# What a Discord bot can see, and what to ask before installing

> The data a Discord bot receives by design, what it must ask permission for, and the questions worth answering before you grant it.

- Canonical: https://discordbots.blog/blog/discord-bot-data-practices-to-check
- Published: 2026-09-21
- Author: The Field Guide desk
- Competitor facts checked: 2026-09-01
- Sources:
  - Discord developer docs — Gateway (https://docs.discord.com/developers/events/gateway) — checked 2026-09-01
  - Discord developer docs — Permissions (https://docs.discord.com/developers/topics/permissions) — checked 2026-09-01
  - Discord developer docs — Audit log (https://docs.discord.com/developers/resources/audit-log) — checked 2026-09-01
- Publication: The Field Guide, an independent site not affiliated with Discord Inc.

---
Installing a bot grants a third party structured access to your community. Most of that access is
necessary and unremarkable; the parts worth thinking about are specific and few.

## What it receives by design

**Channels its permissions allow.** Not what it uses — what it *can* see. A bot with Administrator
bypasses channel overwrites entirely, which includes private staff channels, appeal channels and
anything else the role structure was built to protect.

**Server membership.** With the `GUILD_MEMBERS` intent, the full member list with roles and join
dates.

**Message content.** Only with the `MESSAGE_CONTENT` intent. Without it, a bot receives message
events with the text removed. This is the single most meaningful line in a bot's data profile, and
it is visible before you install: the intents are listed in the developer portal and, for verified
bots, reviewed by Discord.

A moderation bot, an anti-spam system or an auto-responder genuinely needs it. A poll bot, a ticket
panel or an economy game that works entirely through slash commands does not.

## What it keeps is entirely the vendor's choice

The platform imposes nothing. Discord's own audit log keeps 45 days of administrative actions; a bot
may keep message content indefinitely, or for an hour, and nothing external decides which.

That makes retention a question you have to ask, and the answer is a fact about the product rather
than an opinion:

- **What is stored?** Message content, or only IDs and counts?
- **For how long?** A stated number, not "as long as necessary".
- **Where?** Which jurisdiction, which provider.
- **Can you delete it?** Is there a purge on removal, and does it cover backups?
- **Can you export it?** Self-service, or a support ticket?

## Why the export question matters most

Every other answer becomes moot if the data is unrecoverable. A vendor with a long retention window
and a working export is a lower risk than one with a short window and no way out, because the first
is a decision you can reverse.

It is also the cheapest thing to verify: run the export once, before you depend on it. A vendor
confident in their product has no reason to make leaving difficult, and the ones that do have told
you something useful.