Who spreads what on Telegram around the Saxony-Anhalt election?
A network analysis of forwarding among German-language right-wing, “alternative media” and conspiracy-oriented Telegram channels.
01The question
On Telegram, content spreads one forward at a time.
Telegram has no recommendation feed: a post leaves its own audience mainly when another channel forwards it. Forwards are therefore the traces along which content spreads.
We followed those forwards through 47 German right-wing, “alternative media” and conspiracy channels, from the start of the Saxony-Anhalt election campaign to three weeks after the vote on 6 September 2026.
Public debate often describes these scenes as orchestrated campaigns. We wanted to see whether the data agrees.
Three questions to follow
1. Who decides what travels?
2. Is someone pulling the strings?
3. What happens when the polls close?
02The story
Follow the forwards.
Three parts, seven steps. Hover and click the charts to explore.
Who decides what travels?
Start with the basics: how are these channels connected?
Step 1
One network, and content flows one way
The forwards of our 47 channels link 748 channels into a single web: party accounts, “alternative media” and conspiracy channels all hang together.
- Taken, not traded. Only 4.7% of links run both ways (partly by design: only our 47 channels can forward).
- Try it: click a big node such as oliverjanich to see how much one hub connects.
Communities
Source: report Figures 2.1–2.2. Node colour = Louvain community; node size = PageRank; faded nodes are source-only channels. Links run from source to forwarder; width = number of forwards.
One network, then. But is it one conversation?
Step 2
Twelve rooms, each with its own topic
Channels mostly forward within their own group, and each group has its favourite subject.
- Different agendas. Russia/Ukraine is 38% of C5’s forwards; migration 21% of C10’s.
- One shared topic. Government and parties are a top-two topic in 6 of the 12 rooms. Click a row for its profile.
Source: report Figure 4.3. Share of each community’s forwards by main topic, %; rows sum to 100%. Residual = forwards with no keyword hit (39% overall); the second view excludes it and renormalises. Click a row for the community profile. Codes: ELE election/AfD, GOV government/parties, RUS Russia/Ukraine, USA USA/Trump, HLT health, ENE energy/economy, MIG migration, MEA Middle East, MED media/censorship, OTH residual.
If every room has its own agenda, who decides what moves between them?
Step 3
The distributors decide, not the authors
The most influential sources (left) are mostly not the channels doing the forwarding (right).
- The top source barely forwards. auf1tv is the most influential source, yet made only 12 forwards itself.
- One distributor, a fifth of the traffic. oliverjanich alone made 22% of all forwards.
The voices
The loudspeakers
Source: report Tables 3.2 and 3.3. Left: top 15 channels by PageRank among channels forwarded at least 10 times (number = PageRank rank). Right: top 15 collected channels by forwards made (number = betweenness rank). Hover a channel to find it in the other list. Reach = how many different channels; volume = how many forwards.
A handful of operators control the flow. So is someone pulling their strings?
Is someone pulling the strings?
The usual story is an orchestrated campaign. Here is the test.
Step 4
Aligned, not coordinated
We checked every pair of forwarding channels: do they forward the same posts within 10 minutes more often than chance?
- Only 2 of 820 pairs stand out, both around oliverjanich: a lead to check, not proof.
- No conductor. Channels spread similar content most likely because they share audiences and worldviews.
Source: report Tables 5.1 and 5.4. Synchronised co-forward = two channels forward the same original post within 10 minutes. Flag = at least 5 synchronised co-forwards, at least twice the expected number, and Benjamini–Hochberg q < 0.01. Funnel widths are not to scale. Timelines are schematic.
No conductor, then. So what did this network do with an election?
What happens around the vote?
From the campaign to three weeks after polling day.
Step 5
During the campaign, the war out-talked the vote
Before the polls closed, Russia/Ukraine content was forwarded 11.0 times a day; election content only 8.9.
- Bigger than the election. Over the whole period, Russia/Ukraine is 11% of all forwards, the election 10%.
- Not one fringe room. It is the top topic in three communities (C5, C6, C11).
Source: report Figure 4.3 (share of all 12,757 forwards by main topic; the 39% without a topic keyword are not shown) and Section 6.2 (per-day rates).
So the election was not the main topic. When election content did circulate, who carried it?
Step 6
Influencers carried the election, not the party
Two communities built around independent influencers carry roughly 70% of all election content.
- The party is a side player. The 5 AfD channels we collected supply just 0.9% of all forwards.
- No party hub. They are scattered across four different communities.
