Updated October 2, 2026
OpenAI has announced Dots, personal AI assistants within ChatGPT designed to carry out tasks and keep making progress after you close the conversation. Each dot has its own cloud computer and browser, and can work with apps you connect. The rollout began on September 29, gradually reaching subscribers on selected plans.
For current plans, prices and every Maitz article on ChatGPT, see our ChatGPT hub.
The idea is straightforward: give the assistant a job with several steps, and it works through them, coming back with results, questions or requests for approval. According to OpenAI, Dots can prepare documents, analyze data, develop software and update materials as new information arrives. They run on GPT-6 Astra and can handle several tasks at once.
What can you ask a dot to do?
One of OpenAI’s examples focuses on content creators. After an interview, a dot can review the transcript, identify passages suitable for short videos, and prepare a summary and draft posts. The creator reviews and edits the materials, and the assistant can apply that feedback across the other outputs.
Another example is a product launch: when a detail changes, a dot can prepare updates to the presentation and accompanying copy. For data work, it can update an analysis and its charts when new results arrive, and flag findings that need checking. These are uses presented by the company; we have not independently tested the quality of the work or the time it saves.
Context from previous work together is intended to help over time. According to the documentation, a dot can use instructions, earlier decisions and feedback to adapt to your preferences. To make that useful, it needs to know the result you want, the materials it should rely on, and which decisions should involve you.
Cloud work can continue when your computer is off
A dot’s cloud computer is separate from your personal computer. Work happening in the cloud can therefore continue when your home or office computer is switched off. You can also connect your personal computer for tasks that need its files and software. In that case, it must stay online with the ChatGPT app open.
Account access is not automatic either. Being signed in to a website in your own browser does not mean your dot is signed in there. Access to email, documents and other services depends on the connections and permissions you grant. You could connect a file service, for example, without granting access to all your other accounts.
Initial setup takes place in the ChatGPT desktop app or a desktop browser. You can then use the same dot in the mobile app once the supporting update is available. Mobile browsers are not yet supported. Alongside text and voice conversations in ChatGPT, supported communication channels include Slack and Teams. Connecting a conversation channel does not, by itself, grant access to your mailbox or files.
More capabilities make permissions more important
An assistant that can act in your accounts needs clear limits. According to OpenAI, asking it to prepare a draft does not authorize it to send that draft. You can set rules for particular actions, such as requiring approval before messages are sent to customers. Those rules do not override the built-in safety requirements.
Purchases require approval, and sensitive actions such as changing a password remain in the user’s hands. The company stresses that the system can still make mistakes even with safeguards in place. When account sign-in is needed, the recommendation is to use the secure sign-in flow rather than type a password into the conversation.
ChatGPT’s usual privacy settings also apply. On personal accounts, the “Improve the model for everyone” setting affects whether conversations and work are used to improve the models. Content from Business, Enterprise and Edu plans is not used for training by default.
You can follow progress in the Activity view and change your instructions. Pausing the main task is also different from canceling other work: Pause does not necessarily stop separate tasks or cancel future schedules. Those need to be checked and stopped separately, and an action already completed is not automatically undone.
Who can access Dots, and what does the price include?
As of October 2, 2026, Dots were rolling out gradually to Pro 100, Pro 200 and Pro 500 subscribers aged over 18, outside the European Economic Area, the United Kingdom and Switzerland. Even outside those exclusions, the feature might not yet have appeared in an individual account. The service was also rolling out to Business Premium. For Enterprise, it was available as a beta that a workspace administrator needed to enable.
The first dot is included with Pro and Business Premium at no extra charge beyond the subscription. OpenAI’s pricing page listed Pro plans at US$100, US$200 or US$500 per month. Before upgrading, check that access is available in your account and confirm the final price shown there.
Usage is not unlimited. Conversations with a dot do not count toward ChatGPT usage limits, but deeper work has an allowance, with a temporary expansion during the first month after launch. Tasks a dot starts in Work or Codex count toward those products’ allowances.
Dots promise to reduce the need to switch between files, repeat the same instructions and follow every step yourself. Their practical value will depend on how accurate the results are and how much correction they need. A small, recurring task that you can check against its source material is a useful way to see whether the assistant makes your day easier.
