How to stop bots and raids from joining your Discord

It starts with a ping. Someone drops a screenshot into the mod channel: #general is filling with the same scam link pasted over and over, or a wall of slurs, or a hundred accounts spamming @everyone. The member count climbs faster than any human can read join notifications, and half your moderator team is asleep.

By the time you’re online, the damage is done — and now you have a second job: figuring out how they got in so it doesn’t happen again. That’s the whole problem of discord raid protection in one sentence: raids don’t break into your server, they walk in through the same doors your real members use. So the fix isn’t faster mods. It’s making your doors smarter.

What you’ll learn

• The five layers that make up real raid protection — invite hygiene, a gate before the server, Discord’s native controls, automated filtering, and an in-raid playbook
• Why a raw, never-expiring invite link is the biggest structural hole most servers have
• What Discord already gives you natively, and where to verify current behavior
• The lockdown steps that help during a raid — and the panic moves that make it worse
• Why “recruit more moderators” feels like the answer but isn’t

What a raid actually is

A raid is a coordinated wave of joins aimed at disruption rather than participation. The payload varies: scam links, phishing, slurs, porn spam, or just noise designed to make the server unusable. The delivery varies too — sometimes it’s bored humans with alt accounts, sometimes it’s automation running on disposable identities.

The part that matters for defense: raiders exploit ordinary join mechanics. They use an invite link, they pass through whatever join flow you have, and they land in whatever channels your default permissions open up. Nothing about a raid is supernatural. Which means every layer you add to slow down, filter, or screen the join path works against raids by definition — whether the attacker is a botnet or a person with a grudge.

Layer 1 — Invite link hygiene: the front door raiders actually use

Start with the unglamorous layer, because it’s where most raids begin. A raw discord.gg/... link that never expires is a permanent open door with your address printed on it.

Invite links leak in ways you won’t notice:

• Posted in a Reddit thread, a forum signature, or an old tweet you forgot about
• Listed in server directories and listing sites you submitted to once for growth
• Screenshotted into a group chat, then forwarded where you can’t see
• Shared by a member who meant well and pasted it somewhere public

Once a link is out, you can’t chase it down. And there are tools that crawl public pages specifically to harvest invite URLs — so “nobody will find it” is not a plan.

The hygiene moves:

1. Never run your growth on a single permanent link. Create separate invites for each place you post, so you know which link is which.
2. Use expiration and usage limits where Discord offers them. Discord’s invite settings let you constrain links; the exact options shown depend on your server type and have changed over time, so check the invite controls in your own server settings and Discord’s support site rather than trusting a tutorial.
3. Rotate on purpose. If a link has served its purpose — a launch, an event, a collab — revoke it. If you suspect a leak, revoke it immediately.
4. Audit periodically. Discord shows which invites have been used; if one link accounts for a wave of joins you didn’t expect, that’s your leak.
5. Prefer one stable URL you own that points at your current join flow. When the underlying invite changes, the public link stays the same — and you control what’s behind it.

Layer 2 — Put a gate before the server

Invite hygiene controls which links exist. A gate controls who gets a link at all.

The pattern is simple: instead of handing the invite to everyone, you publish a join page — a page that explains what the community is, states the rules, and only then offers the invite, or collects an application first. Between “stranger clicks a link” and “stranger is inside your server,” there is now a moment of friction where you can screen.

What that friction can look like, from lightest to heaviest:

• A landing page with rules and context, then the invite — filters out people who won’t read two paragraphs
• A form or application with a few questions a bot or drive-by troll can’t be bothered to answer
• Manual approval, where a human reviews answers before anyone gets the invite
• Automated checks — email verification, captcha, region rules — if you use a hosted tool

The point is not bureaucracy. It’s that intent is visible at a gate and invisible at a raw link. Someone who answers three questions to join is self-selecting; a bot that joins instantly through an open URL is not.

This pattern works even without any product: a form, a spreadsheet, and a mod who approves people is a perfectly legitimate gate for a small server. It stops scaling when approvals become a daily chore — which is when hosted join pages earn their keep. We’ve written about the pattern itself in how to build a community join page and specifically for Discord in Discord server landing pages.

Layer 3 — Discord’s native controls: verification, screening, onboarding

Discord ships its own layers, and you should use all of them that fit your community.

Verification levels. Servers can require increasing levels of account verification before new members can participate — things like account age or a confirmed contact method. The exact requirements at each level have changed over time, so don’t trust a blog’s table from two years ago: read the current docs on Discord’s support site and test what your own server settings show.

Membership screening. Community-enabled servers can require newcomers to acknowledge rules before they can interact. It’s a speed bump, not a wall — a determined human clicks through — but it filters the laziest attacks and gives you a defined moment to state norms.

Onboarding. Discord’s onboarding flows shape where new members land and what they see first. Done well, newcomers end up in a limited set of channels instead of everywhere at once — which shrinks the blast radius if a bad actor does get in.

Discord also maintains a safety hub with guidance for server owners, including material on handling harassment and abuse. It’s worth reading once when things are calm, not for the first time mid-raid.

The honest framing: native controls raise the cost of raiding you. They won’t stop a targeted attack with aged accounts, and they won’t fix a leaked invite. That’s why they’re layer 3, not the whole plan.

Layer 4 — AutoMod, permissions, and rate limits

This layer is about limiting the damage of whoever does get through.