Share of all election/AfD forwards, by community
Source: report Figure 4.3. C2 = maeklegutelaune. Shares rounded.
Then, on 6 September at 16:00 UTC, the polls closed.
Step 7
The topic changed. The leaders didn’t.
After poll close, election content jumped from 8.9 to 15.5 forwards a day and overtook the war.
- Louder. All forwarding rose (102.7 → 112.1 a day), driven by election content.
- Not different. oliverjanich and auf1tv stayed on top; the network two weeks before and after looks almost the same.
Source: report Sections 6.1–6.3. Per-day rates, because the periods differ in length (97.67 days before, 24.33 after). Descriptive; not tested statistically. Day-by-day view and network snapshots under More charts.
Put the seven steps together, and one picture emerges.
03The answer
An ecosystem, not a campaign.
Around the Saxony-Anhalt election, this corner of Telegram did not work like an orchestrated campaign. It worked like an ecosystem: a few distributors move content between separate rooms, nobody conducts them, and the vote made it louder without changing who leads.
What surprised us
The top source barely forwards
auf1tv is the most influential source, yet made only 12 forwards. One distributor, oliverjanich, made 22% of all forwards.
The party is a side player
In an election campaign, the AfD channels supplied under 1% of forwards. Influencers carried the election content.
The war out-talked the vote
Before polls closed, Russia/Ukraine was forwarded more often each day than content about the regional election itself.
What to remember
Who spreads content matters as much as who creates it
The gatekeepers are a handful of distributors, not the authors or the party.
22%of all forwards made by one channelAligned is not the same as coordinated
Channels say similar things without acting in lockstep, so there is no single control room to switch off.
2 / 820channel pairs flaggedElections amplify what already exists
The vote raised the volume, but it ran through the same influencers who already led the network.
8.9 → 15.5election forwards a day, same leaders
One caveat: we see the network only through the 47 channels we collected, so part of the concentration of forwarding is built into the design (see Limitations).
The story ends here.
The sections below are supplementary material, not part of the main story: the data, method, extra charts, limitations, a description of the channels and our codebook, for readers who want to check how we got our results.
04Appendix · Data
78,392 messages, one small label.
We downloaded every message of a fixed list of public Telegram channels and kept each post carrying a “Forwarded from …” label.
Focus
The Saxony-Anhalt state election (polls closed 6 September 2026, 16:00 UTC). The data covers 1 June – 30 September 2026: the three months of the campaign and the weeks after the vote. Most channels operate nationwide; one (afd_fraktion_lsa) is specific to Saxony-Anhalt.
Tool
Messages were downloaded through Telegram’s official API with the Python library Telethon (v1.45.0) in a Jupyter notebook.
Dates
- 10 Oct 2026Download of the initial list
- 11 Oct 2026Download of the snowball channels
- AfterwardsData not modified
Channels were selected in two steps, following Zehring & Domahidi (2023). Select a step to see the rules.
Four inclusion criteria
- A public channel (not a group or chat)
- German-speaking (predominantly)
- Active during the election period
- Documented in published research, previously monitored, or the official channel of a party
Snowball rule
From the initial channels’ messages in the same period, every channel forwarded or linked (t.me) at least 50 times by at least two different initial channels became a candidate: 23 channels. They were screened with the same four criteria, additionally excluding commercial channels and sister channels of included accounts.
Each of the 47 channels belongs to one of three groups and records how it entered the sample. Select a group for its definition.
Group definitions follow Zehring & Domahidi (2023), Bader et al. (2023), CeMAS (2023, 2025), DFRLab (2023) and CORRECTIV/CorrelAid (2024).
German-language far-right Telegram channels include AfD politicians and party branches, extreme-right protest groups such as Freie Sachsen, Identitarian Movement activists, “alternative media” outlets like AUF1 and Compact, and conspiracist influencers. Mutual forwarding links them into a dense network that also includes Querdenken, QAnon and pro-Russian propaganda channels, so content moves from right-wing populist to extremist accounts in only a few steps. Prior network analyses of public Telegram data document these channels and their connections:
- Breidenbach, S. (2025). Die rechtsextreme Gemeinschaft in Telegram. Akteur:innen, Verbindungen, Vergemeinschaftungspraktiken. ZRex – Zeitschrift für Rechtsextremismusforschung, 5(2), 203–219. doi.org/10.3224/zrex.v5i2.03
- Schoolmann, J., & Ziemann, A. (2025). Politische Stimmung und pro-russische Rhetorik auf Telegram. Netzwerk- und diskursanalytische Befunde. ZRex – Zeitschrift für Rechtsextremismusforschung, 5(2), 255–276. doi.org/10.3224/zrex.v5i2.06
- Zehring, M., & Domahidi, E. (2023). German corona protest mobilizers on Telegram and their relations to the far right: A network and topic analysis. Social Media + Society, 9(1), 1–12. doi.org/10.1177/20563051231155106
1 · A forwarded post
Posted in oliverjanich, originally from auf1tv.
