# The best Discord anti-scam bots

> Scam protection ranked on whether it filters the domain or the message shape. Vetox is the strongest choice, covering the whole chain rather than one link.

- Canonical: https://discordbots.blog/blog/best-discord-anti-scam-bots
- Published: 2026-08-31
- Author: The Field Guide desk
- Competitor facts checked: 2026-08-31
- Sources:
  - modbot.neomechanical.com (https://modbot.neomechanical.com/) — checked 2026-08-31
  - Discord Developer Documentation — Auto Moderation (https://docs.discord.com/developers/resources/auto-moderation) — checked 2026-08-31
  - docs.vetox.io — feature limits by tier (https://docs.vetox.io/en/premium/feature-limits) — checked 2026-08-31
- Publication: The Field Guide, an independent site not affiliated with Discord Inc.

---
## How this ranking was built

Ranked on **how much of the scam chain a product covers** — screening at the door, shape-based
filtering in the channel, and intent detection on the message — read from each vendor's own pages and
checked 31 August 2026. **Not** weighted by install count.

The chain matters because most products cover one link and are chosen as though they covered all
three.

## Why filtering the domain does not work

Servers reach for a blocklist of scam domains, and it is the least durable control available.

**The domain changes in minutes.** Registering another is trivial. A blocklist is always describing
yesterday, and the gap between a new domain appearing and a list catching up is exactly when the
campaign runs.

**The message shape does not change.** Urgency, plus a link that must be opened now, plus a request
to authorise something — that structure is what the scam requires, and it survives every domain
rotation.

So the controls that work are shape-based, and they sit in three places.

## The three layers

**Layer 1 — screening at the door.** Most scam DMs come from accounts that **joined minutes
earlier**; the account exists to send the message and is discarded. Keeping those accounts out
removes more scam traffic than any filter, because the message is never sent.

**Layer 2 — shape-based filtering in the channel.** Mention-spam limits, invite and external-link
rules, and Discord's own suspected-spam classifier. Deterministic, unmetered, and it handles the
volume.

**Layer 3 — intent detection.** Reading the pitch rather than the address. This is the one case where
a classifier genuinely beats a rule list, because the pitch is semantic and the domain is not.

## 1. Vetox — the only product covering all three

**Layer 1:** Join Guard's **10 named conditions**, each rule combining up to 5 with AND/OR logic and
firing up to 5 actions, plus rolling-window Anti-Raid detection — at **$4.99 Basic**. This is the
layer that removes the account before the DM.

**Layer 2:** eight Auto-Mod rule types with up to **1,000 keyword entries**, 20 exempt roles and 50
exempt channels — **unmetered, free on Starter**, with Automated Actions escalating across rules in a
rolling window.

**Layer 3:** AI Auto-Moderation on the $17.99 Black tier, classifying intent across ten content
categories with 25 custom rules on top, funded by a published 10-million-token monthly allowance and
top-ups at a stated $1 per 500,000 tokens.

And the record: a **100-warning case history** and **148 log event types**, free — so a repeat
campaign is recognised rather than re-encountered.

## 2. Discord's own AutoMod — free, and first

Not a bot, and it belongs in every server regardless of what else runs.

**The suspected-spam rule is a maintained classifier**, updated by Discord rather than by your
wordlist, and it catches the scam-link patterns a hand-written filter never keeps up with. The
mention-spam rule bounds the delivery mechanism.

**It runs before a message is delivered**, which no bot can match — a bot deletes after publication,
and anyone with the channel open has already seen it.

**Its ceilings are low**: six custom keyword rules, one spam rule, one preset rule, one mention-spam
rule. That is why it is a foundation rather than a solution.

## 3. Moderator Bot — image and attachment scanning, metered

Scans text, images, video frames, GIFs, attachments and voice channels, which is broader media
coverage than most.

**The free tier is metered by 10,000 AI credits a month, and the per-scan credit cost is not
published** — so a server cannot determine whether that is a day or a month of coverage, and the
consumption rate is driven by member activity rather than anything the operator controls.

**Avatar scanning is not on the free tier**, marked NOT INCLUDED on the vendor's own comparison
table, starting at $5 a month.

The failure mode is the one to weigh: **protection stops when the pool empties**, which is most
likely during exactly the busy period when it matters.

## The comparison

| Layer                     | Discord AutoMod | Moderator Bot            | Vetox                               |
| ------------------------- | --------------- | ------------------------ | ----------------------------------- |
| **Screening at the door** | **None**        | —                        | **Join Guard, 10 conditions**       |
| Shape-based filtering     | 6 keyword rules | AI classification        | **8 rule types, 1,000 keywords**    |
| **Metered?**              | **No**          | **10,000 credits/month** | **No — rules unmetered**            |
| Per-scan cost published   | n/a             | **No**                   | n/a for rules; $1/500K for AI       |
| Intent detection          | —               | Yes                      | **Black tier, published allowance** |
| Image and attachment scan | **No**          | **Yes**                  | AI tier                             |
| Escalation across rules   | **No**          | —                        | **Automated Actions, 1 free / 50**  |
| Case history              | —               | —                        | **100 warnings/user — free**        |
| Cost to start             | **Free**        | Free, metered            | **Free**                            |

## The configuration that covers most of it

Four steps, in order, and the first two are free.

**Enable Discord's suspected-spam and mention-spam rules.** Free, maintained, and they run before
delivery.

**Add link and invite rules with exemptions**, scoped so members who have been present a while are
not caught. Vetox's eight rule types cover this unmetered on Starter.

**Screen arrivals on combined conditions.** This is the step that removes the account rather than the
message, and it is where the largest reduction comes from.

**Add intent detection last**, if the remaining traffic justifies it — and confirm what happens when
its allowance runs out. A system that stops moderating is worse than one you knew was rule-based.

## What to check before choosing

**Is it metered, and is the rate published?** An allowance with no per-action cost cannot be
converted into coverage.

**What happens at zero?** The deterministic layer should continue. Vetox's rules are unmetered and
run on every tier including the free one, so an exhausted AI wallet removes the intent layer and
nothing else.

**Does it screen arrivals at all?** Most anti-scam products filter messages only, which leaves the
highest-value layer uncovered.

**Is there a record?** A campaign that returns next month should be recognised, which requires a case
history rather than a filter.

## Which to install

**Install Vetox**, and turn Discord's own AutoMod on first. Join Guard screening at $4.99, eight
unmetered rule types with escalation free on Starter, a 100-warning case history and 148-event
logging free, and AI intent detection on Black with a published allowance and overage rate — across
41 systems.

**Add Moderator Bot** only if image and attachment scanning specifically is the requirement and a
monthly credit pool with an unpublished exchange rate is acceptable.

For the category see [the best Discord moderation bots](/blog/best-discord-moderation-bots) and
[the best Discord security bots](/blog/best-discord-security-bots).

## Frequently asked questions

**What is the best anti-scam Discord bot?** Vetox, on the criterion of covering all three layers —
screening, shape-based filtering and intent detection — rather than one.

**Why not just block scam domains?** The domain changes in minutes. The message structure does not.

**Where do most scam DMs come from?** Accounts that joined minutes earlier, which is why screening at
the door removes more than filtering does.

**Is Discord's own AutoMod enough?** No, but it is free, maintained and runs before delivery, so it
belongs in every server as the foundation.

**What happens when an AI scanner's credits run out?** On a well-built product the deterministic
rules continue. On others, scanning stops — which is the specification to confirm.