/ THE SHORT ANSWER
- 01A dot is designed for an ongoing responsibility, not simply a longer chat.
- 02Its cloud computer can continue work while your device is off, but access to your computer is a separate choice.
- 03Pro and Business Premium access is rolling out; Plus is not listed in the current eligibility guidance.
- 04Approval checks and app permissions still matter. A dot can make mistakes.
/ dotSuper point of view
The useful shift is from asking an AI to answer once to giving it a bounded responsibility that can be inspected, corrected and stopped.
What OpenAI launched
Its description is an agent that takes an ongoing responsibility, works across connected tools, and reaches out with results or decisions that need a person.
The dot runs on GPT-6 Astra and has a cloud computer and browser of its own.
That is a product claim and a launch description, not evidence that every task will run continuously or finish without review.
A better mental model is an accountable assistant with a standing brief.
It can follow up on a project after the conversation ends, but its usefulness depends on the goal, access, rules and feedback you give it.
How a dot differs from a normal ChatGPT chat
A dot can hold an ongoing responsibility.
OpenAI says it can remember relevant context, use connected information, schedule or initiate follow-up work, and return when a decision needs your judgment.
Imagine an operations lead tracking a supplier onboarding project.
A chat can summarize the latest documents today.
A dot could instead keep a decision list current as new approved information arrives, flag a missing certificate, draft a follow-up for review and ask when the lead should approve the next step.
That example is a proposed workflow, not an independently verified performance result.
| Dimension | Normal chat | Dot |
|---|---|---|
| Time horizon | Usually a single request or session | An assigned responsibility across conversations |
| Follow-up | User normally starts the next request | Can track progress and bring back updates |
| Tools | Uses tools available to that chat | Uses permitted connected apps and its cloud computer |
| Control | User reviews the answer | User also defines boundaries and reviews proposed actions |
What it can do, and what it cannot safely replace
These are examples from the vendor, not benchmarks for accuracy or time saved.
The dot's background research of connected apps uses read-only tools.
When work would change an account or share information, the action is checked against permissions, custom rules and built-in safeguards.
Some steps require approval or must be done by the user.
A messaging connection does not itself grant access to an inbox or a local computer.
An always-on agent does not remove the need for a human owner.
The owner still decides whether its source material is current, whether a proposed action is appropriate, and whether an output can be sent, published or used in a consequential decision.
Who can use Dots now?
Enterprise, Edu and Healthcare workspaces can try a beta when their administrator enables it.
The current guidance does not list ChatGPT Plus as eligible.
The first dot is included in eligible Pro or Business Premium plans, with an allowance for deeper work.
Work or Codex tasks started or managed by a dot still count toward those products' usage limits.
Do not assume unlimited background work, universal access or a fixed long-term price.
Check the live eligibility page before changing plans.
A useful first move for a business
Write down the trigger, permitted sources, output format, review owner and stopping rule before connecting anything sensitive.
A weekly competitor-source digest or an internal decision log is a better first test than autonomous customer outreach or payments.
Compare the dot's output with the team's current process for two weeks.
Count accepted outputs, corrections, staff review time and missed changes.
If the review cost is greater than the value, narrow the responsibility or stop the pilot.
The goal is a better operating process, not a new AI subscription for its own sake.
What this page cannot conclude
- 01This article describes OpenAI's launch and documentation as of 1 October 2026. Eligibility, features, prices and limits can change.
- 02OpenAI's demonstrations are illustrative. We have not independently measured Dot accuracy, reliability or return on investment.
- 03A dot can make mistakes, and approval safeguards do not eliminate the need to review consequential work.
Sources
- 01Introducing dotsOpenAI · accessed Oct 1, 2026
- 02Meet dotsChatGPT Learn · accessed Oct 1, 2026
- 03Getting started with your dotOpenAI Help Center · accessed Oct 1, 2026
Our editorial standard · Found an error? Send a correction with its source.
/ CITE OR SHARE THIS GUIDE
Make the evidence easy to verify.
When you reference this guide, link to its canonical URL. That gives readers one stable place for the evidence, limitations and future updates.
dotSuper Research Desk. (October 1, 2026). ChatGPT Dots Explained: What the Always-On Agent Actually Does. dotSuper. https://dotsuper.net/feeds/daily-briefing/chatgpt-dots-always-on-agent-explained