Methodology

Which communities are included

Public Telegram channels and groups linked as official by crypto projects on CoinGecko and on DEX listing sites, plus communities discovered through forwards and links from those. Communities with fewer than 50 members are tracked but not published. Private and invite-only chats are never read.

Members and growth

Member counts are recorded at least daily from Telegram's public channel page, and hourly for active communities read directly. Growth figures compare the current count with the count recorded 1, 7, or 30 days earlier in our own history. We do not use third-party historical data, so history begins the day a community was first tracked, which is shown on every page.

Activity and views

Posts per day is measured over the most recent 30 posts. For broadcast channels, view counts are read from public posts and re-read 1 hour, 24 hours, and 7 days after posting, since views mature over time. Views per subscriber divides average views by the member count. Groups do not have view counts on Telegram, so groups are measured by the share of members online and by message volume.

Trust score

A 0 to 100 estimate of how genuine a community's size and engagement are. Every community starts at 60 and each signal moves it up or down by a stated weight. Signals are scored against communities of similar size in our own index, so "views per subscriber better than 80% of channels its size" means exactly that. For channels: views per subscriber, how much views vary across posts (bought views are uniform), whether views keep growing after posting, forwards and reactions per view, growth smoothness, who forwards the channel, and age. For groups: how many members actually post, share online, spam score, join burstiness from join events, growth smoothness, and age. How much data exists sets a confidence level and caps the score: low confidence cannot exceed 75, medium 88, and nothing scores 100. Labels: 75 and above looks genuine, 50 to 74 mixed signals, under 50 likely inflated. The score is a statistical estimate, not a finding about any account.

Written pages

Released profiles carry an overview, takeaways, cautions and FAQs written by a language model from the page's own metrics and a sample of recent public messages. The model may only use figures we pass it, never quotes or names a member, and is refreshed weekly. The auto-generated paragraph under the cards states the source data and date.

Controversy score

For groups, 0 to 100 from the last 7 days of member messages: the share that are complaints, the share of replies that are hostile or negative, the balance of negative to positive mood, and the share written in shouting (caps, repeated punctuation). Groups with fewer than 100 messages or over half spam are not scored. The disputes on a page are described in our own words from the messages; no member is quoted or named.

Spam score

For groups, 0 to 100 from the share of messages that repeat earlier messages or come from bot accounts, how concentrated posting is among a few accounts, and messages per poster. Communities with a spam score over 60 are excluded from activity rankings.

Lookalikes

Handles that resemble an official channel, by shared core name, small edit distance, or similar title, found through Telegram search and through channels that forward official posts. Listed so readers can verify before joining. A listing is not an accusation; some lookalikes are legitimate fan or regional channels.

Risk flags

Pattern rules applied to public posts: referral schemes, wallet-connect prompts, DM solicitations, urgency language, listing claims, guaranteed-return claims, airdrop claims, and seed-phrase requests. Flags describe post content, not intent.

Corrections and removal

Channel owners can request corrections or removal on the removal page.