Narrative intelligence · Czech Republic
Engineering Euroscepticism
Five platforms, 310,000 items and 4,462 topic clusters from six months of Czech political media. One question: who is shaping this space, and what are they pushing?
Total items analysed
309,527
Sentences clustered
646,061
Topic clusters
4,462
Research window
Dec 2025 – May 2026
Key empirical finding
The grievances are real. The direction they point is engineered.
The loudest content in this space is not about Russia, and it is barely about the EU. It is domestic: distrust of institutions and the cost of living dominate every platform we monitored. What turns that raw anger into an anti-EU, anti-Western worldview is a switch – the cause of each complaint is quietly renamed, and one idea, lost independence, gives every grievance the same enemy.
And on Facebook, part of the distribution of the converted message is organised: a network of anti-system and Czexit groups places the same message into group after group on cue, hours apart, again and again – and its central group, CZEXIT, has been independently identified by Czech investigative journalists as openly spreading disinformation.
So the answer has two parts: genuine domestic anger sets the volume – and a smaller, organised layer sets its direction.
88%
Adversarial lean of the loudest message in the space – distrust of institutions
74%
Anti-EU lean of the sovereignty frame that every other complaint is poured into
01 · What they push
The sixteen messages, ranked
We measured every piece of content against 16 message pairs – each with a Russia-friendly version and its opposite. Two numbers matter for each: how often it comes up, and which way the content leans. Right of the centre line, the Russia-friendly version wins out; left of it, the opposite view does.
Volume and lean of the 16 tracked messages
All five platforms, Dec 2025 – May 2026 · lean averaged across platforms; single-platform peaks run higher
02 · The frame
The idea that ties it all together
The loudest messages in the chart are domestic. What turns them into an anti-EU worldview is only the tenth-loudest: sovereignty. It is this report's central finding.
The two identities (CZ/EU) the sovereignty narrative sets against each other.
Image: Shutterstock
Why sovereignty, and not something louder
Only mid-table by volume – its power is that every other grievance can plug into it.
Sovereignty (N02) is not the biggest message in the data. At 98,983 sentences it sits tenth of sixteen, far behind distrust of institutions (176,876) or the cost of living (145,606). Its strength is a different kind. It is the one frame every other complaint can be poured into: energy prices, migration, the war in Ukraine – each can be retold as the same thing, outsiders deciding for Czechs.
That makes it the shared destination, the point where unrelated grievances all arrive at a single enemy – Brussels and NATO. And because it never asks anyone to defend Russia, it travels across the whole political range, from open Czexit campaigners to a governing party that still calls itself pro-EU.
03 · The mechanism
How a real complaint gets turned against the EU
The messages that work best are not made up. A real complaint goes in, the cause is quietly switched, and an anti-Western version comes out – the same grievance, given a new label.
What goes in · real complaints
Corruption and elite captureEnergy pricesCost of living and inflationMigrationThe war in UkraineThe switch
The complaint stays real. Only the cause is renamed.
Now the EU, NATO and the West are to blame.
What comes out · anti-EU messages
N02EU sovereignty / CzexitN14Sanctions hurt usN08Institutional distrustN03Anti-NATON05Green Deal backlashA worked example · the energy bill
Take one real grievance: energy bills. In the data these sit inside N07 (cost of living), a cluster that leans anti-EU and anti-Western on 81.8 % of Facebook clusters and 93.3 % on X, surfacing as energy-price anger, pension worries and the charge that the government deliberately impoverishes ordinary people.
The real complaint
“I can’t pay my energy bill.”
The cause, renamed
“Diktát Bruselu” – Brussels diktat.
The same reframing runs through the network in near-identical language: that the country has become “a colony of Germany and the West”, that “Brussels crossed a red line”, and in local wind-turbine opposition groups such as Úpice Without Wind Giants, where a Green Deal complaint is organised into N02 (EU sovereignty) and N05 (Green Deal backlash).
The report calls this the most potent move in the whole environment: a bridge from “Brussels diktat” to “I can’t pay my energy bill” – abstract sovereignty fear tied straight to household pain.
Czechs demonstrating against corruption in Prague, June 2019.
