Every punished member gets a polite, apologetic notice with an appeal button. Telm reviews the appeal automatically — cautious safety rules first, then an independent AI check — usually in under a minute. Cases the AI is unsure about wait for you in the Appeals tab of the dashboard, with a badge showing how many need review. When you approve an appeal, Telm records it as a confirmed mistake and stops similar messages from being punished across all of your groups.
Want to set this up? Do it right in your Telm account.
Open in your dashboard1Why appeals exist
No automated moderation is perfect. A brand-new member sharing a legitimate link, a regular writing in an unusual style, or a message that simply looks like a known scam can all be caught by mistake. What separates good moderation from frustrating moderation is not whether mistakes happen — it is how quickly and fairly they are fixed.
The appeal system is Telm's safety net. It gives every punished person an immediate, dignified way to say 'this was wrong', settles most of those cases automatically within a minute, and routes the genuinely unclear ones to you. It also turns confirmed mistakes into a learning signal, so the same kind of error is less likely to repeat.
2The appeal cycle, end to end
Automation gets it wrong sometimes — what matters is what happens next. The moment Telm punishes someone, it tells them politely and offers a way to appeal. For a soft punishment (a deleted message or a mute) the notice appears in the group and disappears after about a minute. For a ban or kick, Telm tries to message the person privately with the reason and an appeal button.
The member taps the appeal button, an automatic review runs, and they get a decision with a case number. If the review clears them, the punishment is lifted right away. If it is unsure, the case comes to you. Every appeal carries the original reason and the message that triggered it, so nothing is decided in a vacuum.
- Soft punishment (delete or mute) → in-group notice that auto-clears after about a minute.
- Hard punishment (ban or kick) → private message with the reason and an appeal button.
- Each appeal gets a case number so it can be tracked to a final decision.
3How the automatic review decides
The review has two steps. First, a set of cautious safety rules can clear an obvious non-offender — for example, someone with no global spammer record, punished in only one group, with no recent automatic pardons. If the rules do not clear it, the original message is re-checked by an independent AI whose only job is to ask whether the punishment was a mistake.
If the message is cleared, the ban or mute is lifted and the person is told their restrictions are gone. If the AI confidently confirms spam, the appeal is politely declined. If the AI is unsure, the appeal stays open for you to decide. The AI looks only at the single message under appeal — it does not re-judge the member's whole history — which keeps the decision focused and fast.
- Most automatic decisions land in under a minute.
- Step one is rules-only and deliberately cautious, so it never clears a risky case on its own.
- Step two is an independent AI check that asks one question: was this punishment a mistake?
- A clear yes lifts the punishment, a clear no declines the appeal, and anything in between comes to you.
4Anti-abuse limits and scope
An appeal channel is only useful if spammers cannot weaponise it. Telm caps the system on several axes so a flood of appeals can never overwhelm you or trick the automation into clearing real spam.
An appeal is also scoped to a single group. Approving an appeal decides the case for that community, but it does not touch a global spammer (CAS) ban, which is a network-wide signal owned outside your group. That is intentional: one group's owner should not be able to un-flag a confirmed cross-network spammer for everyone.
- One appeal per incident — a member cannot re-open the same case repeatedly.
- One AI re-check per incident, so mass appeals cannot slip real spam through.
- Daily limits cap how many automatic checks run, protecting you from appeal-spam attacks.
- A global spammer (CAS) ban is not lifted by an appeal — that is a network-wide decision, not a per-group one. See CAS blocklist.
5Reviewing appeals in your dashboard
Open the Appeals tab for a group in your dashboard. Anything the automation could not settle is listed there for you, with the member and the reason. Approve to lift the punishment, or reject to keep it — the last word is always yours. A badge shows how many appeals are waiting so nothing slips by.
The tab also shows stats: how many are pending, how many the system approved or rejected on its own, and the share of appeals resolved without you. Reviewing an appeal pairs naturally with the moderation journal, where you can open the exact message the decision refers to before you approve or reject.
6Handling appeals inside Telegram
You do not have to open the dashboard to clear the queue. The same actions are available as bot commands inside the group, so an admin can resolve a case on their phone in seconds.
Use /appeals to list what is pending, then /approve or /reject with the case to act on it. The result is identical to using the dashboard: an approval lifts the punishment and records the confirmed mistake, a rejection keeps the punishment in place.
- /appeals — list appeals waiting for a human decision.
- /approve — lift the punishment and record a confirmed mistake.
- /reject — keep the punishment and close the case.
7How approvals make Telm smarter
When you approve an appeal, Telm treats it as a confirmed mistake and learns from it: the wrongly punished messages are marked as false positives, cleared from spam memory, and recorded as clean examples that apply across all of your groups. One confirmed mistake stops a whole class of similar messages from being punished again.
Automatic pardons deliberately do not train the system — only decisions a human confirms do. That keeps a spammer from tricking an automatic approval into poisoning detection for everyone. If a certain kind of legitimate message keeps getting caught, approving those appeals is the fastest, most durable way to teach the bot to leave them alone.
8Getting the most out of appeals
A few habits keep the appeal queue small and the group calm. A few habits keep that conversation healthy and your group calm.
- Check the badge daily — a small, regular review keeps the queue from piling up.
- Approve genuine mistakes rather than un-banning by hand; the approval teaches the bot, a manual unban does not.
- If the same legitimate message type keeps appearing, pair approvals with a whitelist entry or a softer strictness level.
- Trust the automation for the obvious cases and spend your attention on the borderline ones it escalates to you.