Telegram forwarding network
Which channel forwards from which, over the whole collection: the channels we follow and every channel they forwarded from, in 3D. An edge points at the channel being forwarded from: incoming measures how much a channel is amplified, outgoing how much it amplifies others. The view recomputes from the live graph a few times a day.
This is the forwarding graph. The source-copying graph (identical texts between sources, near-duplicate links) is a different product shown on source profiles. Download this graph: telegram-network dataset.
Loading the graph snapshot…
Channels
Click a row to highlight the channel in 3D. Channels with a public handle are shown by default; channels outside the roster that Telegram names only by a numeric id are listed as “origin …”.
| # | channel | pagerank | fwds recv | amplifiers | out-deg | fwds sent | betweenness |
|---|
Reading this graph
What the colours mean (the “Stratum” colour mode):
Attributed — channels publicly attributed to Russian state-linked influence operations in official reporting (VIGINUM / FCDO / EEAS technical reports). The curated seed roster; every entry carries a provenance citation.
Discovered — channels found by following forwards out of the others and added to the roster after review.
Institutional — official state and institutional outlets (government portals, state communications), monitored for context.
Outside the roster — channels we do not follow: they appear only because a followed channel forwarded from them, by handle where Telegram exposes one, else as origin … with the numeric channel id. Their own forwards are unknown.
Highlights are computed from the published graph.
Metrics. PageRank (weighted, directed) — influence that accumulates along forwarding chains; read it alongside the raw counts, it can spike on sparse chains. Forwards received / distinct amplifiers — raw amplification. Betweenness — bridge position between clusters. Out-degree — amplifying behaviour.
Method. Snapshot computed server-side with igraph: weighted directed PageRank, weighted in/out strength, unweighted directed betweenness; 3D Fruchterman–Reingold layout with log-damped edge weights, far outliers winsorised. The browser only renders the precomputed result. Isolated channels are counted in the tiles but not drawn.