Useful ChatGPT Dots use cases are narrower than the marketing suggests and wider than a scheduled reminder. The honest way to describe them is by what OpenAI has documented a dot doing, which falls into seven shapes. Each one below is tied to a capability in the Help Center or the announcement rather than to an imagined workflow.
Updated October 2026. Dots launched on 29 September 2026 and are rolling out gradually, so check OpenAI own pages for the current position before relying on any detail here.
1. A standing morning check
The example OpenAI gives in its own documentation is asking a dot to review your calendar each morning and tell you what is coming up. Scheduled runs appear in the conversation, and you manage them from the Scheduled section, where you can change the repeat schedule, the time, or the completion notifications.
2. Research that continues without you
A dot can research in the background and suggest ways it could help. OpenAI calls this proactive research, and it is restricted to read-only tools, so it cannot send messages or change anything while it runs. We cover the boundaries in our guide to proactive research.
3. Work that spans several sessions
The difference from a chat window is persistence. A dot keeps making progress between conversations and remembers context for continuing work, so a project does not have to be re-explained. It can hold several projects at once without you juggling separate threads.
4. Software changes, reviewed by you
In the developer example OpenAI publishes, a dot watches customer feedback for recurring requests, scopes smaller improvements and bugfixes, builds and tests them, and brings complete pull requests to review with attached videos of the changes. It can also create tasks in Codex cloud environments, provided you create the environment in Codex first. Our guide to AI coding assistants covers the wider category.
5. Keeping a document true as facts change
The research example is the clearest version of this shape: as new data arrives, the dot reruns analyses, investigates unexpected results, updates figures, revises the supporting explanation, and flags what needs review. The same pattern covers a proposal that has to track changing requirements.
6. Turning one input into a set of outputs
In the content example, a new interview transcript arrives and the dot identifies moments for clips, prepares show notes and drafts social posts for approval, then carries your edits across the materials. The useful part is the carry-across, not the drafting.
7. Catching the thing you forgot
OpenAI reports that an early tester had a dot notice he had forgotten to invoice a publication, prepare the invoice, and send it after his approval. Note the last three words. This is the shape with the most value and the most risk, and it is exactly where the approval settings earn their keep.
What a dot cannot do for you
Some limits are specific and documented, and they rule out a few use cases people assume:
- It cannot call you. A dot cannot initiate calls at launch, though you can call it.
- It cannot have its own email address at launch. You can connect your personal account for it to use.
- It cannot be created on mobile, and dots are not supported on mobile web. Setup happens on the desktop app or desktop web.
- It cannot change a password for you. Certain sensitive tasks always stay with you.
- It cannot act outside your rules without review, but it can still be wrong, so consequential work needs checking.
For the wider idea these sit inside, see our explainers on what an AI agent is and on agentic AI.
Choosing between ChatGPT Dots use cases
Faced with seven shapes, most people pick the most impressive one and are disappointed. A better way to choose is to score a candidate job on three questions, all of which map onto something the product actually does.
- Is the expensive part gathering or deciding? Dots are strong where the work is assembling context from several places and weak where the work is judgment. Every published example has a person making the call at the end.
- How long is the chain? OpenAI own testing found that doubling the number of intervening tasks from five to ten roughly doubled the rate of flagged scope violations, from 8.6 percent to 19.7 percent. Short, well-bounded jobs behave better than sprawling ones.
- What happens if it is wrong? A wrong summary costs you a read. A wrong invoice costs you a relationship. Keep the second kind behind an approval until you have watched the first kind work.
That ordering also suggests a sensible progression. Start with a scheduled read-only check, because nothing can happen without you. Move to research and drafting, where the output is a document you review. Only then allow actions that touch other people, and set the rules for those explicitly rather than relying on defaults.
Where the ChatGPT Dots use cases are genuinely new
It is worth being precise about what is actually different, because scheduled automation is not new and neither is a chatbot. Three properties are:
- Work continues between conversations, so a project does not restart each time you open it.
- The dot can notice something you did not ask about, which is what the invoice example illustrates, and raise it.
- It reaches you where the work is happening, carrying context between ChatGPT, Slack and Teams rather than keeping separate histories.
Everything else on the list is a faster version of something you could already do. Those three are the ones worth designing a workflow around, and they are also the three that make the control settings matter.
Common questions
What are the best ChatGPT Dots use cases to start with? A scheduled daily check is the lowest-risk start, because it is read-only and you see the output before anything else happens.
Can a dot send an email on my behalf? You can connect a personal email account for it to use, and whether it sends without asking depends on your rules. It cannot have its own address at launch.
Can a dot work on several projects at once? Yes. OpenAI describes a dot taking a project and running with it while working on several others.
Does a dot work while I sleep? It can. It runs on its own cloud computer, so tasks continue when your device is closed or offline.
Can a dot write and test code? The published developer example has it building and testing smaller fixes and returning pull requests for review. It can also create Codex cloud tasks.
Sources and further reading
Where the figures and rules above come from, so you can check them:
- The announcement, including proactive research and the control model: OpenAI, Introducing dots
- Setup, connected apps, scheduling, memory, rules and reset: OpenAI Help Center, Getting started with your dot
- Everything else announced alongside dots: OpenAI, DevDay 2026 Recap
- Appendix B, the red-teaming and alignment results for dots: OpenAI, GPT-6 Astra system card
Photo credits: Home-office-336377 by By Free-Photos from Pixabay, CC0, via Wikimedia Commons. Person using a laptop while holding a telephone receiver by Shixart1985, CC BY 2.0, via Wikimedia Commons. Wikiconf NYC 2010 open space by Boris Mindzak, CC BY-SA 3.0, via Wikimedia Commons.
Join the discussion