Skip to content

BlogGuides

Discord Support Glossary: Every Term, Honestly Defined

The vocabulary of Discord support, defined precisely: tickets and tokens, budgets and gates, the metrics, and the marketing words that need a second look.

Dani, Founder, AI Ticket Bot

15 min read

Every term on this page will eventually appear in front of you as a claim: on a pricing page, in an analytics tab, in a sales pitch for a bot you are comparing. Words like unlimited, instant and self-improving do real work in those sentences, and the work they do is mostly hiding a number. So this glossary does something slightly unusual for the genre: each definition is written to be testable. Not what a word vaguely gestures at, but what it has to mean before somebody may honestly use it.

How this glossary works

The terms are grouped by the job they do: running tickets, staffing them, paying for AI, measuring the result, and reading a pricing page. Where a word already has a full article on this site, the entry here stays short and points at it, because a glossary that re-argues everything it links is just a longer way to say the same thing.

Every term on this page is a claim somebody will make to you. A definition is what that claim has to mean before you believe it.

One deliberate omission up front: the five words every ticket conversation is built from, panel, category, claim, transcript and escalation, are defined where they live, in what is a Discord ticket bot. That article explains the machinery those words name. This page covers everything the machinery gets talked about with.

Where the deep treatments live

Panel, category, claim, transcript and escalation: what is a Discord ticket bot, which this glossary deliberately does not repeat.

Escalation as a practice, priorities and routing included: Discord support escalation.

Deflection: how to reduce support ticket volume owns the word, and spends most of its length arguing for fewer tickets rather than more.

Resolved, deflected, escalated and first response time as metrics, with honest numbers: do AI ticket bots actually work.

Memory, training, fine-tuning and retrieval: what a Discord AI bot actually knows and how to train an AI ticket bot, with the closed-ticket route in an AI that learns from your tickets.

Free plan, free trial, watermark, self-hosted and overage: free Discord ticket bots.

Webhook, receiver, signature and the rest of that family: Discord ticket bot webhooks.

Working hours, inactivity window, nudge, grace period and sweep: working hours and auto-close settings.

Modal and required field: ticket form questions. Responsible role, transfer and inherit: ticket categories.

Ticket spam, sniping and impersonation: how to stop spam tickets.

Transcript retention: Discord ticket transcripts.

The ticket words

The lifecycle every support request moves through, whatever bot runs it:

  1. Opened
  2. First reply
  3. Resolved or escalated
  4. Closed
  5. Transcript kept
Ticket
One member's support conversation in a room of its own: a private channel or thread the bot creates on demand and removes at close. The base unit everything else on this page is counted in
Channel mode
The ticket is a real Discord channel with permission rules written onto it. Private by construction, but it spends from Discord's limit of 500 channels per server
Thread mode
The ticket is a private thread instead. It costs no channels, but its visibility follows the channel it lives in, which is the classic way thread setups break
Ticket number
Counts per server, not per panel. Ticket #412 is the 412th ticket opened anywhere in your server, and two members pressing the button at once still get different numbers
Priority
A label that orders the queue by urgency instead of arrival. Staff set it with a command, and when an AI hands a ticket over it attaches one itself: low, normal or high
Close request
Staff asking the opener to confirm a close instead of closing over them. A yes closes the ticket, any new message from the opener cancels the request, and silence lets it close when the countdown the staff member chose runs out
Bulk close
Closing many stale tickets in one action, up to 100 at a time here. Each one still gets its own transcript and its own notice to the member, which is the part worth checking on any bot offering it
Reopen
A path most ticket bots, ours included, deliberately do not have. The same member coming back gets a new ticket with a new number, and the old one stays closed with its transcript intact

Two of these deserve the longer treatment they have elsewhere. What actually happens at close, and why there is no undo for it, is how to close a Discord ticket. Closing tickets automatically after silence, and the timers that warn people first, is auto-close settings.

And one term from the edge of the category: when a different ticket bot creates the room and our AI answers inside it, that is an external ticket. The setup, and the honest limits that come with it, are in add AI to any Discord ticket bot.