Photo: Wikimedia Commons · CC BY-SA 4.0
04 · The organised layer
The coordination behind the posts
Most of what this report measures spreads on its own. The anti-EU push does not. On Facebook it moves through one organised network of anti-system and Czexit groups, with a clear operational fingerprint.
The anti-EU engine · Czexit / anti-system clique
By hand, but on cue.
1–6 h
gap between the same post landing in one group and the next
12–13
coordination strength on the core links
The same message is placed into group after group within hours – the rhythm of a small team posting on a signal, not of readers sharing on their own. Czech investigative journalists at investigace.cz have independently listed CZEXIT among Facebook groups openly spreading disinformation and extremist content.
The groups at the centre
Posts in the monitored period
The pipeline in miniature
Wind-turbine opposition groups – Úpicko bez větrných gigantů and Džbánsko a okolí BEZ větrníků – form their own coordination sub-network. A local complaint about turbines is organised into two of the 16 tracked messages: EU sovereignty (N02) and the Green Deal (N05). Documented false claims moving through this cluster include that turbine blade material causes cancer and that infrasound is dangerous at long distances. The pattern is the whole report in one village: a genuine local grievance, given a Brussels-shaped cause, moved by an organised network.
The network, mapped
Anti-system / Czexit cluster · node size = posting volume · line weight = coordination strength (links at 8.0 and above) · drag nodes, hover for detail
The network is more than a picture of who shares what – it is a machine for gaming the system. When the same post lands in group after group in a short window, Facebook’s ranking reads that repeated, synchronised activity as a wave of genuine popularity and pushes the post to more people. The coordination fakes the “lots of people like this” signal, and that signal turns fringe content into mainstream reach.
The densest hub is Aktuální zprávy z ČR (“Current news from the Czech Republic”), which links to seven other groups in the cluster; its strongest tie runs to JUSTIČNÍ MAFIE (strength 13.8). A news-styled group at the centre of a coordinated political cluster is itself part of the disguise.
05 · Platform spotlight
TikTok: narrow, extreme – and algorithmic
On TikTok the story looks completely different. There are no coordinated networks at all – zero coordinated clusters in 8,279 videos. The amplification is done by the recommendation engine, and the hostility lives one layer down, in the comments.
The videos
The least hostile layer – with the most extreme single score
By count, TikTok video is the least hostile platform in the dataset – only 5 of the 16 tracked messages lean the Russia-friendly way. But all five are economic, and one of them, “sanctions destroy the Czech economy”, scores 94.8 %. The common thread is household-level financial anxiety – energy prices, food inflation, housing – framed as a direct consequence of Western policy choices.
The comments
A different world under the same videos
In comments sampled from under political videos, every one of the 16 messages leans adversarial – even where the video above scores low. Political TikTok videos work as gathering points for hostile audiences whose views the creators’ content does not reflect. (This measures the people who choose to comment on political videos, not TikTok’s Czech audience as a whole.)
The inversion, measured
Adversarial share of the same message – in creator videos vs in the comments underneath them
06 · Conclusions
What this adds up to
Five findings, each anchored in the numbers above.
1
Domestic grievance, not Russia, carries the adversarial load.
Distrust of institutions scores above 80 % on every platform, peaking at 98.1 % on X; the single highest score in the dataset is economic – “sanctions destroy the economy” at 94.8 % on TikTok video. The content channels genuine concerns – corruption, energy prices, housing – into anti-Western and anti-EU frames, while directly pro-Russian framing fails almost everywhere. The framing is selective, not fabricated – which is exactly what makes it hard to counter.
2
Sovereignty is the frame that converts.
One idea – that the country is losing its independence – pulls otherwise separate complaints about energy, migration and Ukraine into a single story with a single enemy. It is only mid-table by volume, but every other grievance can be poured into it, and it never asks anyone to defend Russia.
3
On Facebook, part of the distribution is organised.
11.2 % of Facebook discussion clusters show likely-coordinated patterns – the highest rate of any platform. The anti-system and Czexit network shares near-identical content on cue, its central group has been independently identified by Czech journalists as a disinformation channel, and wind-turbine opposition groups form their own coordination cluster around the sovereignty and Green Deal messages.
