The AI Is Paused: What Happens When a Discord AI Bot Runs Out of Tokens

Every AI bot has a limit somewhere, because each reply costs the company running it real money. What differs between bots is what happens when you reach it. This page covers what happens on AI Ticket Bot, for the member in the ticket and for the admin behind it.

What does "The AI is paused" mean?

The server's AI has used its allowance for now, and a person will answer. If you are a member and saw this in your ticket, that is the whole story:

Your ticket is still open and your messages are still there. Stay in it. The notice pings the staff role for your ticket, and they answer in the same place. Opening a second ticket will not bring the AI back.

What still works when the AI is paused?

Everything except the AI's replies.

One detail is worth knowing. The notice in a ticket appears when the AI was already answering in that ticket and had to stop. A ticket opened while the AI is paused gets the normal welcome message and no AI reply, so staff should watch new tickets during a pause.

What counts against the allowance?

Every reply the AI writes, wherever it writes it. The allowance is counted in tokens, the unit AI providers bill in. A longer conversation uses more.

There are two allowances and they work together.

On the Free plan, at the time of writing in October 2026, the AI answers more than 300 average tickets a month. Premium is more than 2,000 and Pro more than 11,000. Those went up sharply when the default model became Claude Haiku 5.5, because each ticket now costs far fewer tokens. The pricing page has the current figures.

Tickets, AI FAQ channels, /ai ask, the website chat and training the AI all draw from the same allowance. A server that connects its own AI provider on Pro is the exception: its replies are billed by that provider and do not count.

When does the AI come back?

By itself, when the allowance resets.

/ai usage shows both allowances: how much is used, how much is left, and the exact reset time for each. It also shows the model the server is on and any extra tokens it holds.

The first time a server reaches its limit, we add a one-time gift of extra tokens so the AI keeps answering, and send a direct message explaining what happened. It is given once per server. The next time, the AI pauses as described here.

How do you make the allowance last?

Spend fewer tokens per ticket before buying more tokens. In order of effect:

The first row matters most. If a server moved to a heavier model and now runs out mid-month, moving back is the whole fix. Choosing a model explains the trade.

The second is the one most servers skip. An AI that does not know your rules asks questions, guesses, and hands over late. Training it shortens every conversation. For the last row, see how to stop spam tickets.

Should you buy tokens or upgrade?

A simple test: if the AI paused once because of an event or a launch, a pack covers it. If it pauses every month, the server has outgrown the plan and a pack is the expensive way to keep going.

There is a third option on Pro. A server can connect its own AI provider, and then the AI's replies no longer use the plan's allowance at all.

What the limit does not do

The honest list.

The fourth limit is the one to plan around. If your server relies on the AI answering at night, look at /ai usage once a week. An allowance that is nearly spent by early evening is a sign to change something before members notice.