---
title: "How Do You Use ChatGPT Dots? 3 Background Agent Prompts"
url: "https://goremy.ai/blog/articles/chatgpt-dots-agent-prompts-how-to-use"
author: Luis Chavez-Mattos
editor: Luis Chavez-Mattos
editor_title: Director, Product
published: "2026-10-25T19:40:06.440Z"
updated: "2026-10-25T19:40:08.834Z"
topic: Always-On Background AI Agents
description: ChatGPT Dots are always-on agents with their own computer and connected apps. Here are three prompts for price drops, flight delays and email commitments.
---

# How Do You Use ChatGPT Dots? 3 Background Agent Prompts

> ChatGPT Dots are always-on agents with their own computer and connected apps. Here are three prompts for price drops, flight delays and email commitments.

Edited by Luis Chavez-Mattos, Director, Product · Published 2026-10-25

## What are ChatGPT Dots and how do you use them?

ChatGPT Dots are always-on AI agents. Each one can use its own computer and your connected apps to get work done in the background. You use a Dot by giving it a standing responsibility in plain language, such as "watch my flights" or "track my purchases," instead of a one-off question. Because it keeps working without repeated reminders, the prompt you write acts more like a job description than a chat message.

## TL;DR

- Dots are **always-on agents** that work in the background using their own computer and your connected apps.
- A Dot prompt works best as a **standing instruction** with a trigger, an action, and a point where it stops and asks you.
- The first prompt tracks recent purchases for **price drops and refund policies**, then queues price adjustment requests for your approval.
- The second prompt monitors upcoming flights and, after a delay or missed connection, **finds alternative routes and compares costs**.
- The third prompt scans your email and calendar for **commitments you made** and reminds you before they are due.
- The source material does not cover pricing, plan availability, or limits, so check those in the product before relying on a Dot.

## How are Dots different from a normal ChatGPT prompt?

A normal prompt is a single exchange. You ask, it answers, and the work ends. A Dot keeps working on a responsibility after you close the window. You don't have to remind it, and it doesn't need you to come back and ask again.

Two things make that possible, based on how Dots are described. First, each Dot has its own computer, so it can do tasks that need a browser or other software rather than just generating text. Second, it can reach your connected apps, which is where the information it monitors lives: email, calendar, and similar services.

That changes how you should write prompts. You are no longer asking for an answer. You are delegating an ongoing duty.

## How do you write a good Dot prompt?

The three prompts below share a pattern worth copying.

1. **Name the thing to watch.** Purchases, flights, or commitments.
2. **State the trigger.** A price drop, a delay, a due date approaching.
3. **State the action.** Check refund policy, find alternatives, send a reminder.
4. **Keep yourself in the loop.** Ask for approval before anything with money or consequences.

Prompts that skip the fourth step are the ones to be careful with. A good standing instruction tells the agent when to stop and ask.

## How do you track purchases for price drops and refunds?

This prompt turns a Dot into a post-purchase watchdog. If something you bought gets cheaper, the Dot helps you get money back.

> Track my recent purchases for price drops, check refund policies, and track price adjustment requests for me to approve.

Here is what each part does:

- **Track recent purchases.** The Dot needs a source of purchase history, which usually means a connected email account with order confirmations.
- **Check refund policies.** Price adjustment rules vary by store. A price drop is only useful if the retailer will honor a refund or adjustment, and often only within a set window.
- **Track requests for me to approve.** This is the safety rail. The Dot prepares the request, and you decide whether it goes out.

You can tighten the prompt for your own habits. For example, you might add a minimum savings threshold so you aren't pinged over small differences, or limit it to purchases from the past month. Those are your own choices layered onto the base prompt.

## How do you monitor flights and handle delays?

The second prompt gives you a backup plan before you need one.

> Monitor my upcoming flights and if one gets delayed or I miss my connection, find alternative routes and compare costs.

The trigger here is an event, either a delay or a missed connection. The action is research: find other routes and compare what they cost. That is tedious work to do from an airport gate with a dying phone, and it is exactly the sort of thing a background agent can prepare ahead of time.

Some ways to make this prompt more useful:

- Tell the Dot which trips to watch, or let it pull them from your connected email and calendar.
- Say what you care about most, such as arrival time, price, or avoiding overnight layovers.
- Ask it to present options rather than rebook on its own. Changing a ticket can cost money, and the source prompt asks only for alternatives and cost comparisons.

## How do you keep track of commitments from email and calendar?

The third prompt targets the small promises that slip.

> Check my email and calendar for commitments I've made and remind me before they're due.

This one depends on your connected apps more than the other two. The Dot reads your messages and events, identifies things you said you would do, and reminds you ahead of the due date. The value is in the "before they're due" part. A list of commitments after the fact is just a record. A reminder in time is what prevents the missed deadline.

Since "commitment" is a fuzzy word, you may want to define it. Examples include replies you promised, documents you agreed to send, and follow-ups after meetings. Specific wording helps the agent separate real promises from casual language.

## What do these three tasks have in common?

| Task | What the Dot watches | Action it takes | Where you stay in control |
|---|---|---|---|
| Price drops and refunds | Recent purchases | Checks refund policies, drafts adjustment requests | You approve requests |
| Flight disruptions | Upcoming flights | Finds alternative routes, compares costs | You choose the option |
| Commitments | Email and calendar | Reminds you before due dates | You do the task |

All three are monitoring jobs. Each has a quiet default state and a moment when the agent acts. That is the sweet spot for an always-on agent: work that is boring to check by hand but costly to miss.

## What should you check before handing a Dot your accounts?

Connected apps are what make a Dot useful, and they are also the main risk. Before you set one up, consider a few questions.

- **What can it access?** Email and calendar contain sensitive material. Connect only what the task needs.
- **What can it do without asking?** Keep money-related and booking actions behind approval.
- **How will it reach you?** A reminder you never see is worthless, so confirm where notifications land.
- **What are the limits?** The material here doesn't cover pricing, usage caps, or which plans include Dots. Verify that in the product itself.

Start with one Dot and one task. Watch how it behaves for a week before you add more responsibilities.

## Frequently Asked Questions

### What is a ChatGPT Dot?

A Dot is an always-on AI agent in ChatGPT. It uses its own computer and your connected apps to complete tasks in the background. It keeps working on a responsibility without you reminding it.

### Do I have to keep the chat open for a Dot to work?

The point of a Dot is that it works in the background, so you shouldn't need to keep prompting it. How it schedules and reports its work is something to confirm in the product, since the source material doesn't go into detail.

### Can a Dot get me a refund automatically?

The price-drop prompt asks the Dot to check refund policies and track price adjustment requests for you to approve. That means it prepares the request and you make the final call. Whether a refund is granted depends on the retailer's policy.

### Will a Dot rebook my flight if I miss a connection?

The prompt described here asks the Dot to find alternative routes and compare costs, not to book them. If you want it to stop at options, say so in the prompt.

### Are these the only prompts I can use?

No. They are three examples of a pattern: a watched source, a trigger, an action, and a point of approval. You can apply the same structure to other recurring responsibilities.
