OpenAI announced dots at DevDay on September 29, 2026. These persistent agents extend ChatGPT from answering individual prompts to working on goals over time. Access is rolling out gradually on eligible plans and in supported markets. OpenAI's announcement describes cloud computers, connected apps and specialist enterprise pilots.
For brands, the useful question is how to make product information easier for an agent to find, check and compare. Answer Engine Optimisation (AEO) means improving how a business can be discovered and represented in AI-generated answers. Persistent agents add another possible research workflow to that work.
The commercial recommendations below are PingAura's analysis. They are not an OpenAI ranking specification, and no content format or visibility score guarantees selection.
What Are OpenAI Dots?
Dots are personal agents inside ChatGPT that can continue working on delegated goals. OpenAI describes them as powered by GPT-6 Astra, with a cloud computer and browser for research and task execution. They can draw on connected tools and learn preferences supplied by the user. The launch announcement describes connections to more than 4,000 apps. Source: OpenAI.
A dot can gather information, track changes and help prepare work across tools. Its ability to act depends on permissions and the task. Background research does not mean uninterrupted crawling of the entire web, and an agent's memory does not guarantee perfect recall or accuracy.
Where users interact with dots
Initial setup is through desktop web or the desktop app; mobile web setup is unsupported. OpenAI also describes Slack and Microsoft Teams workflows. Texting is a limited beta for US Pro users, rather than a universal channel. Check the current getting-started guide for access and setup requirements.
In an authorised workflow, a team could ask a dot to research a vendor and share findings in a work conversation.
Specialist enterprise pilots
OpenAI is testing specialist dots in focused enterprise pilots across functions such as procurement, finance, customer support and marketing. These pilots are not a generally available, self-service feature. OpenAI is also working with Microsoft on Agent 365 integration; the announcement does not establish that integration as generally live. Source: OpenAI's launch announcement.
For a seller, these workflows could increase the value of accurate specifications, clear documentation and comparable pricing. They do not establish a proven increase in sales for any particular brand.
Launch and Availability
The announcement date and the date an individual user receives access are different. OpenAI's availability guide describes a staged rollout:
| Plan or workspace | Launch access |
|---|---|
| ChatGPT Pro | Gradual rollout in eligible markets; the EEA, Switzerland and UK are excluded at launch |
| Business Premium | Supported across ChatGPT regions, with rollout timing still relevant |
| Enterprise, Edu and Healthcare | Admin-enabled beta access |
Availability and plan terms can change. Use OpenAI's current documentation before making an access or procurement decision.
Persistent agents also overlap with other products. Meta's Muse includes personal tasks and business offerings, so a strict “consumer shopping versus enterprise work” division is misleading. Our companion guide examines Meta Muse and brand visibility.
Why Persistent Agents Matter for Business Research
A team can delegate a goal such as comparing software vendors against a budget, security requirement and integration list. An agent may gather sources, build a shortlist and revisit it when circumstances change. Human reviewers still need to check consequential conclusions.
This makes specific information useful: supported integrations, implementation steps, service boundaries and pricing conditions. A clear product page can help an agent evaluate a requirement. Independent documentation and third-party references may also contribute evidence.
Pre-existing model knowledge is not a documented prerequisite for finding a brand. Retrieval can surface current information. Instead of assuming that an unfamiliar brand will be skipped, focus on whether a researcher can locate reliable information that matches the buyer's needs.
How to Set Up and Use a Dot
Follow OpenAI's getting-started instructions for the current interface and eligibility rules.
| Step | Action | What to review |
|---|---|---|
| Check access | Confirm eligible plan, market and workspace settings | Rollout timing and any admin-enabled beta requirement |
| Create your dot | Use supported desktop web or desktop app setup | The goals and preferences you provide |
| Connect apps | Choose the tools needed for the task | Scope of access and whether changes are permitted |
| Configure rules | Set approval requirements and action boundaries | Consequential actions and any pre-approved work |
| Delegate a goal | Describe the outcome, constraints and deadline | Whether the brief includes enough context |
| Review progress | Inspect findings and activity | Accuracy, source quality and actions needing attention |
Read-only proactive research is different from task execution that can make changes. Dots include permissions, action checks and configurable rules, but can still make mistakes. Review consequential work and connected-app access. OpenAI explains these controls in its safety, security and privacy overview.