The staffing words

Staff role
A Discord role wired into the bot with its own switches: whether people holding it can see tickets, take ownership of them, close them, or teach the AI. Each switch is separate, per role
Quick response
A saved reply staff insert with a short command. It comes out word for word every time, which is the point: policy text should not be improvised
Log channel
A channel where the bot posts a notice for every ticket event: opened, claimed, closed, transferred, including the close reason. Free here, and the right tool when your staff just need a feed
Blacklist
Blocking a person, or everyone holding a role, from opening new tickets. It is not a mute and it touches nothing else: an existing ticket stays visible to them until it closes

The distinction hiding in this section is between telling people and telling software. A log channel tells your staff what happened. When a machine needs to know, that is a webhook, and Discord ticket bot webhooks covers that whole vocabulary, signature checks included. Taking ownership of a ticket, which Discord support calls claiming, is one of the five building-block terms defined in the ticket bot explainer. And the vocabulary of people abusing the ticket button, spam, sniping and impersonation, has its own article in how to stop spam tickets.

What is an AI token?

The word that confuses the most buyers, because it prices something invisible. A token is a small piece of text, a few characters, and it is the unit AI models read and write in. Every message in a ticket the AI handles spends tokens twice over: once for what it reads, once for what it writes. Long conversations spend more than short ones, which is why serious AI products meter tokens rather than counting replies. A one-line answer and a two-hour troubleshooting session are not the same work, and a price that pretends they are is hiding the difference somewhere else.

AI token
The metered unit of AI work: a few characters of text read or written by the model. Both the question and the answer spend them, so a conversation's cost scales with its length
Token budget
The daily and monthly token allowance a plan includes. When it runs out, the AI pauses and tickets carry on manually. It never converts into a surprise bill
Token pack
A one-time purchase of extra tokens on top of the budget. It tops up the meter without changing the plan
AI-resolved
A ticket the AI closed itself, with no handoff to a human at any point. Our share sits around half of everything handled, and the live count is public
Knowledge base
The pool of answers the AI may draw from. Shared across every server in some products; per server here, holding only what your staff deliberately taught
FAQ channel
One public channel where the AI answers in the open, so the same three questions stop becoming tickets in the first place
Hallucination
An AI stating something false with confidence. Every AI product carries the risk; the honest question for a vendor is what the AI answers from, and when it must hand over instead of guessing

The token budget's failure behaviour is the part worth reading twice, on any bot. Here, hitting the cap pauses the AI cleanly: the panel keeps working, staff keep answering, and nothing breaks or bills. If a vendor cannot tell you what happens at their cap, assume you will find out the hard way. What the numbers translate to in practice, plan by plan, is worked through in how much does an AI ticket bot cost.

Two more behaviours belong to this vocabulary without needing their own entries. Our AI runs on Anthropic's Claude models through NEXUS, our own AI platform built for Discord communities. And when a question genuinely cannot be answered from what staff taught, an outside service's status for example, the AI can check the live web as a fallback: your taught facts always win over anything found there, and only the search terms leave the conversation, never the ticket itself. You can switch that off per server.

What the AI knows, and just as importantly what it never sees, is what a Discord AI bot actually knows. How the knowledge gets there is how to train an AI ticket bot. When it should stop answering and bring in a person is Discord support escalation.

How the AI-resolved figure is produced

Sample
Every ticket the bot has handled since launch, the original version and the current rebuild combined, more than 180,000 at the time of writing
Window
Lifetime through August 2026, refreshed continuously at the live endpoint linked at the foot of this page
Definition
AI-resolved counts a ticket the AI closed itself with no human handoff. A ticket any staff member stepped into does not count

The measuring words

Resolution time
How long a ticket stays open, from first message to close. It moves independently of how fast the first reply came, and conflating the two flatters slow teams
Backlog
The tickets open and unanswered right now. The number that grows quietly until somebody notices, and the first thing to check before blaming volume
Ticket volume
How many tickets arrive in a period. The number worth reducing rather than staffing for, which is a whole discipline of its own
CSAT
A satisfaction score members give after a ticket, usually out of five. Common in helpdesk products, rare on Discord: most bots, ours included, do not run surveys
SLA
A promised ceiling on how long a member waits. In a community it is usually a promise you make to yourself rather than a feature you buy, and the nearest real features are schedules and reminders

The metric missing from that block on purpose is first response time, the number members actually feel, and its precise definition already lives in do AI ticket bots actually work alongside the honest resolution numbers. Deflection, the art of a question never becoming a ticket at all, belongs to how to reduce support ticket volume. And reading all of these off a dashboard without fooling yourself is Discord support analytics.

