ChatGPT Dots: How OpenAI’s Always-On Agents Work and How to Use Them

Last updated: 9 November 2026

What a dot is

ChatGPT used to work one conversation at a time. A dot is different: it is a named, persistent agent that keeps working on your goals when you are not chatting. Think of it as an assistant that has its own computer, remembers what you care about and checks in when it has something for you.

Who can use dots

OpenAI says users will later be able to add more dots and scale each dot’s output, but had not published prices for extra dots. See ChatGPT Plus vs Pro and ChatGPT pricing.

How to set up your first dot

1. Open ChatGPT on desktop or desktop web. Dots cannot be created on mobile.
2. Describe its job. Be specific: “Keep my client project tracker up to date and tell me each morning what is overdue.”
3. Connect only the apps it needs. Start with one or two.
4. Set custom rules. Choose, by type of action, whether it acts without asking, acts only if pre-approved, asks first or hands off to you.
5. Decide on laptop access. Leave it off unless a task really needs local files.
6. Review its first outputs closely and give feedback. Dots learn from it.
7. Check Activity View to see what it has done.

Controls and safeguards

• Proactive research: when you are inactive, a dot can look for ways to help using read-only tools that cannot send messages, change app content or control the browser or computer.
• Custom rules: the four-level approval system above.
• Auto-review: checks actions that could affect accounts or share information.
• Passwords: changing a password always stays with the user.
• Safety monitoring: a system can pause or stop a dot for safety concerns.
• Memory: a dot forms memories from connected apps. Disconnecting an app does not delete what it learned; you need to reset the dot for that.

Enterprise rollout: what admins should do before 18 November

For ChatGPT Enterprise, Edu and Healthcare, admins control access, pausing, usage review and EU or US data residency, and set permissions for browser, network, computer and password manager access per workspace. Because dots switch on by default after the admin preview ends on 18 November 2026:

1. Decide which groups should have dots at first.
2. Set workspace permissions for browser, network, computer and password manager access.
3. Choose data residency if relevant.
4. Update your AI usage policy. See our AI usage policy template.
5. Brief staff on approval rules and what not to delegate.
6. Plan how you will review usage and incidents.

OpenAI says it does not use Business, Enterprise or Edu workspace content for training by default. See AI data privacy for business.

Good first tasks for a dot

OpenAI’s own marketing examples follow the same pattern: the dot drafts, a person approves.

What not to delegate yet

• Payments and anything involving money movement.
• Messages to customers without review.
• Legal or contractual commitments.
• Changes to production systems.
• Anything involving sensitive personal data you are not permitted to share.

Usage limits and costs

Conversations with a dot do not count toward ChatGPT usage limits. Tasks a dot runs in Codex or ChatGPT Work do count. Pro and Business Premium users get extra capacity and extended limits in the first month. For Pro users, OpenAI also halved the Codex and ChatGPT Work usage included in Pro 200 from 30 October 2026, so heavy agent work may hit limits sooner. See ChatGPT message limits.

Dots vs other agents

Specialist dots for businesses

OpenAI is also previewing specialist dots that hold a fixed responsibility in an organisation, with their own identity, credentials and system access. Early testing covers procurement, invoice processing, email marketing, customer support and commercial contracting, in focused enterprise pilots. OpenAI is working with Microsoft to bring these under Microsoft Agent 365 governance, without a date.

What dots mean for brands

When people delegate research to dots, those agents browse the web and compare options for them. Your website, product data and reputation become inputs to an agent’s recommendation. See how to make your website agent-ready and ChatGPT SEO. For the launch event, see OpenAI DevDay 2026.

Writing instructions that make dots useful

A dot works best with a clear job description. A good one covers:

• Goal: what success looks like.
• Scope: which apps, projects or clients it should watch.
• Rhythm: when it should report, such as every weekday at 9am.
• Style: tone, length and format for drafts.
• Boundaries: what it must never do and when it must ask.

Example:

“You manage my content pipeline. Each weekday at 9am, check the content calendar and shared drive, list anything due in the next three days, and draft social posts for any article published yesterday in our brand voice: friendly, plain English, no hype. Ask before posting anything. Never contact clients directly.”

Reviewing a dot’s work

1. Check Activity View daily for the first two weeks.
2. Correct mistakes with specific feedback, such as “Too formal; shorten to two sentences.”
3. Tighten rules where it acted when it should have asked.
4. Loosen rules gradually for tasks it handles well.
5. Reset the dot if it has learned something wrong or you disconnect sensitive apps.

Privacy checklist for personal and business use

• Connect only apps you are allowed to connect under your employer’s or clients’ rules.
• Avoid connecting accounts with highly sensitive data unless necessary.
• Remember that disconnecting an app does not delete what a dot learned; reset the dot to remove it.
• On personal plans, decide whether you allow the dot’s work to be used to improve models.
• Keep laptop access off unless a task truly needs local files.

Common problems and fixes

Frequently asked questions