Practical AEO Steps for Brands
Publish facts that answer a buyer's constraints
Write useful answers about price, capabilities, eligibility, geography and limitations. Explain what your product supports and when a person should confirm details with your team. Keep comparisons fair and make the underlying sources easy to inspect.
Specific answers can help research. They do not guarantee a recommendation: customer preferences, competing offers, availability and the evidence available to the agent all matter.
Distinguish browsing, search and training crawlers
Review how your site serves ordinary browsers and the search systems relevant to your customers. OpenAI documents OAI-SearchBot separately from GPTBot and user-initiated fetching. Allowing search retrieval is a different decision from permitting training crawlers. OpenAI crawler documentation.
Use accessible navigation, meaningful headings and stable URLs. Check that important product facts are available without broken forms or unnecessary access barriers. A blocked first-party page does not prove that a brand cannot be found through other sources, and opening every AI crawler is not a requirement for this checklist.
Keep evidence consistent across sources
Make sure your website, listings and documentation agree on current features and business details. Cite reliable evidence for factual claims and correct stale information. Structured data can describe information accurately, but it is not a special shortcut to guaranteed agent selection.
Measure what is observable
Identifiable AI referrals, consented analytics, server or bot logs, and conversion evidence can provide partial insight. They measure different things: an automated request is not necessarily a human session, a citation is not necessarily a click, and a referral is not proof of a purchase.
GA4 can help assess observable sessions and conversions. Search Console reports search performance. Neither reveals private agent conversations or completely attributes every background research interaction.
Use PingAura to identify work to improve
PingAura's visibility, citation, site-health and content workflows can help a team organise its AEO work. Prompt research helps explore relevant buyer questions; it should not be interpreted as access to private customer conversations. Product availability and scope depend on the current offering.
| Brand task | Practical purpose | Relevant PingAura workflow |
|---|---|---|
| Review AI visibility | Compare observed answers and identify omissions | Visibility and competitor analysis |
| Explore buyer questions | Cover useful requirements and use cases | Prompt Research |
| Improve verifiable content | Keep product facts clear and current | Article workflows |
| Inspect technical access | Find site issues that hinder reading or retrieval | Site Health |
| Review supporting evidence | Understand observed citations and source gaps | Citation analysis |
| Assess observable traffic | Separate identifiable referrals from uncertain attribution | AI traffic analytics |
These are measurement and improvement workflows, not official OpenAI ranking factors. See PingAura's platform overview for the current offering.
FAQs
What are OpenAI dots?
Dots are persistent personal agents inside ChatGPT that can work on delegated goals using a cloud computer, browser and connected tools. Their capabilities depend on access, permissions and the task. OpenAI's announcement describes the launch.
What model powers dots?
OpenAI describes dots as powered by GPT-6 Astra. That does not imply flawless recall, uninterrupted research or guaranteed accuracy.
Who can use dots?
Access is rolling out gradually on eligible plans and in supported markets. Pro excludes the EEA, Switzerland and UK at launch; Business Premium covers supported ChatGPT regions, while Enterprise, Edu and Healthcare require admin-enabled beta access. Check the availability guide.
Do dots cost extra?
The first primary dot is included in eligible plans under the launch terms. Heavier delegated Work and Codex tasks have usage limits; check current plan guidance rather than assuming unlimited background compute.
How are dots different from Meta Muse?
Both products support personal tasks and have business ambitions. Their integrations, availability and controls differ. See our Muse guide.
How can brands prepare for dots?
Keep product information discoverable, current and verifiable, and answer specific buyer requirements. Those steps can support agent-led research, but OpenAI has not specified a dots-specific ranking formula or guaranteed path to recommendations.
Are dots safe to use?
Dots include permissions, action checks and configurable rules, but can make mistakes. Review consequential outputs and connected-app access, including pre-approved actions. Read OpenAI's safety explanation.
Conclusion
Persistent agents give brands another reason to make their information useful to a researcher. Start with a concrete checklist:
- Check that key product pages load and contain current specifications.
- Explain pricing conditions, limitations and supported integrations.
- Review relevant search-crawler access separately from training permissions.
- Keep third-party business details consistent with your site.
- Track identifiable referrals and conversions while documenting attribution gaps.
Use PingAura to review your observed AI visibility and prioritise the next content or site-health improvement.