If a vendor quotes any of these numbers about their own product, the two-word test is: says who? A metric with no stated definition and no place to check it is an adjective wearing digits.

The buying words

Plan limit
A cap on a number: panels, staff roles, quick responses. The feature exists on every plan and the ceiling moves when you upgrade
Feature gate
A capability that exists only on some plans, such as forms at the door, working hours or custom branding. You either have it or you do not
Token pack
The one-shot top-up from the AI section, listed twice because it is the answer to a buying question: a busy month does not require a bigger plan
Whitelabel
The bot appearing under your server's own name and avatar instead of the vendor's, applied per server through Discord's official bot profile system
Website chat widget
The same per-server AI answering on your own website, handing the visitor into a private Discord thread when a human is needed
Per-seat pricing
Being charged per staff member who answers. Standard in helpdesk products, rare in Discord bots, and worth ruling out before your team grows

The limit-versus-gate distinction is the one that decides arguments, so here it is against our own live catalogue:

A limit: the number moves

  • Panels: 1 free, then 5, then 15
  • Quick responses: 5, then 50, then 200
  • Staff roles: 3, then 10, then 25
  • AI tokens: a budget on every plan, free included

A gate: the capability appears

  • Forms at the door: paid plans only
  • Working hours: paid plans only
  • Whitelabel branding: paid plans only
  • Website chat widget: paid plans only

The rest of the buying vocabulary, free plan against free trial, watermark, overage and what self-hosting really costs, is defined in free Discord ticket bots, and the deeper branding treatment is whitelabel Discord ticket bot. The widget has a full article too: an AI support widget for your website.

How do helpdesk terms map to Discord?

Plenty of people arrive at Discord support from helpdesk products, and the two worlds use different words for the same things. The mapping, and where it genuinely is not the same thing:

Discord support saysHelpdesk products sayThe difference that matters
TicketCase, conversationA ticket here is a real Discord channel or thread, created on demand and deleted at close
PanelContact form, portalIt lives inside your server where members already are. Nobody is sent to a website
Quick responseMacro, canned replyThe same idea exactly: a saved answer, inserted word for word
Staff roleAgent, seatAccess comes from a Discord role you wire once, not from a per-person licence
TranscriptConversation historyThe channel is deleted when the ticket closes. The transcript is the copy that survives
EscalationHandoff, tier twoOn Discord it means notifying one wired role, not moving a case between queues

The row that surprises helpdesk people is the first one. In their world a case is a row in a database; here it is a place, with the permissions of a place. Most of what feels different about Discord support follows from that, and what is a Discord ticket bot unpacks it properly.

Which marketing words deserve a second look?

Not accusations, a test. A claim survives it when it names a unit, an owner, or a place to check. These pairs are the same promise written both ways:

  • Replies in seconds, with the live average published

    A speed claim with an endpoint behind it is checkable, and being wrong would show

  • Instant, human-like AI

    Neither word names anything you could measure after installing it

  • The AI pauses at your token cap and tickets continue manually

    States the failure behaviour, which is the part you will actually live with

  • Unlimited AI replies

    Somebody pays for every reply. Find the unit that is metered, and what happens when it runs out

  • Learns when a person picks a closed ticket and submits it

    Deliberate learning has an owner, a cost and an off switch

  • Trained automatically on your entire server

    Ask what was ingested and who approved it. Automatic ingestion of member messages is a privacy question wearing a feature's clothes

We hold our own copy to the same test, which is why this site says the AI "learns from your real tickets" and means exactly that: a person picks one closed ticket where staff corrected the AI, submits it, and the correction sticks. One ticket per action, nothing in the background, described in full in an AI that learns from your tickets.

