Skip to main content
Fair Moderation

Ban appeals with an
AI arbiter

Every anti-spam system makes mistakes — Telm is the one that fixes them itself. Punished members get a polite notice with an appeal button, an independent AI arbiter reviews the case with your group's context in mind, and wrong verdicts are reversed in seconds. No angry DMs to admins, no silent bans.

  • Verdicts in seconds
  • Notices in 25 languages
  • Near-zero admin work
Message removedPolite noticeAI arbiterRestoredConfirmed spam

How fair moderation works

Polite in-group notice

When a message is removed or a member is muted, they instantly see a courteous notice with an apology for a possible mistake and an appeal button. The notice cleans itself up after a minute — no chat clutter, no public shaming.

25 languages, auto-detected

The notice speaks the language of the removed message itself — all 25 languages of the product, from English, Spanish and German to Arabic, Hindi and Chinese. International communities get the right tone automatically.

AI arbiter with group context

Appeals are judged by a dedicated top-tier AI model that reads your group's description, the original detection reason, the removed message and the member's own argument — so a trade offer in a trading chat is not judged like random spam.

Automatic approvals and rejections

A clean verdict lifts the restriction instantly — unmute, unban and reputation restored without any human involved. Confident spam is declined automatically, so spammers can't waste your moderators' time.

Case number for every appeal

Each appeal gets a #case number shown to the member and to the admins. If someone insists a decision is wrong, admins find the full case — the message, the verdict, the reasoning — in one click.

Automation you can measure

The dashboard shows how many appeals were resolved automatically, how many were approved or declined, and how many actually needed a human. Most groups see the vast majority handled by AI.

~2s
Typical AI verdict time
13
Notice languages
60s
Notice auto-cleanup
≈0
Manual reviews needed

See the public appeal page

Punished members sign in with Telegram, see the exact messages that triggered the action, and file an appeal — all on one page.

Open telm.com/appeal

Why appeals make your anti-spam stronger, not weaker

Aggressive spam filtering and happy members usually pull in opposite directions: tighten the rules and you catch innocent people, loosen them and spam gets through. Appeals break that trade-off. Telm can keep detection strict because every mistake has an instant, self-service escape hatch: the member taps one button, an independent AI arbiter re-reads the message with the group's own description as context, and a wrong call is reversed in seconds — restrictions lifted, reputation restored.

The arbiter is deliberately separate from the detection pipeline and runs on a stronger AI model — it re-judges the case from scratch instead of double-checking its own homework. Decisions of human admins always outrank the AI: if a moderator already ruled on the same content, that ruling wins. One appeal per incident, daily limits and case numbers keep the system abuse-proof, and everything is logged so you can audit any decision later.

Frequently asked questions

What happens when Telm removes a legitimate message?

The member immediately sees a polite notice in the group with an appeal button. One tap opens the bot, shows the removed message, and lets an AI arbiter review the case. If the verdict is 'not spam', the restriction is lifted automatically within seconds — no admin needed.

Can spammers abuse the appeal system?

No. Each incident allows exactly one appeal, AI checks are limited per user per day, and confident spam verdicts are declined automatically. Repeat submissions after a rejection go to human review, and case numbers make every decision traceable.

What does the AI arbiter actually see?

The full context: your group's name and description, why the system flagged the message, the message text itself, and the member's own explanation. That's how it can tell an on-topic question apart from disguised spam — something keyword filters can't do.

Do admins keep the final word?

Always. Human verdicts outrank the AI in both directions — content an admin marked as spam stays punished, content an admin cleared stays cleared. Admins can also review any appeal manually by its case number, in the bot or on the dashboard.

In which languages are members notified?

Twenty-five — every language the product speaks, from English, Spanish and German to Arabic, Hindi and Chinese. The language is detected from the removed message itself, so members are addressed in the language they actually write in.

Ready to Get Started?

Start protecting your community today. Free plan available.