What are ChatGPT dots?
A dot is a personal agent in ChatGPT that keeps working between your conversations. In a regular chat, ChatGPT stops as soon as it has given its answer. A dot picks up ongoing work, keeps track of where it left off and carries on while you do something else. OpenAI announced dots at DevDay on 29 September 2026, in the launch blog.
A dot has its own name in the form @yourname-dot. You talk to it in ChatGPT, and also in Slack and Microsoft Teams.
Dots are part of a new wave of always-on agents. Claude Duet from Anthropic and Grok Bot from xAI work in a similar way. If you want to get the concept clear first, read what an AI agent is.
How does a dot work?
Its own computer in the cloud
Each dot has its own computer and browser in the cloud. That means your laptop and phone can be switched off. You can watch what the dot is doing, take control with "Take over" and then hand it back. Access to your own computer is off by default. If you turn it on through the desktop app, the dot can work on one computer at a time.
A dot remembers what is going on through ChatGPT's memory and its own notes. It can run several tasks side by side and do its own research. That research only uses tools that read and change nothing. According to OpenAI, this limit is enforced in the code.
Models
At launch, dots ran on GPT-6 Astra. Since 3 November 2026, according to OpenAI, a dot can also use GPT-6.1 Terra and use a different model for each part of a task. You do not choose the model yourself.
Where do you talk to your dot?
- In ChatGPT on desktop, web and the mobile app, including by voice. Mobile through the browser does not work.
- In Slack and in Microsoft Teams. Teams is still in limited availability.
- By SMS or iMessage, since 28 October 2026 as a beta for a limited group of Pro users in the US.
Since early October, a dot can also start work in Codex, OpenAI's coding agent, and read and change automations in ChatGPT Work.
Which apps can you connect?
A dot works with the same plugins as ChatGPT, ChatGPT Work and Codex. In the launch blog, OpenAI mentions more than 4,000 apps. Examples from OpenAI's documentation are Gmail, Google Drive, Outlook, Slack and GitHub. Exactly which apps you see depends on your plan and region.
What a dot cannot do yet: it has no email address of its own and cannot call you.
What does a dot cost?
The first dot comes at no extra cost with ChatGPT Pro and Business Premium. You cannot buy extra dots yet; OpenAI says that will come later.
Conversations with your dot do not count towards your usage limit in ChatGPT. Tasks your dot passes on to Codex or ChatGPT Work do count. Higher limits applied in the first month. On OpenAI's pricing page, a Business Premium seat costs €105 per month with annual billing and €130 per month with monthly billing (as of 6 November 2026).
For Enterprise, Edu and Healthcare, OpenAI charges a fixed rate per type of work or a usage-based rate. OpenAI does not publish the amounts. A free preview for admins runs until 18 November 2026; after that, dots are on by default and billing starts, unless the admin turns them off.
Can you use ChatGPT dots in the Netherlands?
That depends on your plan.
| Plan | Dots in the Netherlands? |
|---|---|
| Free, Go, Plus | No, dots are not included in these plans |
| Pro | No, excluded in the EEA, Switzerland and the UK |
| Business Premium | Yes, in supported regions; the European pricing page lists dots with Premium seats |
| Enterprise, Edu, Healthcare | Yes, generally available since 4 November 2026 (source) |
For Pro, OpenAI gives no reason and no date for Europe. According to the help page, Business Premium is available "in supported ChatGPT regions", with no exception for Europe; the Netherlands is not mentioned separately there. With Enterprise, you can have a dot's data stored and processed in Europe (as of 6 November 2026).
To see how this compares with Claude Duet and Grok Bot, read the comparison of the three always-on agents.
How safe are dots?
OpenAI describes the measures in a FAQ on privacy and safety and a separate blog.
Rules you set
For each type of action, you choose how the dot behaves:
- Act on its own without asking
- Act when you tell it to
- Ask first
- Hand over to you
Rules that always apply
- Changing passwords and transferring money always go back to you.
- Permanent deletion and installing software ask for approval every time.
- Purchases with a card saved on a website ask for approval.
- A separate system checks planned actions. Your own rules cannot turn that off.
- Saved passwords stay out of the model's reach.
In the desktop app, an activity overview shows what your dot has done, and you can pause it. Admins of a Business or Enterprise workspace see all dots, can pause or delete them and have audit logs.
Your data
- With Business, Enterprise and Edu, OpenAI does not train on your data by default.
- With personal plans, the "Improve the model for everyone" setting applies, including to your dot's actions and the data from connected apps.
- A dot's context exists for as long as the dot exists. You cannot view or erase a dot's individual memories.
- If you disconnect an app, what the dot has already learned is kept.
OpenAI itself adds that dots can make mistakes and that the protection against prompt injection reduces the risk without removing it. Prompt injection is an attack where a web page or email contains hidden instructions that try to mislead an agent.
What do dots mean for your business?
Of the three big always-on agents, dots are the only one where your data can stay in Europe, through Enterprise. Claude Duet is not yet available here and Grok Bot runs in the US. For many Dutch businesses, that makes dots the first always-on agent they can actually use. It also makes the question concrete: what should such an agent be allowed to do?
- Start with reading and preparing. Have the dot do research, write summaries and draft documents first. Set sending and changing to "ask first".
- Connect with care. Every connected app gives the dot access to data, often including customer data. Under the GDPR, the European privacy law, you remain responsible for that personal data.
- Sort it out internally. If employees work with dots, that falls under the AI literacy obligation in Article 4 of the EU AI Act, which has applied since 2 February 2025. Set out in an AI policy which actions a dot may take on its own. A one-on-one AI training is one such measure. This is general information, not legal advice.
A dot is a general agent for knowledge work. If you want an agent for one fixed process, such as quotes or customer questions, you can have an AI agent built. AI agent vs ChatGPT explains where the difference lies.
Frequently asked questions
What is a ChatGPT dot? A dot is an always-on agent in ChatGPT with its own computer and browser in the cloud. It keeps working between your conversations, even when your devices are off, and you talk to it in ChatGPT, Slack or Microsoft Teams.
What does a ChatGPT dot cost? The first dot comes at no extra cost with ChatGPT Pro and Business Premium. According to OpenAI, a Business Premium seat costs €105 per month with annual billing (as of 6 November 2026). Extra dots are not for sale yet.
Can I use ChatGPT dots in the Netherlands? Yes, through Business Premium and Enterprise. With a personal Pro plan you cannot: it is excluded in the EEA, Switzerland and the UK. OpenAI gives no date for Pro in Europe.
Which model do dots run on? At launch, dots ran on GPT-6 Astra. Since 3 November 2026, a dot can also use GPT-6.1 Terra and use a different model for each part of a task.
Does a dot do things without asking? You set that yourself for each type of action. Changing passwords and transferring money always go back to you, and permanent deletion always asks for approval.
Does OpenAI train on my dot's data? With Business, Enterprise and Edu, not by default. With personal plans, it depends on the "Improve the model for everyone" setting, which also applies to your dot's actions.
Want to learn to work well with ChatGPT and agents?
A dot only does useful work when you know which tasks to give it and which rules to set. In the one-on-one AI training you set up ChatGPT or Claude for your own work in a single half-day session: Advanced for €449 excluding VAT, or Specialist for €579 excluding VAT if you want to work with agents and integrations. After the training you get a 30-minute follow-up call, so you have somewhere to go with your questions once your dot is at work.