4
Platform structure determines the threat.
Facebook produces coordination through organised groups. TikTok’s algorithm amplifies economic grievance with zero coordinated clusters – and its hostile layer sits in the comments, not the videos. X concentrates adversarial framing (10 of 16 messages) without detectable coordination. Coordination on news sites is editorial syndication, not disinformation machinery.
5
Counter-strategies must be platform-specific.
On Facebook the priority is the cross-posting infrastructure – it depends on shared tooling, which makes it brittle if enforcement targets the tooling rather than individual posts. On TikTok the comment layer is the concern, and algorithmic amplification responds to recommendation tuning rather than takedowns.
The data behind this report was collected with Apify. Content from all five monitored platforms was gathered through Apify’s maintained scrapers – which is what made a 309,527-item, six-month dataset practical to assemble and keep fresh.
The platform behind this report
Why we named it Havel
Repsense is a narrative-intelligence company based in Lithuania. This analysis was done by Havel, the Repsense narrative-intelligence platform. Václav Havel – the dissident playwright who became the last president of Czechoslovakia and the first of the Czech Republic – spent his life describing exactly what this data measures: the slow swapping of a convenient story for an inconvenient fact, what he called living within a lie.
“Truth and love must prevail over lies and hatred.”
“Pravda a láska musí zvítězit nad lží a nenávistí.”
Václav Havel · 1936–2011 · president, dissident, playwright
The platform does the same job at scale: it reads five platforms at once, groups hundreds of thousands of posts by meaning, measures them against a set list of messages, and finds the coordination that makes a borrowed grievance look spontaneous.
Listens across platforms
Facebook, TikTok, X and the open web in one pass, and follows how a campaign moves between them.
Watches the video too
Deep video analysis on TikTok, Reels and YouTube: what is shown on screen, not only what is written.
Sees where it is heading
Predicts a narrative’s trajectory before it escalates, with around 90 % accuracy on direction.
Catches the bursts
Copy-and-paste matches, shared topics and synchronised timing.
Havel by Repsense · narrative intelligence for governments, NATO, EU and enterprise.
Scope, methodology and caveats
Scope
This report analyses a targeted sample of Czech political media content, not the entire media environment. Sources on each platform were selected to capture the politically active information space: the accounts, groups, and outlets most likely to carry or amplify political narratives. The findings describe what this monitored sample looks like, not a census of all Czech-language content. Platform comparisons (e.g. “Facebook has more adversarial content than TikTok”) reflect the sampled sources and should be read accordingly.
Methodology
We collected Czech-language content from five platforms over the stated periods – Facebook and TikTok (Dec 2025 – May 2026), X and the open web (Apr – May 2026). That gave 646,061 sentences, grouped into 4,462 topic clusters, each scored against a set of 16 message pairs (N01–N16) by an AI that judges which side a piece of content takes. The “lean” is how much of the content tips toward the Russia-friendly version of a pair, weighted by how big each cluster is. We spotted coordination through copy-and-paste text matches, shared topics and synchronised timing. Data collection across all five platforms was powered by Apify, whose maintained scrapers made it practical to assemble a dataset of this size.
The 16 message pairs we tracked
These message pairs were written for the Czech information space. “Pole A” is the version that, if pushed hard, helps Russia’s messaging goals – whether or not the content actually came from Russia. “Pole B” is the opposite view. Some pairs (N08, N07) are about real domestic grievances that happen to overlap with the messages an outside campaign would want to amplify.
Caveats
These scores come from patterns in the words and the data, not from any legal finding; “Russia-friendly framing” describes which way the content leans, not a proven actor or campaign. The comment figures reflect the kind of people who choose to comment on political content, and may lean more hostile than each platform’s wider Czech audience. The shorter X and web periods fall just after the new government formed, which may push their scores higher than the longer-running platforms. The coordination we saw on news sites is normal editorial syndication, not disinformation machinery. Photographs of Prague and of Václav Havel are sourced from Wikimedia Commons; the 2019 Prague demonstration photograph is from Wikimedia Commons, licensed CC BY-SA 4.0; the flags image is licensed via Shutterstock.