2 · An arrow
A link from source to forwarder; its weight is the number of forwards.
3 · A network
1,393 links join 748 channels into one connected network.
Excluded: 311 forwards whose source is hidden by privacy settings and 1,516 self-forwards. Only collected channels can forward; all other channels appear as sources.
05Appendix · Method
Six steps from forwards to findings.
Each step builds on the forward network from step 1. Click a step to see what it does and why.
06Appendix · More charts
For readers who want to dig deeper.
Extra detail behind the story. Not needed to follow it.
S1Channel–topic network over time
Each forwarding channel (dot) is linked to the topics it passes on (squares). Channels that share topics sit close together, so the separate thematic camps become visible.
Source: report Section 4.5. A bipartite network of the 41 forwarding channels (dots, coloured by community; size = forwards on the selected topics) and the topics (squares). Link width = number of a channel’s forwards with that main topic. The nine topics sit on a circle; each channel is pulled towards the topics it forwards most. Other / no topic sits below, does not influence positions and is off by default. Drag the slider from left to right (or press Play) to move through time; each step shows only the forwards made in that period, and periods containing poll close (6 September) are marked in red. Hover for numbers, click a node to highlight its links, tick topics on or off.
S2Forwards per day, topic by topic
The day-by-day view behind step 7. Pick a topic and switch between before and after; the dashed line marks poll close.
Source: report Figure 4.4 (7-day centred rolling mean of forwards per day by main topic, five largest communities and the rest combined) and Sections 6.1–6.3 (per-day rates before vs after poll close). Changes are descriptive; none is tested statistically.
S3The network two weeks before and two weeks after
The network in the 14 days before poll close (1,496 forwards) and the 14 days after (1,616 forwards) looks almost the same: the same hubs dominate both windows.
Source: report Figure 6.5. Before: 23 August – 6 September, 16:00 UTC; after: 6 – 20 September, 16:00 UTC. Node positions, node-size scale (PageRank) and edge-width scale are identical in both panels; the same 15 channels are labelled. Colour = community, using the report’s colours (legend above).
07Appendix · Limitations
What this study cannot tell us.
The design is descriptive and exploratory: it maps associations, it does not test causes. Click each limitation for details.
08Appendix · The channels
Who these channels are.
This is the description of the channels based on the literature we drew them from.
German-language far-right Telegram channels are public broadcast channels run by AfD politicians and party branches, extreme-right protest groups such as Freie Sachsen, Identitarian Movement activists, “alternative media” outlets like AUF1 and Compact, and conspiracist influencers. They are linked through mutual forwarding into a dense network that also includes Querdenken, QAnon and pro-Russian propaganda channels, so content moves from right-wing populist to extremist accounts in only a few steps.
We know these channels and their connections exist from prior network analyses of public Telegram data by Zehring and Domahidi (2023), Breidenbach (2025), and Schoolmann and Ziemann (2025).
References
Breidenbach, S. (2025). Die rechtsextreme Gemeinschaft in Telegram. Akteur:innen, Verbindungen, Vergemeinschaftungspraktiken. ZRex – Zeitschrift für Rechtsextremismusforschung, 5(2), 203–219. https://doi.org/10.3224/zrex.v5i2.03
Schoolmann, J., & Ziemann, A. (2025). Politische Stimmung und pro-russische Rhetorik auf Telegram. Netzwerk- und diskursanalytische Befunde. ZRex – Zeitschrift für Rechtsextremismusforschung, 5(2), 255–276. https://doi.org/10.3224/zrex.v5i2.06
Zehring, M., & Domahidi, E. (2023). German corona protest mobilizers on Telegram and their relations to the far right: A network and topic analysis. Social Media + Society, 9(1), 1–12. https://doi.org/10.1177/20563051231155106
09Appendix · Codebook
Our codebook: how we coded the topics.
Our codebook assigns each forwarded post one main topic out of nine, or the residual category OTH. It records what a post is about, not the position it takes.
