Where The Com Talks: Cross-Platform Presence Across 2 Million COM Telegram Messages

Across 2,062,068 messages collected from 796 Telegram channels, we identified 27,612 deduplicated platform mentions spanning 80+ platforms. Discord is where members are most actively directing each other off Telegram, accounting for 40% of all external social mentions. Gaming platforms, led overwhelmingly by Roblox, make up a significant and often underappreciated share of where this network shows up. When it comes to actual account handles shared in the channels, Discord and Twitter/X are the two platforms members use most to connect.

What the data shows

Social media and comms

Of the 27,612 total deduplicated mentions, 23,660 (85.7%) fall under social media, communications, and web platforms across 44 distinct services. Telegram itself accounts for 13,512 of those. Since the corpus was collected from Telegram, that volume reflects the collection environment rather than external reach. The more meaningful picture is the 10,148 mentions pointing outward to other platforms:

Below the top six, platforms from very different corners of the internet show up with consistent frequency: 4chan (49 mentions), Truth Social (25), and SimpleX (22). The mix of mainstream and fringe platforms in the same dataset reflects how these networks move fluidly across the internet rather than staying in one corner of it.

Gaming platforms

Gaming references totaled 3,952 deduplicated mentions across 38 platforms and 750 source channels, or 14.3% of all platform activity. It is easy to treat gaming as background noise in this kind of analysis. The data suggests otherwise. Roblox alone accounts for 1,885 mentions, nearly half the entire gaming category, and puts it ahead of Instagram, YouTube, and Twitter/X in the overall platform rankings. Fortnite (453), Minecraft (329), and Xbox (275) follow at a significant distance.

Roblox’s dominance is consistent with its role as a documented recruitment platform for The Com. These are not gaming communities that happen to overlap with the network. They are platforms the network actively operates on.

Social links extraction

Platform mentions tell you where the network is talking about. Extracted handles tell you where members are actually pointing to each other. The pipeline pulled 1,469 unique handles across 30 platforms from 1,847,099 messages.

Discord produced 299 unique handles, the highest of any platform in the dataset. Twitter/X is close behind at 287. TikTok follows at 139, then Instagram at 27. The gap between Discord and Twitter/X at the top and everything else reflects how concentrated off-platform contact actually is. Members are funneling toward two platforms.

Methodology

Data collection

796 Telegram channels were monitored passively and continuously beginning in August 2025, capturing messages in real time as they were posted. The corpus includes historical messages reaching back to December 2015, though the bulk of activity spans 2024 through March 2026 when this analysis was run. In total, 2,062,068 messages were captured across the collection period.

Platform detection

165 platforms were compiled across 15 categories. Each was assigned detection patterns covering two vectors: URLs and domain variants (including short links and alternative domains), and plain-text name mentions using word-boundary matching. Platforms that operate under multiple names are merged into single entities. Twitter and X count as one. 8chan and 8kun count as one.

Ambiguous terms required stricter treatment. “Signal” is common enough in ordinary language that bare mentions were excluded; only “signal app” or “signal messenger” qualified. The same logic applied to other terms where a false positive rate would distort the counts.

Three counts were produced per platform: every individual match, distinct messages containing at least one match, and a deduplicated count that filters repeat mentions from the same source within a short time window. All figures reported here use the deduplicated count.

Deduplication

The deduplication window is 120 seconds. If the same sender mentions the same platform more than once within that window, it counts as one. This keeps bot activity and rapid-fire spam from inflating the numbers while preserving genuine repeated references that occur hours apart. 82 of the 165 platforms scanned produced enough mentions to appear in the final results.

Handle extraction

Beyond counting mentions, the pipeline extracted actual usernames and handles shared in messages. This runs as a separate process, parsing profile URLs, conversational patterns (”add me on snap @handle”), and structured link metadata embedded in messages that does not always appear in the visible message text. Three filter layers remove noise: a blocklist of common words that pattern-match as handles but clearly are not, a list of navigation URL paths that are not usernames, and platform-specific validation rules. Each handle is deduplicated per platform, per sender, and per channel.

Limitations

The analysis covers text only. Links embedded in images, videos, or other media that are not reflected in the message text are missed. Some platform names overlap with common words, and while high-risk terms are qualified or excluded, some false positives may remain. The corpus originates entirely from Telegram, which means platform mention frequencies reflect what this network talks about on Telegram specifically, and mention counts carry no signal about whether a reference is promotional, critical, or incidental.