AutoMod. Discord’s native AutoMod lets you define rules that intercept messages — keyword blocks, spam-style filtering, and related controls. The rule types available have expanded over time; set them up from your server settings and check Discord’s support site for what’s currently offered. Configure it for your actual failure modes: if raids arrive with a specific scam phrase, that phrase belongs in a filter today.

Permissions for newcomers. Design default permissions for your worst day, not your best day:

• New members shouldn’t be able to post in every channel by default
• @everyone and @here should be locked down hard — mass pings are a raid’s favorite amplifier
• Media, links, and embeds can be restricted for fresh members if your server type allows it

Slow mode and rate limits. Per-channel slow mode throttles how fast any one account can post. Combined with limited default channels, it turns a spam flood into a slow drip your mods can actually see and act on.

None of this is exotic. It’s the difference between a raid that owns your whole server in ninety seconds and one that sputters in a single channel where a moderator can contain it.

Layer 5 — The in-raid response playbook

Layers 1–4 are prevention. When a raid lands anyway, you need a written response your mods can execute without a meeting. The order matters:

1. Stop the inflow first. Use whatever server-level emergency controls your server currently has — Discord has shipped controls for pausing new joins and similar emergency measures; find them in your server’s safety and community settings, and confirm the exact behavior on Discord’s support site while things are calm. Then revoke the invite link being abused — if joins are still arriving, an open link is still out there.
2. Raise the drawbridge. Temporarily tighten verification and screening, and restrict posting in the affected channels. You can relax this after the attack; you can’t unsend what got posted while you hesitated.
3. Contain, don’t obliterate. Freeze the affected channels’ permissions. Resist the urge to delete channels wholesale — destruction feels productive and destroys evidence.
4. Capture evidence before cleanup. Your server’s audit log is the record of who joined, who was invited by which link, and who acted. Screenshot the worst of it. When you ban, ban on what you can verify, not on vibes.
5. Communicate once. A single pinned message for real members: what happened, what you’re doing, what they should ignore. Silence breeds screenshots and panic; one calm message ends most of it.
6. Clean up after stabilization. Purge the spam, lift the temporary restrictions, and restore normal permissions deliberately.
7. Run the post-mortem. Every raid tells you which layer failed. A leaked link? Layer 1. No screening? Layer 2. Permissions too open? Layer 4. Fix that layer before the next one — there is usually a next one.

Why open invite links fail structurally

Zoom out and the pattern is clear. An open, permanent invite link is not a growth tool — it’s an unscreened surface exposed to the entire internet:

• Scrapers harvest invite URLs from public pages automatically, so obscurity is not protection
• Directories and listing sites republish your link into contexts you never chose
• Leaks compound — every reshare multiplies copies you don’t know about and can’t revoke
• The link carries zero signal — it cannot tell an excited future regular from a bot, and it never expires on its own

Every other platform treats this the same way. WhatsApp’s raw group links have the identical disease — we broke down the structural version of this problem in our piece on the WhatsApp community member limit — and the fix is identical too: own the door instead of handing out keys.

Why “more moderators” is not the fix

It’s the instinct every community reaches for first, and it fails for a simple reason: raids scale faster than humans do. A raid adds dozens of accounts per minute; a moderator reads, judges, and acts in seconds per case, needs sleep, and has a timezone. You cannot out-staff an automation problem.

There’s also a burnout math nobody budgets for. Moderators who spend their shifts mopping spam leave. The ones who stay get cynical, start banning too broadly, and the community’s tone curdles. You’ve paid for the raid twice: once in damage, once in the people who cleaned it up.

Moderators are essential — someone has to run layer 5. But the goal is to give them a playbook and a server where raids arrive slow, few, and visible. Architecture beats headcount. The servers that feel “well-moderated” are usually the ones whose design means there isn’t much to moderate.

A practical decision tree

Situation	Next move
Raw invite link posted publicly somewhere	Rotate it; put a page you control in front of the new one
Bots joining with no public link you know of	Audit invite usage to find the leak; revoke that link
Raid happening right now	Pause joins, lock affected channels, raise screening — in that order
Spam flood in one channel	Freeze the channel’s permissions before any cleanup; keep the audit log
Want screening without any tool	Application form + manual approval before the invite is shared
Approvals becoming a daily chore	Move the gate to a hosted join page with automated checks

How Inviter fits

Inviter is a hosted front door for communities that stay on Discord: you publish a join page, collect applications with manual or auto approval, email verification, captcha, and country allow/deny rules, and connect the server you already run. Inviter is the front door, not the house — your Discord stays your Discord. It’s free to start; Pro is $19.99/month or $191.90/year, with details on pricing. It also works with Telegram, WhatsApp, and Slack if your community spans platforms, and 1,200+ communities use it. Browse /communities/discord for public examples of the pattern.

It’s optional. Everything in layers 1, 3, 4, and 5 works without it, and a manual gate covers layer 2 for small servers. Inviter earns its place when screening has to scale past one mod with a spreadsheet.

Make the door boring

Nobody writes case studies about servers that never get raided, which is the point: good discord raid protection is uneventful by design. Expired and rotated links. A gate where intent gets checked. Discord’s own verification and screening turned on. Permissions built for the worst day. A playbook every mod can run cold.

You don’t need to out-type the bots or out-hire the attackers. You need five boring layers, each one making the join path a little more expensive for anyone who isn’t actually here to join. Build them once, test the playbook once, and let the raids be someone else’s emergency.