It was applied automatically with the German keyword dictionaries below (exactly as used, so results can be reproduced); scope and limitations are in Section 07.
Unit of analysis and material
- Unit: one forwarded post. The text of the forwarded message is coded, not the channel it comes from.
- Material: all 12,757 forwards dated 1 June – 30 September 2026 (UTC), excluding forwards with a hidden source and self-forwards.
- Aggregation: each forward is counted for the community of the channel that forwarded it.
Coding rules
- Search the text of the forwarded post for every keyword of all nine topics. Matching is case-insensitive.
- Count the hits per topic: every occurrence of a keyword counts once (a post that mentions “Putin” twice and “Ukraine” once has 3 RUS hits).
- Assign the main topic: the topic with the most hits. Each post gets exactly one main topic.
- Ties: if several topics have the same highest number of hits, the post goes to the tied topic whose first keyword appears earliest in the text (the topic the post opens with).
- Residual: a post with no hit at all is coded OTH. This includes posts without text.
- Shares: every post counts exactly once, for one topic, so all counts are whole numbers and the topic shares of a community add up to 100%.
Keyword notation
| Notation | Meaning | Example |
|---|---|---|
plain stem, e.g. migra | matches anywhere in a word (prefix, inside or end of a compound) | migra → Migration, Migranten, Massenmigration, Emigration |
\b before a term, e.g. \bkrieg | the term must start a word; any ending is allowed | \bkrieg → Krieg, Kriegstreiber; not Weltkrieg |
\b on both sides, e.g. \bafd\b | whole word only | \bafd\b → AfD, AfD-Verbot; not AfDler |
Category definitions
| Code | Definition | Keywords (as applied) | Known false hits |
|---|---|---|---|
| ELE Election / AfD / Saxony-Anhalt | Elections, campaigns and results, the AfD as a party or its candidates, politics in Saxony-Anhalt | \bwahl · landtag · sachsen-anhalt · magdeburg · \bafd\b · siegmund · wahlkampf · wähler · stimmzettel · wahlergebnis | \bwahl also matches “choice” (Wahlfreiheit) and elections abroad; landtag any state parliament; siegmund other people of that name |
| GOV Government / Merz / parties | Federal government, chancellor, Bundestag, established parties, coalition politics, firewall debate, party bans | merz · regierung · bundestag · \bcdu\b · \bspd\b · grüne · ampel · koalition · brandmauer · verbot | verbot any ban; grüne the colour; regierung foreign governments; ampel traffic lights |
| RUS Russia / Ukraine / war | Russia, Ukraine, the war, its leaders, NATO, the Kremlin, regardless of stance | russland · russisch · ukrain · putin · selenskyj · selensky · \bkrieg · \bnato\b · kreml · moskau | \bkrieg any war; russisch also weißrussisch |
| USA USA / Trump | US politics, Donald Trump, the MAGA movement | \btrump · \busa\b · amerika · washington · \bmaga\b | amerika also Latin and South America |
| HLT Corona / health / vaccines | Corona pandemic, vaccines, WHO, pharmaceutical industry, health in general (incl. health-product advertising) | corona · covid · impf · pandemie · \bwho\b · spritze · gesundheit · pharma | \bwho\b the English word “who”; gesundheit greetings and advertising; corona the beer brand |
| ENE Energy / climate / economy | Energy prices and policy, climate policy, heating, inflation, insolvencies, the economy | energie · klima · heizung · inflation · wirtschaft · strompreis · co2 · insolven · rezession | energie, klima very broad; wirtschaft also Gastwirtschaft |
| MIG Migration / asylum | Migration, asylum, refugees, deportation, foreigners, knife crime discussed with migration | migra · asyl · flüchtling · abschieb · einwander · remigration · ausländer · messer | messer every knife; migra also Emigration |
| MEA Israel / Gaza / Middle East | Israel, Gaza, Hamas, Iran, Palestine, the Middle East conflict | israel · gaza · hamas · iran · nahost · palästin | iran can match inside other words |
| MED Media / censorship / freedom of speech | Media, public broadcasters, propaganda, censorship, freedom of speech | zensur · meinungsfreiheit · mainstream · propaganda · lügenpresse · staatsfunk · \bard\b · \bzdf\b · desinformation | mainstream non-media uses; propaganda used for all sides |
| OTH Other / no topic | Posts without a hit in any dictionary: no text, personal posts, advertising, entertainment, topics outside the nine categories | – | – |