What this glossary leaves out

Four things, on purpose.

The building blocks are defined elsewhere, as said at the top: panel, category, claim, transcript and escalation belong to the ticket bot explainer, and repeating them here would just give search engines two half-copies of the same page.

Helpdesk-suite vocabulary as a system, views, triggers, routing rules in the email sense, is out of scope because the product is out of that scope: this is Discord plus one website widget, with no email, SMS, phone or Telegram side to have words for.

How our AI's memory works internally is not documented here or anywhere public, deliberately. What is published is the boundary: exactly what the AI can see, what it can never see, and what leaves the conversation, all of it in what a Discord AI bot actually knows.

And Discord's own vocabulary, roles, threads, what Discord itself calls a category, is Discord's to define. One collision is worth flagging: a Discord category is a container of channels, while a ticket category is a button on a panel, and ticket categories untangles the two.

Use this page for

  • Reading any bot's pricing page without taking its words on faith
  • Naming what your analytics are actually counting
  • Translating helpdesk vocabulary into Discord reality
  • A map of where every deep treatment on this site lives

Do not expect from it

  • Definitions for email helpdesk suites and their automation vocabulary
  • A tutorial: nothing on this page sets anything up
  • Internals of how the AI memory works, which are deliberately not public
  • Discord's own primitives, which Discord documents better than a bot vendor should

Check the words against a live product

A glossary written by a vendor should end by exposing itself to its own test, so here are the claims this page leans on, with the places they can be checked:

180,000+
tickets handled over the bot's lifetime
about half
resolved by the AI with no human stepping in
3,400+
Discord servers running it
in seconds
typical first reply

Frequently asked questions

A token is the unit AI usage is measured in: a small piece of text, a few characters long, that the model reads or writes. Both the member's question and the AI's answer spend tokens, and longer conversations spend more. Plans meter AI work in tokens because replies vary enormously in size. On AI Ticket Bot, every plan includes a daily and a monthly token budget, and when the budget runs out the AI pauses while tickets keep working manually.

Because replies are not the same size. A one-line answer and a long troubleshooting exchange can differ in cost by a factor of a hundred, so a per-reply price would overcharge simple questions and undercharge hard ones. Tokens meter the actual work done. As a rough anchor, the 9.99 euro plan's monthly budget covers around 200 to 250 average ticket conversations.

Here it means a ticket the AI closed itself, with no handoff to a human at any point. If a staff member stepped in, the ticket does not count. Roughly half of the more than 180,000 tickets handled over the bot's lifetime were resolved this way, and the live count is published at a public statistics endpoint, so the figure is checkable rather than a marketing claim. Vendors define this differently, so ask any bot you compare what its number excludes.

A limit caps a number: how many panels, staff roles or quick responses you can have. The feature exists on every plan and the ceiling moves as you upgrade. A gate is a capability that only exists on some plans, such as forms at the door or working hours. The distinction matters when comparing bots, because you can outgrow a limit gradually, while a gate is something you either have or do not.

The pool of answers an AI support bot is allowed to draw from. In some products it is one shared base serving every server; here each server has its own, and it contains only what that server's staff deliberately taught. Nothing is ingested from your channels automatically, which is worth confirming about any AI product you evaluate.

A quick response is a saved reply a staff member inserts by hand, and it comes out word for word every time, which is exactly what you want for policy text. An AI reply is generated fresh for the specific question, from what the server's staff taught the AI. If the exact wording is the point, use a quick response. If understanding the question is the point, that is what the AI is for.

No. A ticket bot works without any of this vocabulary, and most admins set one up without it. The words start to matter the day you compare two bots, read a pricing page, or try to work out what your analytics are counting, because that is where a loose definition costs money. This page exists so each word has one precise meaning you can hold a vendor to.

It’s not just an AI, it’s your AI.

See it on your own server.

Add the bot free, teach it a few of your most common answers, and watch it clear the repeat tickets on its own.

Free plan, no card. Your first panel starts 14 days of Premium.