OpenAI Dots are always-on AI agents (AI that takes actions on your behalf, not just answers questions) that run tasks in the background across more than 4,000 connected apps. Announced at OpenAI’s DevDay keynote, Dots are powered by the GPT-6 Astra model and are OpenAI’s direct response to Meta’s Muse, which launched just weeks earlier.
What are OpenAI Dots and how do they work?
A Dot is a personal AI agent that uses its own cloud computer to access a web browser and over 4,000 supported apps. You interact with it through a text-message-like interface or via voice calls in ChatGPT on web, desktop, or mobile. Dots also plug into Microsoft Teams and Slack, carrying context from your sessions across devices.
Key capabilities at launch, according to The Verge:
- Cloud computer access with a browser and 4,000+ app integrations
- Text and voice interaction via ChatGPT on web, desktop, and mobile
- Microsoft Teams and Slack connectivity with persistent cross-device context
- Proactive task suggestions: Dots can surface work to complete “before you even think to ask,” per OpenAI
- Collaborative workspace through a new feature called ChatGPT Space
- SMS support described as coming soon
Each user can currently create one Dot. OpenAI says it plans to allow multiple Dots per user in the future, with options to scale each Dot’s speed or the total work it handles per month.
How do OpenAI Dots compare to Meta’s Muse?
Meta’s Muse, which rolled out just weeks before DevDay, can browse the web, make purchases, generate documents, and track goals, per The Verge. The two products share a similar text-message-like chat interface. The main difference OpenAI is emphasizing is control: Muse has already drawn attention for a privacy incident in which a user reported it shared their address with a Facebook Marketplace buyer without their knowledge. OpenAI is positioning Dots with more structured guardrails baked in from the start.
Dots’ safety controls, as described by OpenAI via The Verge, include:
- Built-in rules governing when Dots act without asking you first
- Custom rules users can set to block specific actions or require explicit permission
- An “auto-review” feature that checks whether an action complies with your rules before executing it
- Automatic handoff of sensitive tasks, such as changing a password, back to the user
What can Dots actually do in practice?
OpenAI offered two concrete examples at DevDay. A developer can ask their Dot to pull customer feedback, build and test app updates, then present a video showing the changes. A content creator can have their Dot review an interview transcript, identify clip-worthy moments, write show notes, and draft social posts. In both cases, the Dot messages the user with updates and questions as it works, and the user can follow along by accessing the Dot’s cloud computer directly.
Dots also learn from how you work over time, personalizing output and proactively identifying follow-up tasks. Work with colleagues is possible through ChatGPT Space, the new collaborative workspace OpenAI announced alongside Dots.
Who can access Dots today?
As of the DevDay announcement, OpenAI is rolling out Dots to ChatGPT Pro, Business Premium, and Enterprise users. Conversations with a Dot do not count toward standard ChatGPT usage limits, per OpenAI. The company is also testing a feature that lets organizations create “specialist” Dots configured to handle specific roles inside a business.
Before you connect an AI agent to your business tools: a quick checklist
Giving an AI agent access to your apps, files, and communications is a meaningful step. A few things worth working through before enabling it:
- List which apps you plan to connect and what sensitive data each one holds
- Decide which actions require human approval (purchases, outbound messages, file edits)
- Check whether the Dot’s activity log is accessible and reviewable by your team
- Understand how your data is handled when it routes through the Dot’s cloud computer
- Set your custom rules before the first autonomous task runs, not after something goes wrong
- Assign someone to spot-check Dot actions during the first week
What this means for AI-search visibility
Here is the simplest way to frame it: AI assistants used to answer questions. AI agents like Dots go out and do things, and who they do things for depends partly on which brands they can find and trust.
When a Dot is asked to, say, “research software vendors in category X and shortlist three options,” it is not pulling up a search results page for you to scroll. It retrieves, evaluates sources, and makes a recommendation, often without any human review of what was skipped. That changes which brands surface and which ones do not.
A few observations from Hingewise’s perspective, distinct from what the source reports:
- Agents amplify existing visibility gaps. A brand that is already underrepresented in AI training data and live retrieval will not close that gap by chance. An agent asked to research a category will return what the model has been consistently exposed to and what appears structured and credible. Brands that have invested in AI-readable content and consistent third-party mentions are better positioned when an agent makes that call autonomously.
- The 4,000-app integration footprint creates new surfaces. A brand’s presence inside tools like Slack, its documentation quality in project management platforms, or how it appears in connected business software all become potential retrieval surfaces, separate from the open web. This is a layer of AI visibility that traditional SEO does not address.
- Structured safety controls likely create citation preferences. Dots’ auto-review feature means the agent is evaluating whether a source appears trustworthy before acting on it. Brands whose online presence is consistent, well-cited, and structured are more likely to pass that implicit trustworthiness check than brands with thin or contradictory signals.
Hingewise’s assessment: the arrival of two competing always-on agent platforms from OpenAI and Meta within the same quarter indicates the agent layer is not a future scenario. It is already rolling out to enterprise and professional users. The question worth watching is not which platform wins, but whether the content and brand signals companies have built are structured in a way that agents can retrieve, evaluate, and act on without a human steering the query. That gap between agent-ready and agent-invisible brands will become measurable in the months ahead.
Sources: The Verge (OpenAI DevDay coverage, linked above). Author: Lam Nguyen, Hingewise.
Sources
This article was drafted with AI assistance from the sources cited above and checked against them before publishing. The "What this means for AI-search visibility" section is Hingewise's editorial analysis, built on the frameworks and positions we use in our GEO work. How we work: Editorial policy.