How to vet new members at the door

Start here, because most admins are looking for something that does not exist. Discord does not publish a platform-wide ban list: a ban it issues closes the account, and there is no record of it queryable from outside. A ban issued by a server lives only in that server’s own records — another server cannot see it.

Discord equally does not expose a “this account is an alt of that one” field. The user object carries account age, badges and premium status; it carries no alt-account relationship. So alt detection is not an API call, it is reading signals together.

The practical consequence: what you can do at the door is not read an answer the platform provides. It is weigh records shared by communities against behavioural signals, side by side.

Order matters. Work from the hardest evidence down to the weakest signal, because no single signal is enough to decide on.

• Cross-server ban records — if another community shared a record, you see its source, date and reason. This is the hardest evidence, because it is a record rather than somebody’s claim.
• Cheat-network membership — holding a customer or staff role in a cheat seller’s server. That is not an accusation, it is a relationship indicator.
• Alt-account links — shared game identifiers, the same avatar, overlapping servers. These produce probability, not certainty.
• Account age and history — a newly created account has no history. That is not a signal, it is a limit on what the other checks can return.

Most servers conflate these three and try to close all of them with one tool. They ask different questions.

At application time the person is not inside yet, and batch checking is most efficient: screening an ID list from one application form at once is both faster and more consistent than checking them one by one. At join time the question changes — they are arriving now, so the decision has to be instant and the gate has to be automatic. Once they are inside, the question turns historical: this person has been here a while, so how do their behaviour and their record read together?

As volume grows, the real mistake is trying to read the results one at a time. The right move is to separate hard evidence first: those with a record, those without, and those carrying a link signal.

Those with a record are fast — it has a source and a date, and that is usually enough to decide. Those without are not “clean”: an absent record means they have never entered a community that keeps records. Those carrying a link signal need human review; a signal is grounds for a decision, not the decision.

• Batch screening tool (10 IDs, no account)
• Batch checking guide

This is the section a company selling screening has to write too. Direct messages are invisible. Unmonitored servers are invisible. Closed communities are invisible. No query ever shows an IP address. And a record is the claim of the community that shared it — Argos carries it, it does not guarantee it.

Alt detection produces no certainty either. A link is grounds for a decision, not the decision itself. And an empty result means “no record”, not “clean”. Keeping that distinction in writing reduces both appeals and wrongful rejections.