All featuresAI Advertisement Protection

AI Advertisement Protection

AI that catches Discord advertisements, invite links and scams in any language — even text hidden inside images.

Overview

Sylon reads every message in your Discord server the way a human moderator would. It combines a fast pattern engine, image vision (OCR + QR reading) and a large language model to catch advertisements, off-topic Discord server invites, crypto and investment scams, and paid-promotion or recruitment pitches — in any language. Crucially, it only removes a message when there is concrete evidence, so ordinary Discord conversation is never wrongly deleted.

See it in action

Watch on YouTube

How it works

Risk & signal scoring

Every message gets a user-risk score (account age, tenure, history, prior violations) and a language-agnostic pattern score (links, invites, contact solicitation, currency markers, mass mentions, obfuscation).

Vision + context

Images are run through OCR and QR decoding so links hidden inside pictures are caught. Recent conversation, trusted counterparts and media-channel context are weighed before any decision.

Evidence-gated AI verdict

The model returns a typed verdict, but Sylon blocks only when real evidence exists — a link, invite, image link/QR or a recognizable scam/pitch pattern. No blocking on probability alone.

Key capabilities

Detects ads, invites, scams and recruitment in any language

Reads text and QR codes inside images (OCR + vision)

Cross-server ad-history and domain reputation memory

Learned spam clusters that improve over time

Full multi-channel cleanup of an offender's messages

Configurable punishment: warn, timeout, kick or ban

An optional VirusTotal check on the destination, so a phishing page first reported minutes ago is caught before anyone here has seen it

Setup

1.Open the dashboard, go to the AI Ad Protection tab and enable it.

2.Set how many detected ads before punishment, and choose the punishment.

3.Set a log channel so every action taken is reviewable.

4.Exempt trusted roles, members or channels that legitimately post links.

How it works in detail

How detection works

Instead of matching a list of known-bad links, message content is classified for advertising and spam intent. That means it catches phrasing and patterns a link-based filter would miss, and does not need constant updates to a blocklist.

Built to avoid false positives

Sylon only blocks on objective evidence — a link, an invite, an image containing a link or QR code, or a recognisable spam pattern. Circumstantial suspicion on its own is never enough, which keeps ordinary conversation from being flagged.

When the classifier is unavailable

If detection genuinely cannot run, content is left for manual review instead of being punished automatically. Sylon never guesses when it cannot actually check.

Escalating response

A configurable number of detected ads is allowed before punishment kicks in, so a single ambiguous case does not immediately cost a member a timeout.

Settings

Ad ProtectionAutomatically detects and removes advertisement links and Discord invites using AI.

PunishmentThe action taken against the user when the limit is reached. Timeout temporarily restricts, Kick removes from server, Ban permanently bans.

Timeout Duration (min)How long a user will be timed out when caught spamming. Only applies when Timeout is selected.

Log ChannelThe channel where the bot will send log messages for this feature.

Exempt RolesMembers holding any of these roles will be exempt from this protection.

Exempt MembersSpecific members this system completely ignores, in addition to the exempt roles.

Exempt ChannelsMessages in these channels will not be checked by this protection.

Good to know

Sylon needs Moderate Members for a timeout, or Kick/Ban Members for stronger punishments.

This system looks at message content — for adverts placed in a member’s status or activity instead, that is Activity Ad Protection.

Detection runs on every message, so there is nothing to schedule or trigger manually.

The VirusTotal check is optional and does nothing unless an API key is configured. It is only asked about links Sylon's own reputation table cannot already judge, it never submits anything for scanning, and a timeout or a rate limit reads as "unknown" — never as "clean".

Where a link actually leads is decided by the URL Scanner, which runs inside this system: it shares this punishment and this log channel, so there is no second switch to find.

Frequently asked questions

Will it flag normal chat?

No. Sylon requires concrete evidence — a link, invite, image QR or a real spam pattern. A new or low-trust user chatting normally is never blocked on suspicion alone.

Does it work in non-English servers?

Yes. Detection is language-agnostic; scam and recruitment templates are caught regardless of language.

Can moderators undo a mistake?

Yes. Every protection log has an Undo button that reverses the punishment and clears the record, plus a report button to flag false positives to the Sylon team.

Related features

Spam Protection

Activity / Status Ad Protection

Server Guard (Anti-Nuke & Recovery)

Join Gate

URL Scanner