by quoting them is counterproductive. The reaction window is a day, and quoting the attack reopens it. Counter-framing has to lead the agenda rather than rebut it – and land within the day, before the same-operator network lifts the narrative onto the audience-facing platform.
05The coercion arc
The coercion arc followed a textbook escalation, generating a high-volume media event every 48–72 hours into the final week.
The arc ran as designed and the audience absorbed it – and rejected it. The pressure did not convert votes; what it did instead was raise the salience of the European question to the highest point of the campaign and lay the groundwork for post-election contestation. The most election-relevant narratives – that the vote would be rigged, that democracy is a facade – are pre-built for either outcome, and the infrastructure that spent eight months shaping the campaign can pivot to contesting the result within hours of polls closing.
06It worked on the narrative, not the ballot
The operation succeeded at one level and failed at another, and the distinction is the central finding of this report.
Measured across twelve binary narrative scales over the nine-week campaign window, seven ended on the constructive side and five on the hostile side. The constructive wins cluster where facts anchor the debate – the peace process, institutions, the Turkey border. The hostile wins cluster on identity and alignment, where framing, not facts, decides.
Moscow's open pressure – gas-price warnings, the Wildberries and Ozon suspensions, the Jermuk water block, the ambassador's recall – turned the vote into a referendum on Europe, and the referendum went Europe's way. The pro-European vote held at 49.8% on the highest turnout since 2017; international observers named Moscow directly. The diversification preference the public held before the campaign survived three weeks of escalating economic coercion intact.
07Two campaigns, two outcomes
Russia ran not one operation but two – in parallel, on different timescales, with very different results. Telling them apart is the key to understanding what happened, and what remains.
The loud campaign: open coercion
In the final weeks, Moscow escalated to open, attributable pressure: trade frictions after late-April customs changes, the suspension of Armenian sellers on Wildberries and Ozon, a block on Jermuk water exports, gas-price warnings, and the recall of its ambassador on 30 May. It was loud, visible and counterproductive. By attacking the European choice so openly, Russia turned the election into a referendum on Europe – and elevated the very choice it opposed. The pro-European vote held at 49.8% on the highest turnout since 2017. The loud campaign did not move the ballot.
The quiet campaign: narrative infrastructure
Beneath the coercion ran a second, quieter campaign – the narrative infrastructure built steadily across the full eight months. The "democracy is a facade" narrative outpaced its counter 3:1 and widened every week; the thirteen-account TikTok clique operated uncontested throughout; the Armenian-language Telegram layer reproduced the Russian-state script daily. None of it depended on any single event, and none of it was ever effectively contested. This campaign did not need to win on 7 June – and unlike the coercion, it does not demobilise after election day. The channels, the clique and the narrative stack all remained in place after the vote.
The loud campaign drew the attention; the quiet campaign was the one that mattered. Open coercion is self-limiting – it provokes the resistance it seeks to overcome. Narrative infrastructure is the opposite: patient, cumulative, and unbound by the electoral calendar.
What the engagement established
Eight months of continuous monitoring produced a coherent account of a coordinated foreign influence operation: how it was built, how it ran, what it achieved, and why it ultimately failed at the ballot box.
A coordinated operation, documented in full
The monitoring established beyond reasonable doubt that the Armenian information environment was the target of a structurally coordinated Russian influence operation – pre-positioned infrastructure, a self-reinforcing stack of six narratives, a four-layer distribution pipeline, and a coordination signal beyond statistical chance. It followed a documented playbook, transposed from Moldova and Lithuania, and operated at industrial scale, with nearly a quarter of all monitored content matching an identified influence narrative.
Built to suppress, not to persuade
The operation was designed to demobilise pro-Western voters rather than convert undecided ones. It concentrated on identity and grievance, framed the Prime Minister with contradictory accusations whose only shared purpose was demoralisation, and amplified the hostile position far beyond actual public opinion – a 4:1 reach advantage and a near-total inversion of apparent consensus, against a public position that did not move.
Measurement was the decisive capability
The engagement's most valuable contribution was the ability to measure the gap between what the information environment delivered and what the public actually believed. Quantified across all twelve narrative pairs and anchored to a representative survey, that gap turned the assessment from impression into evidence – allowing the operation to be characterised precisely, its failure to be demonstrated rather than asserted, and its persistent infrastructure to be identified for what it is.
Detection preceded disclosure
Throughout the engagement, the monitoring detected the operation's components – the playbook template, the diaspora network, the fabricated-video frames, the coordinated account clusters – months before external investigations confirmed them. Continuous narrative intelligence delivered a lead time that reactive analysis, however skilled, cannot. That lead time, and the analytical framework that accompanied it, is the core value the engagement demonstrated.
In sum
Over eight months, Repsense monitored 1.8 million media items, resolved a coordinated Russian influence operation into its component parts, measured its effect against real public opinion, and documented its two-track structure – the coercion that failed and the infrastructure that endures. The European choice prevailed at the ballot box. The record of how it was contested, and at what scale, is the substance of this report.
None of this happens in isolation. The engagement was carried out in partnership with the International Practitioners' Partnership Network (PRACTNET) and with Sihtasutus Liberaalne Kodanik (SALK), whose sociological research anchored the analysis in representative public opinion – a reminder that reliable narrative intelligence rests on strong partnerships as much as on the monitoring itself.